0:02

you have taste in a

0:04

way that's meaningful to software people hello

0:08

, I'm bill gates I

0:12

would .

0:12

I would recommend , uh , typescript

0:14

.

0:14

Yeah , it writes a lot

0:16

of code for me and usually it's slightly

0:19

wrong .

0:20

I'm reminded , incidentally , of rust here

0:22

, rust this almost

0:25

makes me happy that I didn't become a supermodel

0:27

.

0:28

Cooper and Netties .

0:31

Well , I'm sorry guys , I don't

0:33

know what's going on .

0:34

Thank you for the opportunity to speak to you today about

0:36

large neural networks . It's really an honor to

0:38

be here Rust Rust Data Topics

0:41

.

0:41

Welcome to the Data Topics .

0:57

Welcome to the Data Topics Podcast . Hello and welcome to Data Topics

0:59

Podcast Tim , fan favorite friend

1:02

of the pod . What's up , tim ? How are you

1:04

?

1:05

I'm doing good , thank you .

1:06

Thank you .

1:06

And I'm joined for a first-time joiner .

1:08

Yep , yep , sander . How are

1:10

you doing , sander ? I'm good Nervous . No

1:13

, not at all . Oh , not at all Never is

1:15

. He's

1:24

used to the first time . Would you mind , uh , getting

1:26

giving the people a bit of a background ?

1:28

um , I

1:30

vaguely remember that I am 0.81

1:33

sam stall . That's

1:35

maybe a bit of a throwback . I

1:38

don't know if you do that anymore it

1:40

was hard to keep up . It was hard to keep up okay

1:42

um , yeah , I to keep up , okay , yeah , I'm

1:44

Tim . I work in

1:46

the commercial department of Data Roots , which

1:48

means that I am today

1:51

. You're not going to hear the very technical statements from me or

1:53

the very technically accurate lingo . I'm

1:57

here as resident NBA expert . I've

2:00

been informed by Murillo , so I

2:03

have a big passion for sports analytics , especially in

2:05

basketball . So , um , with

2:07

the addition that we're gonna have , I hope

2:09

that's gonna be a relevant experience I'm

2:12

sure you will , I'm sure it will .

2:13

And sunday , yes , for the people that

2:15

don't know you yet um , is

2:17

there people that don't know me

2:20

but the ? People like any life updates

2:22

, you know just love that the people your fans are like manander

2:24

. Tell me what happened with you . Like what's up .

2:27

No , no . So I'm Sander

2:29

. I'm a

2:31

data engineer as well as the

2:33

team lead of our platform team , but I'm

2:36

also a big sports fan . I play football . I

2:38

watch football like every weekend , so that's

2:41

why I got involved .

2:43

Yes , yes , yes , yes . I feel like you guys are

2:45

saying like , as I snatch you from the

2:47

hall you gotta sit , you

2:49

gotta eat your broccoli .

2:50

You gotta , you know , you gotta talk to me about sport

2:53

.

2:53

Maybe a bit too far , but so

2:57

, yeah , this is the Data Topics Unplugged . The summer

2:59

edition , let's say so

3:02

we do have a bit of a different format . Um , the

3:05

euro just finished , right , the euro

3:07

2024 , the

3:09

final big final was yesterday was

3:11

, uh , spain and england

3:13

. Maybe I'll just put on the screen here the brackets . Um

3:17

, and I want to say maybe

3:19

how much , alex

3:21

, actually do you remember how much time ago was the other

3:23

euro predictions was

3:25

like three weeks ago , yeah , maybe a month or something

3:28

. So I thought it would be interesting to kind of

3:30

revisit and see how how

3:32

these ai predictions do . Also , the

3:34

plug with nba , because the team's here , so

3:36

we need to talk about me too . I just came um

3:39

, but yeah , well , like is what , did

3:41

they get it right ? Did they get wrong ? So I was a bit curious

3:44

in hearing your thoughts . I know sunday you also had some thoughts

3:46

, and tim as well .

3:47

Um , the first , congrats to spain , big

3:50

winners the the true mission of the data

3:52

topics podcast is holding random medium articles

3:55

accountable for their predictions that they're making

3:57

, finding those guys .

3:58

Oh , you said that , get

4:00

him , cancel him . Uh , no , that's

4:02

not the point , but I do think that predicting

4:05

uh scores

4:07

in sports I think it's an interesting

4:10

challenge . I think some sports are more challenging than

4:12

others . I also think that's . Uh , there's probably some

4:14

interesting reasons there , and even the more

4:16

philosophical is like do we want to make it easier

4:18

to predict which ? We'll get to that in

4:20

a bit so cool . So

4:23

, uh , maybe to start here

4:25

, uh , like I mentioned

4:27

, we did have some predictions already

4:29

, namely , we had one prediction

4:31

from uh . I want to say snowflake

4:34

, using snowflake ml . Let me see if I can find the

4:36

link . They

4:39

had predicted , so I'm putting on

4:41

the screen as well and

4:43

we'll share the links as well , as usual . They

4:46

had predicted . I think I want to say Spain . No

4:49

, not .

4:49

Spain , France , England .

4:51

But , Spain .

4:53

it was either France , portugal or

4:55

England that was

4:57

.

4:57

France , portugal or England .

4:59

Yes , so yeah , this is it .

5:04

There we go and

5:06

hands-on match . I have a Portugalugal won . Friends have come through several times in testing

5:08

and if I had to pick a top three winning team winners would be england

5:10

. Friends , maybe before the competition started would

5:12

you have agreed with this prediction or no ? Probably

5:15

yeah , yeah , you think portugal

5:18

would .

5:18

Uh , you were very optimistic about portugal as well I

5:20

mean , I predicted the final to be england , portugal

5:23

really oh , wow if

5:25

you look at , if you look at the pure quality of

5:27

the players that they have and where they play

5:29

which teams , then it really makes

5:31

sense for portugal , portugal .

5:33

They have a great roster maybe

5:35

I'm just out of touch .

5:37

Who's uh the , the portuguese uh

5:39

stars , I guess but you

5:41

have bruno fernandez , you have bernardo silva

5:43

, you have stars

5:46

, I guess , but you have bruno fernandez , you have bernardo silva , you have leo from is

5:57

I

6:06

mean these heroes .

6:07

We got the record of the oldest outfield player as well as the youngest

6:09

one . Yeah right , that's crazy . Yeah , so , uh . And the youngest one was the yamal

6:11

yeah , and his birthday was on saturday

6:13

really but so he became 17

6:15

years old .

6:16

On saturday , sunday , he became the

6:18

youngest I promise you , I have a list of , like

6:20

random statistics right , all right . Tim , take us home

6:22

he also like as a as

6:24

a 16 year old . He also made another record

6:27

. He is the tied uh

6:29

single tournament

6:32

biggest assist giver , as in . He gave

6:34

four assists in the in the entire

6:36

championship and it's the . It's the

6:38

highest for spain ever and it's

6:40

the highest single tournament , tied with

6:43

others and I don't know the others , but

6:45

it's crazy oh wow , he's

6:48

also the youngest , youngest scorer ever , the

6:51

youngest in everything , basically yeah , 16

6:53

the goal against France

6:54

so if he was born , he's 16 and his birthday

6:56

he just turned 17 . That means he was born in . Actually

6:58

, I don't need to do the math , don't look it's gonna be painful .

7:03

Declan Rice . In his in his interview

7:06

before the game he he said something

7:08

really funny . He said like oh yeah , when COVID started

7:10

, this guy was 12 .

7:13

He was born in 2007 . So it's crazy

7:15

man .

7:17

But did you see like he

7:19

always got sucked in , like the 80th

7:21

minute ? Did

7:27

you something like the 80th minute ? Did you see why ? No , why , like um ? In

7:29

germany it's like , uh , prohibited for like someone who's under the age of 18

7:31

to work past the hour of 11 o'clock in the evening , so

7:34

that meant like yeah , yeah really

7:36

so he had to be taken off because he had

7:38

to shower before 11 o'clock really

7:40

otherwise he was breaking the law it's a lot

7:42

.

7:42

It's a lot , it's like it's a lot against child labor

7:45

in germany , so if the year

7:47

wasn't in germany and they had a different

7:49

law in the other country ?

7:50

or

7:52

if it would have been you know more strict in

7:54

another country , when , but they would have

7:56

taken the fine . Probably it was .

7:58

It was a 30k fine , yeah still

8:00

yeah , if you're up 3-0

8:03

, it doesn't matter so let's see popular , popular

8:05

2007 movies . Spider-man

8:07

3 that's funny , that's the

8:10

first one . The Bucket List yeah

8:13

, actually , alex , do

8:15

you know these films that I'm saying , or no , it's just so . Yeah

8:18

, okay , the Bee

8:20

Movie .

8:21

The Bee Movie . Yeah , okay .

8:24

The Underdog . Yeah , there's some uh resident

8:27

evil extinction . All these movies were uh

8:29

came out . Yeah , mr bean's

8:31

holidays , you

8:33

know this one we

8:39

put this up .

8:40

We put this up for the next like hour and a half .

8:41

Yeah , we're just gonna list movies from the rest of the podcast

8:44

. Actually . No , no

8:46

, but indeed so . Yeah , super young kid , it's crazy

8:48

. Like what was I doing at 16 ? I was like

8:50

, hmm , maybe . Actually I know

8:52

an anecdote . I think it was by Leverkusen

8:55

. They also had a very

8:57

young player . He was also in

8:59

high school and I remember they announced on Twitter maybe I can

9:01

try to find it it

9:05

was a Champions League match and they were put

9:07

on Twitter like who's playing and who's not playing

9:09

, who's out because of injuries , who's out because of yellow

9:11

cards and all these things . And they're like oh , this guy

9:13

, they're recovering from injuries . This guy , they received a third

9:15

yellow . This guy's got a red card . And then

9:17

this guy , he cannot go

9:20

, has an important exam with the school the next day .

9:21

Like really , I

9:24

think at Mala as well , when he was still

9:26

in training camp , he was making homework

9:28

can you imagine he cannot be in the Euro

9:30

final because the teacher gave him an

9:32

exam ? The next day his grades were not good enough

9:34

.

9:35

He was grounded by his dad yeah , it's

9:37

like you didn't do your chores .

9:39

Yeah , yeah , florian

9:41

, florian Wirtz , yeah

9:43

ah , he's pretty good as well right now yeah

9:46

, yeah , yeah , so uh , school exam .

9:48

Let me be sure this is dead he

9:51

wasn't that good in the championship yeah

9:53

, but it says school exam made 17 year old Florian Wirtz miss

9:56

Leverkusen's Europa League clash . Yeah

9:58

, it's funny , huh , I

10:00

mean it's funny , but but it's

10:02

good In Brazil that wouldn't fly In Brazil

10:04

. You'd be like you're probably not in school anyway , if

10:08

you're serious about football , you're probably not in school

10:11

. Okay , so this is one . So actually

10:13

you put England and Portugal

10:15

. And actually we can check it out . No , because at Data

10:17

Roots I think we did a little predictions

10:20

. Yeah , we did yeah

10:22

but I haven't even checked .

10:25

I should , yeah , yeah , we did , yeah , but I haven't even checked . I should have

10:27

checked before we record . But the thing is like with the , with the data roots one

10:29

, I think you could like uh , every round of

10:31

the , it progresses update or

10:33

you just have to fill it in by then . I see

10:35

, because the the pro ?

10:36

no , I did , I really had to predict the whole yeah

10:39

, yeah , yeah , but

10:41

I think it's like so well , we're

10:43

talking about the euro , right ? So the way the Euro is set up

10:45

, there are different groups and then the

10:48

. I think the first two of each group goes

10:50

through , but then some third places also

10:52

go through . Something like this no four out of six yeah

10:54

so I guess it's like , yeah

10:56

, predicting the actual matches , like who's

10:58

gonna play against two , and the outcome is super

11:01

unlikely or even the

11:04

turds , like the turds , like

11:06

, it depends on how many goals they scored and stuff so

11:08

it's impossible to have them . Ah , maybe also

11:10

talking about the group stage . I think this euro I think

11:12

was the first time ever that after the second round

11:14

, uh , all the teams in the group were

11:17

tied right belgium , romania

11:19

, slovakia and ukraine yeah

11:21

, they all had the same . I think was the first time ever . As

11:23

well , they have four points right because they uh , they

11:26

know after your third .

11:27

Even right was after

11:29

the third . It was an entire group

11:32

stage in our in our group . Yeah , the belgian group

11:34

was just . Everybody was tired at the end because

11:36

I was .

11:36

I was explaining to my partner , like the the , how the

11:38

who's gonna classify because she's romanian

11:41

, and I said she was actually the second , second time , I think

11:43

, romanian , or maybe first time since 2000

11:45

and something that Romania actually goes through and I think

11:47

maybe the first time that they go to the next phase . So she

11:49

was super like oh yeah , we're gonna go through , we're not

11:51

gonna go through , ah , but the points are times , ah , and

11:53

I was like no , there's a system , you know

11:55

usually goes this , then

11:59

it goes cards . How many this ?

12:00

there were two teams that like the deciding

12:03

factor between them was like a qualifying

12:06

game they played and there even

12:08

there was tight and it was something like it was

12:10

. They had to go to , like criteria number six

12:12

.

12:13

But in the end they didn't go to that because

12:16

one of the teams got a yellow card

12:18

in the 90th or something yeah it's

12:20

crazy , it's crazy .

12:22

Um , but yeah , who did you ? Who did you have

12:24

? Because you mentioned England and Portugal

12:27

, you had it on the final . Who did you have ? Tim

12:29

, or I don't know if you actually picked anyone

12:31

, but before the tournament started .

12:33

I didn't do Prono on the bracket , but

12:35

my guess was just France .

12:38

Just France , france Okay .

12:45

Because just the way that they were able to and they showed it as well during the championship

12:47

the way that they were able to and they showed it as well during the during the championship , the way that they were able to like ice games they , they do that

12:49

so good . They weren't like Spain

12:51

was just better , but just in the

12:53

mentality that they have in winning games is

12:55

the way that they become became world champions

12:58

.

12:58

Yeah .

12:58

Still like in terms of talent .

13:00

Indeed , I actually saw maybe a sure

13:03

Another time talent

13:06

. Indeed , I actually saw maybe uh sure another tab . Now , another

13:08

prediction was

13:14

from the kill love and uh research center , right , um , yeah , just kind of skipping a bit through . I think we covered

13:16

this last time as well . They had germany and friends on the final and friends winning

13:18

. So uh doesn't really disagree with what you were saying

13:20

, tim . I , before the

13:22

tournament started , I thought friends . I thought

13:24

england actually and I thought germany . Those are the three . I don't . I were saying Tim

13:26

, before the tournament started , I thought France , I thought England actually .

13:28

And I thought Germany .

13:30

Those are the three . I don't follow football as closely these days , but

13:32

Spain , I think , definitely surprised me . I think also , when

13:34

you look at the roster

13:37

as well , it's like they

13:40

have some names that I know , but a lot

13:42

of them are really young . A lot of them they play

13:44

really really well together . Yeah

13:46

, and I think one thing also that I hear from um

13:48

these national competitions

13:50

, that it's way

13:54

harder because the players don't train together the whole

13:56

time , right . So I think sometimes , like , even if

13:58

you have big names but they , even if

14:00

you have like five big names , but in their

14:02

current teams the people are set up to play for

14:04

them .

14:04

Uh , that wouldn't happen as well

14:07

on the national team I think what you see as well

14:09

is that , um , especially with

14:11

france , has become really clear in the tournaments

14:13

that you in

14:15

in club football , you have like teams that

14:17

are just really constructed as

14:19

in you really like by by

14:22

buying players here and there and then training

14:24

some players here and there , you get a team that

14:26

is usually they're more balanced . And then

14:28

in these kinds of teams and France

14:30

, for example , in their midfield they had like three , six

14:34

, eight players like not a

14:36

creative kind of midfielder . That

14:39

was very difficult . Only when Griezmann dropped back . Then

14:41

you saw that and I think that's

14:43

in general as well what you see in international football . Because

14:45

you cannot sort of oh yeah , we want Ronaldo

14:47

to be a Belgian , no , no , you're

14:49

stuck with the players that have your nationality

14:52

and then you get way more unbalanced

14:54

teams where you have to make a distinction . Kamavinga

14:56

is a really good player , chouamini is a really good player , kante

15:00

is a really good player , but they're all sort of the same

15:02

profile , like these breakers , defenders

15:04

, um , and then you don't have the creativity

15:06

and like I think it was at

15:08

some point in philippe osa said it , one of the commentators

15:10

he said like the coach is really trying

15:13

to do the best he can , but he just doesn't have . You

15:15

don't have the players , you don't have the same balance

15:17

as you're doing club football .

15:19

It's not like I think in club football you can kind of

15:21

say I want my team to work like this and

15:23

then you can find the right pieces . And I think

15:25

in national teams is a bit the opposite . Like this is what

15:27

I have . What can we do ?

15:29

yeah , I also think the thing with national teams

15:31

is you kind of feel pressure

15:33

to pick your best players right . Yeah , it's difficult

15:36

to say like , okay , this guy I saw it with

15:38

england yesterday like kane , he

15:40

scored like 40 goals at

15:42

bayern this year and he's playing like super

15:44

deep and I'm like play him to his strengths

15:47

right , but then or don't play him .

15:49

Yeah .

15:50

So it's difficult to pick

15:52

those players which fit your system if

15:54

that means dropping , maybe , a big name

15:56

player .

15:57

Yeah , we saw it a bit Belgium as well

15:59

. Yeah , like Opelna fit way better with the system

16:01

at Lukaku you play him , indeed , indeed , way better with the system at Lukaku when you

16:03

play him indeed , indeed

16:05

, I think

16:07

we mentioned he feels sometimes

16:10

not super objective .

16:11

We're going to touch a bit more on that as well . Because

16:14

Lukaku , I think one thing that I I

16:16

always feel like he makes a presence

16:19

on World Cups , but

16:21

during the rest of the year I never hear

16:23

much about him . I feel like now he's an inter , but I think last World Cup and but during the rest of the year I never hear much about him . You know , I

16:26

feel like now he's an inter , but I think , like

16:28

last World Cup , he he's

16:30

a Roma . He's a Roma , but he was

16:32

an inter before .

16:32

He was an inter the year before .

16:34

Yeah , he left there in really nice terms

16:36

but I also

16:38

think there's also like some players like Kieb , I

16:41

think there's some goalkeepers as well that

16:43

they go crazy well , and

16:45

I also think there's a bit .

16:47

Pickford is a clear example . You don't hear anything

16:49

from him when he's at Everton , and then

16:51

he's playing for England but to be honest , even

16:53

for England I was like man , he's

16:56

an okay goalkeeper , right he's

16:58

funny but

17:00

I think he's like , I think

17:02

, in 2014 , the US goalkeeper .

17:04

I think Howard's like , I think , in 2014, . The US goalkeeper .

17:06

I think Howard .

17:06

Howard yeah . He did crazy , good , crazy . The

17:08

Mexico goalkeeper as well , ochoa

17:10

, he was doing super . But , like during the year , we don't see as

17:13

much and I do think there's a bit of the can

17:16

. These players , how

17:19

do they deal with pressure ? Yeah Right

17:21

, and I think sometimes it's a bit hard to say okay

17:23

, you have this guy that is a star , or even like lukaku

17:25

, but he has a track record like maybe this other guy's playing

17:27

well , but maybe for a very short-term competition

17:29

, maybe he's going to do super great , you know . So it's a bit

17:31

unpredictable . I think it's very hard to to predict

17:33

these things .

17:35

Um , in the in in the nba

17:37

they call these like floor razors or ceiling

17:39

razors , like the , the people

17:41

you have . You have people that can really play exceptionally

17:44

at a high level , but not always that consistently

17:46

, and they're like ceiling raisers , like at

17:48

the top . They can make your top better . They can

17:51

make the absolute best of your team , make

17:53

it better . And then you have the , the sort

17:55

of floor raisers . I think conte is a very

17:57

clear example of like a floor raiser . He's

17:59

always going to be super consistent . You know

18:01

what you got in him and he's just going to make the team

18:03

the worst your team can be . He's going to make

18:05

it better , like you're saying . You're not going to be worse

18:08

than that just because he's there , but he's

18:10

not going to make you know the best franz better

18:13

. He's going to be an essential part of it

18:15

, but he's not going to make it better . That's where mbappe

18:17

comes in . He's a see and

18:19

I think you have . You have these

18:21

kind of players that can be because of their

18:23

unpredictability

18:25

. They can be great and they're going to be ceiling

18:28

raisers for a specific tournament . They're going to make it

18:30

because everything , like the coincidence

18:32

and just the great shape and whatever

18:35

comes , they're going to be the ceiling raisers .

18:37

And do you , can you quantify ? Like , how are

18:39

? Can you quantify ? I mean , you mentioned NBA

18:41

and we talked a bit before we

18:43

started recording that NBA in general has way

18:46

more statistics . Um

18:48

, is there like a metric or something

18:50

, then ? Be it they use , or is

18:52

it more like feeling ? Or how do they actually say

18:54

this guy's a ceiling razor or the

18:57

razor ?

18:58

in the in the nba way more than than in

19:00

, because , especially now

19:03

in the NBA , there is not as

19:05

much position anymore . There's

19:07

a very large shift towards

19:09

position of this basketball . So you can , they're

19:11

trying to have metrics that sort

19:14

of represent the impact of people and then , and

19:16

what you can see , like there's metrics that

19:18

are like there's a metric called LeBron

19:20

. The metric is called

19:22

LeBron because it's , you know , it tried to capture

19:25

what the value was of lebron versus michael

19:27

jordan , and

19:29

this metric is used as sort of and what

19:31

is the ? But there's also box plus , minus , whatever

19:33

, like there's different metrics that try

19:36

to represent the individual impact the player

19:38

has , which is very difficult in soccer

19:40

because you have the positions and different positions

19:42

already have a different impact . So , um

19:44

, but there you could . You could just look

19:46

at the , the , the

19:48

impact this person has in this

19:50

, this , in an individual game , and then

19:53

the , the variability on that , like

19:55

if , if there's a on a specific

19:57

game , it's like super high , and but

19:59

there's also super lows , and then you have players

20:02

that have a very steady impact because they're all in

20:04

basketball . Then you have , like , players like drew holiday

20:07

, who's a very just , like , similar

20:09

to conte , defensive , disciplined

20:11

, does all the right things , all the

20:13

dirty work . They , these kinds of players

20:16

, super valuable and you can see it in any , any

20:18

kind of sports . You always need this , but

20:20

that's the way you could . You could try to measure

20:22

. I'm not sure if it , if it is done , I don't know

20:24

to that extent if , because it doesn't

20:27

hold a lot of value , I think to teams to

20:29

classify somebody as a ceiling razor

20:31

or a floor razor , but that is the way that you would

20:33

measure it in basketball okay

20:35

, cool , maybe .

20:37

Um , now there are different ways to measure , like

20:39

players . Now we're talking about individual players

20:41

and I want to move on there , but but just before

20:44

we do , I just wanted to wrap up the discussion

20:46

on the predictions . We

20:49

also internally at Data Roots , we're also commenting

20:51

a bit on this . One thing that

20:53

someone mentioned I think it was Nick , another

20:56

data engineer . He was mentioning that the

20:58

predictions seem very much aligned with how

21:01

expensive the players were . And then when you

21:03

look into the actual um predictions

21:06

right , so the first thing they said is the elo rating

21:08

. So the elo rating , I think , is like each team

21:10

in this case has a score and then if

21:13

you have a high score , you have a expectation

21:15

that you're going to win , I guess um

21:17

, and if you don't win , then I guess you

21:19

lose a lot of points . But if you confirm that expectation

21:21

with a little bit of points , it's a bit like these point systems

21:24

. So they actually compute the ELO rating

21:26

for the recent international match

21:28

results , which I think makes sense . Then

21:31

they have individual offensive and offensive ratings

21:33

. So

21:36

teams go score and consider in recent games and

21:38

then , yeah , basically it's not easy , but etc

21:41

, etc , etc . And not

21:43

so surprisingly , they include the cumulative market

21:45

value from transfer market . So

21:47

I

21:49

don't know how much this weighs in in

21:52

the predictions , right , but

21:54

I mean it would be interesting

21:56

to see a breakdown , right , how much that contributed to

21:58

this prediction . But uh , there's .

22:00

There's such a strange bias

22:02

there in the cumulative market value . Like a

22:04

player like modric is a clear issue

22:07

with that . Like modric

22:09

is not going to be a valuable player at this point , he's what ? 36

22:11

, 37 ? His value , his

22:14

value on something like transfer market

22:16

is not going to be a lot because a player , a team , is

22:18

not going to pay 80 million anymore

22:20

for modric because he's that old . But

22:22

in terms of impact , if you were to able

22:24

, if you're able able to measure sort

22:26

of the individual impact of a player I don't know if there's

22:29

metrics in that in soccer , um

22:31

, like the , the forward passes

22:34

, the assists , the crosses , the

22:36

accuracy of passing , all these kind of things , sort

22:38

of combined into a metric , modage

22:40

would be super high in that .

22:42

Still , yes , maybe , uh

22:44

, I also . We also brought it . I didn't put

22:46

it on the list , but you mentioned like a metric

22:48

. One thing that I've seen that I thought was pretty

22:51

interesting is also from the cure live in . Uh , research

22:54

, which is the vape , is the valuing actions

22:56

by estimating probabilities , and I think it's somewhat

22:59

similar to what you were saying . So , like , um

23:02

, basically , they compute

23:05

the probability that an action contributes to a goal

23:07

, right ? So I think the example

23:09

here and I have some video is that

23:11

there is a people on the baseline and

23:13

then they pass sideways , basically , actually

23:16

maybe I don't know if this is the best visualization here um

23:19

, maybe this is better . So they

23:21

have , like , let's say , a four-action

23:23

sequence . There is a pass

23:25

that is sideways from the defenders , right

23:28

, and that's arguably a very easy pass

23:30

. So in statistics sometimes they just say that's a

23:32

pass completed , but that's a very easy pass

23:34

. And then there's a through ball from two to three

23:36

that puts someone in

23:38

the edge of a corner . So that

23:40

pass is also just a completed pass , but

23:45

it creates a lot of probability of scoring a goal . And then actually there's a cross in the middle

23:47

and then it's a shot . So then , once they

23:49

know that it was a goal . They kind of back

23:51

, propagate and attribute the value back , and

23:54

the idea is that they try to do this with every action

23:57

in the game . So I think , even

24:00

in the presentation we saw from Kiel

24:02

Levin , actually they were showing

24:05

how , just

24:07

using the quote-unquote simple statistics

24:09

, you see a lot of players' value going high

24:11

. But then once you apply this framework , like

24:14

Kevin De Bruyne was super valuable because

24:16

the actions that he does , even if he doesn't

24:18

have this high percentage of completed passes

24:20

, the passes that he does complete

24:22

are super valuable , right . So I thought it was a very

24:24

interesting and a very data-driven

24:27

way of assessing these things .

24:29

Yeah , and actually maybe , while I'm talking about

24:31

this , so yeah , maybe

24:33

first question I have is does this , does

24:36

this have any way to represent off-ball

24:38

actions ?

24:39

I think so it depends a lot , and I

24:41

think it's a bit intrinsic to football that

24:43

the way that they gather data is literally

24:46

someone watching a game and annotating these things

24:48

right , uh . But even if you say a pass

24:50

is a pest where there's

24:52

a lot of information you need to capture , right , if you manually

24:54

annotating um , I don't

24:57

remember if they have uh

25:00

off ball actions , maybe they do

25:02

I'm just

25:04

saying like there's there , were there used to be players

25:06

.

25:07

I'm not sure if you you still have the the archetype

25:09

, but like the alessandro del piero archetype

25:12

, where you have somebody who's just always

25:14

the right time and the right spot

25:16

to and and it's a skill , right yeah

25:19

to read spaces and but

25:21

it's very difficult to put into a mathematical model

25:23

yeah , I agree , I also feel like I

25:25

feel a bit about that , about ronaldo , to be honest now

25:28

ronaldo .

25:28

Yeah , because I feel like a lot of yeah

25:30

I think , yeah , in the beginning of his career he was way more of a

25:32

playmaker and whatnot . But I think the

25:34

the years that he was scoring a lot of goals . A

25:36

lot of the goals were like tap-ins or headers or something

25:39

, and a lot of times it's like

25:41

, yeah , okay , he was just tapping it in , right , but

25:43

at the same time , to be in that position to do these

25:45

things like there's , it does take a lot of

25:47

judgment calls and it's not easy . Yeah , no , uh

25:50

indeed . So let me see if they have .

25:52

Uh , I'm not

25:54

sure , tim maybe I remember

25:56

it's also difficult to create these

25:58

models and like it's very arbitrary , like what do you

26:00

include into a model that sort of represents

26:02

the impact of the player ?

26:04

yeah , and I also think that every time you make a decision

26:06

like that , you're also including your bias

26:08

in a way . Yeah , right , because you're saying that this

26:10

well , this feature , you don't include . So

26:12

even if it is valuable , you're not .

26:14

Uh , most of my thinking on this subject

26:16

is like shaped by the thinking basketball

26:18

podcast and and they do

26:20

a very interesting thing there where they sort

26:23

of rank players according to their impact metrics

26:25

and then you look into does

26:28

this make sense ? Like , does this make sense ? Does

26:30

the eye test , if

26:33

we ask 20 people to rank these players

26:35

, does it hold true

26:37

to the impact metrics that we see ? And then you

26:39

often see outliers and you're like , okay , is

26:41

this player ? Does

26:44

he or she , do they have the

26:46

narrative against them because of some things

26:48

that happen off the court , which happens as well

26:50

. Like , is that's why that ? Why people like

26:52

them less , is um , or

26:55

is this some ? Is there just

26:57

a part of the game that we overrate ? Like

26:59

somebody who scores a lot , is that a good player

27:01

or is the person who provides them with the pass

27:03

that a good player ?

27:05

and yeah , yeah , and

27:07

I do think in football the build-up for a goal

27:09

is well

27:12

, yeah , it's complex , it's really complex , it's

27:15

really quote-unquote long right , I think in basketball

27:17

it's more . It's still somewhat complex , but it's

27:19

still shorter . If you look at baseball , it's

27:21

even shorter , you know like , and

27:23

I think that's , and I think that the shorter it is , the

27:26

more direct or

27:29

less complex in a way , the analytics

27:31

. Like if you look at money , ball and all these things , you

27:33

know it's , you can almost

27:35

like , you can almost not , you can almost be a

27:38

scout and not watch any

27:41

match , you know

27:43

what ? I think that's a bit unthinkable for

27:45

football Right yeah . You

27:48

don't like baseball , do you ? That's what I was going to say .

27:50

But yeah , it happens and they

27:52

try to apply the monoball principles to basketball

27:55

. I think today you could

27:57

not apply them to soccer , like

27:59

not at all . You could try and , you know

28:01

, adopt some of them and try to be a bit more data-driven

28:03

, but today you cannot run soccer

28:06

by no means . And there's a dimensional problem

28:08

as well . We talked about it during lunch with uh

28:10

, with yonas , soon as well , shout out yonas

28:12

, um , but uh

28:15

, like there's on baseball

28:17

, you have basically two people that are

28:19

involved in an action . Like there's a pitcher and there's

28:21

a one that you know hits it . I

28:23

think there's the most important people . You

28:25

got all the outfielders . I don't know what . I'm not a baseball

28:28

expert , but like in basketball you have 10 players

28:30

and in soccer you already

28:32

have 22 .

28:33

Like there's a dimensionality problem , like the amount

28:35

of coordinates you have to track , and it's true

28:37

. But then what about for sports like tennis , for

28:39

example ? There's a there's a lot of analytics

28:42

in tennis there is right , but it's not as

28:44

much as baseball , I would say um

28:47

or maybe the

28:49

reason why maybe I don't hear as much is because

28:51

tennis is still an individual sport , right ? It's not like

28:53

you're going to recruit someone for I gave a

28:55

.

28:56

I gave a presentation on that a couple of years ago on roots

28:58

conf . There's a . There's a number of factors

29:00

that play into the analytical

29:03

mindedness of a sport . You have the

29:05

amount of people , the size of the court , which

29:08

sort of , and the how

29:10

discreet the sport is

29:12

. Baseball is super discreet sport

29:14

. It's not continuous , it's discreet because

29:17

you have like action , stop , action , stop , action

29:19

, stop , action , stop . Tennis is already

29:21

more fluent . You have like a number of hits and

29:23

it's a . You don't

29:26

know how many hits , like how many times

29:28

the ball is going to get kicked across

29:30

the field . Soccer is a super fluid sport

29:32

like you can . You can have 15 minutes of gameplay

29:35

without stoppage yeah true , without the

29:37

ball ever going out , so you cannot divide

29:39

it into individual parts , which makes

29:41

it more difficult to model the

29:43

game . Um , that's

29:46

why that's why you basically you can see

29:48

, like the amount of teams in in

29:50

america , like baseball is almost

29:52

all the coaches are do have an analytical

29:54

background and have a heavy

29:56

scouting team and then , like it drops off , like

29:58

um basketball and and

30:01

and ice hockey . They sort of have similar

30:03

complexities to the game , similar amount of players

30:05

, similar gameplay , fluidity . Then

30:08

it drops down to american football already , although

30:10

with the American football , with the fantasy football , yeah

30:13

huge increase .

30:14

And then soccer you barely see any so

30:17

are you saying that predicting NBA

30:20

is going to be easier than predicting football ?

30:23

soccer football yeah

30:25

, just want to make the distinction .

30:27

Uh yeah , predicting nba is going to be easier because I

30:29

think so too I

30:32

know

30:34

right it's like I didn't

30:37

find this this morning at all um

30:40

, predicting to be a champion with machine

30:42

learning . So this is also medium post right

30:44

. So I looked I

30:46

maybe I should . I mean , I wanted to look more

30:48

into it , but what I have gathered

30:50

is for the last four years they made some . I have gathered is for the last four years they

30:53

made some predictions and they actually got the last four

30:55

years right , but not only that , they were able to predict

30:57

the eight out of the last ten NBA

30:59

championships in the last , well , eight

31:01

out of the last ten years right . So

31:06

to me this was a , if it is possible

31:09

, I would imagine , even

31:11

though so they read , if I

31:13

correct me , if I'm wrong here , tim . But I think that the way then

31:15

they works , there's like two conferences

31:17

right , the east and west , and I think in the beginning

31:19

everyone plays against everyone or no . Everybody's

31:21

everyone against everyone , and then the the top of

31:24

each each or west the conferences

31:26

. So there's still like the no , actually

31:28

no .

31:30

So everyone plays against everyone , and then they have like a knockout

31:33

stage you play , I think

31:35

, um , but don't quote

31:37

me on that one you play four times within

31:39

conference , uh , against the team you know

31:41

inside your conference , and two , two times

31:43

against the team outside your conference , something like

31:45

that , um , in the regular

31:47

season . And then you have the playoffs , which is a knockout

31:49

stage within your conference , which ends

31:52

up a final , the winner of the

31:54

you know , the Western Conference champions against the

31:56

Eastern Conference champions .

31:57

So there is , there is still like there is a

31:59

points part , but there is still like a knockout

32:03

part . Yeah , right , which , as we mentioned

32:05

before I guess I don't know if you mentioned before I definitely mentioned

32:07

before the recording that I think that's also

32:10

something that is super , like the an outlier

32:12

, someone that is not doing good , like

32:14

maybe the striker is not feeling well , is sick , or something

32:16

. It has a huge impact for

32:19

euros , especially because it's a one match knockout

32:21

. So

32:23

in nba there's still a knockout , but the person was

32:25

still able to predict fairly accurately

32:27

, right . And one of the reasons why I hypothesize

32:30

and that's what I wanted to hear from you is that the

32:32

knockouts in nba's

32:34

are seven games right . So

32:36

even if there is an outlier there , it's

32:39

still diluted a bit across the , the multiple

32:41

they need to win , for a team needs

32:43

to win four times at least to go through

32:45

.

32:46

Yeah right , yeah , so you have , you know , more

32:48

statistical relevance . For starters , there's

32:50

also another trend trend in there that is going to make

32:52

it very like very

32:54

much , and I'm wondering how applicable

32:57

this is to soccer , for example . One

32:59

of the trends that you see the last couple

33:01

of years in NBA basketball is the fact that the

33:03

game speeds up , as in there's more possessions

33:06

typically per game . They

33:08

go for shorter possessions , more possessions , and it's a , it's

33:11

it's statistical reasoning . Um

33:13

, the better teams want to , you know you

33:15

want to up the tempo because

33:17

the more possessions you have it's a lot

33:20

of large numbers the more possessions they have , the

33:22

more likely that the best team is

33:24

going to win . So you try to up the uh

33:26

, the tempo you you really like as a

33:29

, as a good team , you really want to up the tempo . And then

33:31

there's sort of the , the

33:33

thing where you know good teams are built to have

33:35

a lot of good possessions and then the less

33:37

good teams start to look at the good teams , to how they

33:39

build their teams , and then at some point everybody

33:42

starts . You can see the pace increasing

33:44

, and so the offensive rating goes up .

33:45

When you say pacing , you literally mean the amount

33:48

of seconds . For example , it

33:50

would always go down to the amount of

33:52

seconds it takes for the

33:54

possession to turn in basketball yeah , they measure it in

33:56

amount of possessions per game . Oh , wow .

33:59

And it goes up steadily for a

34:01

long time .

34:03

For any team .

34:05

Average . There might be a team sometimes that

34:07

slows it down because they specifically have a

34:09

type of basketball that they want to play

34:12

because they're not as good , so they want to

34:14

slow it down . Yeah it happens

34:16

, it happens , uh , the lakers in 2020

34:18

. They were a very slow paced team because

34:21

they had a very good

34:23

interior defense , so they wanted to sort

34:25

of constrict it every time to a half

34:27

court basketball , where this

34:30

, you know , the , the , the features

34:32

that they had came out really well . Oh

34:34

, wow , like , um . But because

34:36

of this , like , the statistical

34:38

models that you have will become more

34:40

relevant because you're sort of already

34:43

, by increasing the pace , you're averaging

34:45

out the noise yeah , a bit fair

34:47

enough so each game becomes statistically

34:50

more relevant .

34:51

So as a whole , the predictions

34:53

should also be more easy to make

34:55

, because there's more signal compared to the

34:57

noise as a machinery engineer

34:59

, maybe also draw a bit the parallel with football , because in football

35:02

, like you know

35:04

, I mean it's a bit I don't

35:06

know how relevant this is today or how

35:08

accurate this is today , but I know

35:10

that , like , maybe a

35:12

few , like four years ago , whatever the

35:15

Spanish way you know , the tiki taka , high possession

35:17

and all these things was really the scene as the the

35:20

top , you know , which is like the opposite of

35:22

what Tim is saying because you have less turnovers

35:25

, right , but it's like I

35:28

think football is different because there's like there's no

35:30

clock running down right , so as long as you have the ball

35:32

, that's true that's true

35:34

, like , yeah , you see what you're saying right , because

35:36

actually I heard that argument as well , that

35:38

one of the reasons why possessing the ball

35:40

is so good is because it's also a way of defending , like

35:43

if the team doesn't have the other

35:45

ball .

35:46

You said the clock runs , right

35:48

yeah , um , but

35:50

I the reason why I wondered and

35:53

I wanted to look this up , so I went into the , into

35:55

the statistics that you can find on the

35:57

uefa website and

35:59

for the current european championship , you can find

36:01

the amount of possessions and that

36:03

the teams have , and I think I found it here

36:06

, um , so there's

36:08

, for example , spain had 123

36:10

attempts on goal , versus England

36:12

75 only . But the final

36:14

no no , just in general in general

36:17

, the entire tournament but

36:22

Spain , for example , also had 411

36:25

attacks where possessions

36:27

, whereas England only had 344

36:30

. So on the same number of games they had like 70

36:33

possessions less , which

36:35

means that on average each game that's 10

36:37

possessions less . They played

36:39

7 games and I

36:41

wonder . I don't know if there's

36:44

any statistical relevance . I hope that there's somebody

36:46

exploring this . But , like for

36:48

me as

36:50

a neutral supporter , I would love it if there's more

36:53

possessions because , like the fast-paced game where

36:55

you know they go up and down , up and

36:57

down , it's , it's , it's amazing . But

37:00

it should also like if you , if

37:02

you have a lot of , if you're a good offensive

37:04

squad , you also want to have a lot of possessions . You want

37:06

to try a lot of things out .

37:07

Yeah , I think I also feel like so

37:09

a bit . I also live in the us , right

37:11

, and I know that how they feel about soccer , alex

37:13

, I know how you feel about soccer as well . Um

37:15

, that , uh , it's a very like

37:18

the . The scoring

37:20

moments are not as frequent as like basketball

37:22

or football and all these things , right

37:24

, and their eyes . It becomes

37:27

a bit boring . Yeah , right , um

37:29

, but then I also think that it

37:33

becomes way more impactful . Right , like in in basketball

37:36

, you cannot every time someone scores , they cannot celebrate like

37:38

this in football , right , it's not

37:40

like free throw , you know

37:42

, like , yeah , yeah take

37:45

off your jersey .

37:46

Technical foul . Second time you do it

37:48

, you're out . But uh , and I also .

37:49

I also think , like , because you mentioned

37:51

it , you're out , but and I also think like , as you mentioned , yeah

37:54

, you would like to see more actions

37:56

, right , like more back and forth , more

37:58

things like this , and I was also thinking how

38:01

, yeah

38:03

, I think also that the hardship to predict

38:05

football is that these things are very

38:08

like

38:10

, it's not like I think , if you were scoring , if the expectation is that you score , are very , um , like , it's not like I think if

38:12

you're scoring , if the expectation is that you score 100 points per match , maybe

38:15

there is . Also , it makes it easier to predict who's going

38:17

to win , because yeah there you know like

38:19

statistically these things say they , once

38:21

you get more data , the the likelihood of an outlier

38:23

, the impact of an outlier , is smaller

38:25

there's more statistically relevant

38:28

events in basketball .

38:29

Yeah , like the things that you're trying to predict , like

38:31

which are your outcome metrics ? In the end ? There's

38:33

like an average game in basketball , I

38:36

think , the average offensive rating , which is amount

38:38

of points per 100 possessions . So

38:41

they try to normalize for the possessions because otherwise

38:43

you cannot compare between error anymore . In basketball the

38:46

offensive rating I think it's like 118

38:48

last year , which is crazy

38:50

. Like that means every 100 possessions

38:52

and there's a there's roughly 100

38:54

, a bit more than 100 possessions each game . You

38:56

get like 120 points as

38:58

a team on average each game that's

39:01

a lot of things to predict that there's a lot of actions

39:03

, yeah , plus the fact that you have only

39:05

five people . So a single action

39:08

is very is way easier , like there's . It often happens that there's

39:10

two people doing a pick and roll . A single action is very is way easier , like there's . It often

39:12

happens that there's two people doing a pick and roll a

39:15

single type of action the other three are standing still in

39:18

soccer . It's almost , in think , unthinkable

39:20

that there's two people doing something and the other nine are

39:22

just like .

39:23

Oh , awesome guys , you do you unless

39:25

you're messy like a few years ago yeah

39:28

, like messy and Iniesta , but yeah , yeah

39:31

it's true , but I also think it's like yeah , I

39:34

think in football it's a bit , maybe in basketball

39:36

I think it's also more clear who's active in the play

39:38

, and in football I think it's less

39:40

clear as well yeah , yeah .

39:42

I think you can argue that in a football action

39:45

like everyone's involved actually

39:47

. I think in a task everyone's involved actually

39:49

I think it's like he probably has an instruction for every

39:51

instruction

39:55

for every player in every situation where he should

39:57

be very I think it's like it's less of

39:59

a .

39:59

I think that the elbow right of influence kind

40:02

of it's not like an elbow , it's not like it drops , I think it's more

40:04

like of a slowly fades out right

40:06

, like someone that is further from the action has . But

40:08

still , yeah , you know , could

40:10

right and never know , you

40:12

know . And I think now we're also talking about on the game

40:14

level . But I think again , if you take more of the

40:16

competition level , um

40:19

, it also becomes

40:21

it's the same thing . For

40:23

example , I also think it's probably easier

40:26

to predict the uh

40:28

, who's the winner of a points based

40:30

competition than a knockout competition yeah

40:33

, very simple thing as well .

40:35

Like I think and I know

40:37

, an average basketball season is

40:39

82 regular season games , and

40:41

then you have at least 16

40:44

playoff games . So you have , on average

40:46

, each season 98 games . That's

40:49

not like in soccer you play

40:51

less , huh so you already have less statistical

40:54

relevance . That is true that

40:57

is true , so that means that a single like

40:59

, a single sum , like

41:01

if , if you

41:03

, we've all seen like random goals happening

41:05

I think own goal was the top scorer of the european championship

41:08

um , but like if

41:10

something like that happens , where somebody

41:13

kicks in a cross and you hit it with your

41:15

knee and it ends up in the top bin and you're

41:17

like , yep , all right , this , this might

41:20

impact , they have a huge impact on the rest of

41:22

the season , whereas somebody who

41:24

accidentally bounces a basketball and it ends

41:26

up in the yeah , that's true , it

41:29

doesn't have an impact . Right , that's , that's true , that's in-game

41:31

, but also because there's 81

41:33

other games .

41:34

Yeah , that's true , that's true . So I guess it's like which

41:36

statistically makes sense , right , when you

41:38

have more data , the impact

41:41

of this , and that it's like the we're

41:43

doing a long run of explaining the law of large numbers

41:45

. Yeah , law of large numbers , yeah

41:47

exactly , exactly , all right , thanks

41:49

very much . That was the conclusion of

41:51

this , but I also think there's

41:54

a psychological aspect , right ? So

41:56

one thing we mentioned as well when we were discussing the predictions from

41:58

Snowflake , ml and the

42:01

K-11 research , someone

42:03

brought up that this

42:05

reflects on the complexities of the psychological

42:08

, like the complexity of football , which includes the psychological

42:11

effect on players

42:13

. I guess , like you're losing the match it's

42:15

80 minutes , what

42:18

decisions are you making ? And in

42:20

that I think it was the same presentation you were also there

42:22

, right , sander for the K-11 , data Science

42:24

11? . They also try to

42:26

quantify the impact of

42:28

psychological pressure . So

42:31

the title here is choke or shine quantify

42:33

quantifying soccer players abilities

42:35

to perform under pressure . I

42:37

think they took the vape still the , the

42:40

framework before about the decisions and

42:42

the valid decisions bring . But then from that

42:44

they try to say okay , when it's high

42:46

pressure moments or low pressure moments , and they

42:48

they split it around . So this is a very short

42:50

post , but they also have the , the banner

42:52

here , um , but then they kind of say

42:54

what is high ? First , they try to quantify what is

42:56

high impact and then there's a before match

42:58

and aftermath and during match . So before

43:01

match is like if you lost the last three games . Maybe there's

43:03

a high pressure , right , and there's also

43:06

the image like you got scored the 80th minute , right

43:08

, or this is a final . So they also try

43:10

to quantify these things and

43:12

then they try to see for each player , what's the impact

43:15

of that . So , in terms of making the

43:17

right decisions and in

43:19

terms of executing those decisions correctly

43:21

, and , for example , one

43:23

thing that they see here and the people that are following

43:25

the live that are just on the audio , this is

43:27

an image with neymar has

43:30

some statistics every contribution per per 90

43:32

minutes . If it is high pressure , I

43:35

guess he has higher contribution , I

43:37

guess uh . But then it also

43:39

talks about , like , how many shots , how many uh

43:42

passes , dribbles , and you can see

43:44

that um , more actions

43:46

come when there is a well

43:48

, I think actually this , this graph is a bit confusing but

43:51

uh , he dribbles more when there's

43:53

high pressure and

43:55

these actions have high ratings , I guess it's

43:58

there that tries to make stuff happen indeed

44:00

, he tries . So I think it's like there

44:03

are players that but his contributions drop

44:05

on average .

44:06

Like he has a lower average contribution for

44:08

high pressure situations because the

44:10

density is more moved to the left .

44:12

Like the book , I think so I

44:14

think this graph is a bit confusing . Maybe we can actually open

44:17

, and I did try to spend . I remember the presentation

44:19

and I also was trying to , and

44:21

I remember the conclusion of the presentation is that neymar

44:23

chokes on the pressure . I'm just gonna ignore that , I'm gonna

44:26

pretend that I didn't understand that , but

44:28

, uh , that was something . That was something like

44:30

that , right , but they do try to . So I

44:32

think this is a bit . Now I'm opening up for people

44:34

following the live stream . This is the , the banner

44:36

. So when you do research , you also have like a banner

44:38

. You present stuff . So I think this is a bit the

44:40

same . Uh , and here

44:43

they show this is really laggy derive

44:46

insights about players performance under pressure . We

44:48

aggregate his performance metrics under different pressure

44:51

levels and then they see what's the actions

44:53

with high ratings , average ratings , low ratings , and you

44:55

see amount of passes , amount of shots and all these things

44:58

. And then they also kind

45:00

of flip this around into recruitment , for example

45:02

, player acquisition , right . The

45:05

use case here is that you have modest , so

45:07

you're trying to find a replacement for that person , and

45:09

then try to see which players and how do they perform

45:12

under pressure , right , so

45:14

which players actually , uh , perform

45:16

very well under pressure , the shot execution

45:18

, for example , under pressure , which ones are average

45:21

and which ones are like the contributions over 90

45:23

minutes .

45:24

And then you can also see , kind of like I think they have here

45:26

somewhere- sort

45:29

of a weird statement

45:32

to make that you want to have people who perform better

45:34

under pressure , because there's like a , there's

45:36

a . There's a reverse statement as well like

45:39

it's when there's not enough pressure

45:41

, they don't perform well . Like yeah you

45:43

could be mario balotelli and be like 90

45:46

of the games . You're like man yeah

45:48

I don't even want to run like you also

45:50

don't want to have these kind of like .

45:51

But I think that he probably scores really well on this

45:54

because high pressure moments , but for example there this

45:56

guy , for example , they say jiru , jiru

45:58

, jiru is a clutch goal scorer , so

46:00

when there's high pressure , apparently he scores a lot of goals , which

46:03

is good , but at the same time , it's

46:05

like it's like the same thing you mentioned with the

46:07

floor razors and the ceiling raisers right

46:09

, if your team is already good , you also want people

46:11

that perform well under pressure . Because if

46:13

you're already Like , I think actually Spain

46:16

is a good example , at least Spain from

46:18

some years ago . Right , they

46:20

had a very possessed ball style

46:23

, but if they were falling behind , I

46:26

feel like they had a hard time converting to golf . Yeah

46:28

, yeah , but I feel like , because they never fell behind

46:30

, because they always have possession , they were always controlling the game

46:32

, you know . So I think it's like it's almost

46:34

like you never get to the high pressure situations

46:36

because your team what is under pressure performs really well

46:38

. Yeah , so I'm saying yeah

46:40

, so I think it's a bit of . I mean , of course , ideally you

46:42

want someone that performs well under pressure and , uh

46:45

, without pressure , right , but

46:47

I think , in reality

46:49

, if you have some trade-offs , you kind of need a bit of both players

46:51

you get trade-off .

46:52

Or you can say like you want , you want people

46:55

to have um sort of robust

46:57

curves with regards to pressure

46:59

yeah , indeed they keep executing the

47:01

same way . That's what I like I I've

47:03

been a basketball coach for eight years . That's what I

47:05

would be looking for is like a player

47:08

that , when the pressure amounts , they

47:10

keep executing the way that I expect

47:12

them to but I do feel there are some players

47:14

that perform better in the pressure yeah , yeah yeah

47:18

, but is that because they're not living up to their potential

47:20

when there's no pressure ?

47:22

I don't know . I don't know , I

47:24

mean , okay , I don't think I don't think anybody is

47:27

like like , like for example

47:29

giroux as well yeah .

47:30

I find it an interesting example , because he also

47:32

only got substituted when

47:35

there was a high pressure . Yeah , Like when

47:37

he , he rarely started for

47:39

Chelsea . So when they need a goal

47:41

in the final 10 minutes , they put him up

47:43

top and just two balls long balls

47:46

.

47:46

Okay , Actually , you're gonna

47:48

be more impactful right If they tailored

47:50

the game for you yeah , well

47:55

, I guess it's like there's always the argument of if they put someone else

47:57

into the ball for that person ? They wouldn't

47:59

score .

48:00

Yeah , fair enough , right ? Yeah , it's a bit of yeah , but how many ? How ? How large

48:02

is this statistical sample size

48:04

in this ?

48:05

so I don't know I hope it's , because it's kul

48:07

.

48:07

It's going to be good probably . I think so too right

48:09

. A lot of respect for their statistical analysis

48:11

.

48:13

And actually I wanted to find someone

48:16

from the that works on the

48:18

KUL 11 research for sports to

48:20

to be with us , but not yet , but

48:22

one day to get them .

48:24

I think in these kinds of things it's also like

48:26

what is interesting is that you can . You

48:32

can construct your roster to sort of have people that take up this role ? Yeah , like you have

48:34

. We used to have three smertens in belgium

48:37

who used to , in belgium , always

48:39

act as a super sub like and it's

48:41

you come on the court .

48:43

Super sub is the best , like backhand

48:45

slap , kind of like . You know , it's like , I feel like it's . I

48:48

don't know . I don't know if film calls me a super sub

48:50

.

48:50

I don't know if I'm gonna be happy or

48:52

if I'm gonna be like , yeah , you're just calling me , you know

48:54

I think at some point in in in

48:57

belgium there was sort of the agreement that driss

48:59

mertens was one of our best players and

49:01

yet he was the most valuable for the team coming

49:03

off the bench yeah , but and that actually

49:05

I have like I

49:08

do think there are some players that need

49:10

either to come out or they need

49:12

to go win , because sometimes , like in the beginning , the nerves

49:14

are once the ball is rolling .

49:15

there's a psychological thing as well .

49:18

I think like if you know that that's your

49:20

role and you accept and fully embrace that role

49:22

, there's a liberating part with mentally

49:24

where you come on the court and you

49:27

know they're like , oh , now we expect

49:29

this from you . And you know like like , oh , now we expect this from you . And you know like

49:31

, all right , cool , they're looking towards me . I like this , like

49:33

I've . I've been in basketball

49:36

. I've been in situations where you come on the court and people

49:38

sort of look at you like you

49:40

do this now , and you know like , okay

49:42

, it's from , it's okay for me to take initiative . Nobody's gonna

49:44

blame me if I take initiative and it goes wrong

49:46

, because that's why I came on the court like I'm

49:49

expected to shoot the ball .

49:50

But there's also an argument of why . Why

49:52

do you need to come off the bench ? For that ?

49:53

why can't ?

49:54

you just start the match and people just look at you to do

49:56

the amvc because

49:58

tactics like

50:01

uh , how many , how many coaches

50:04

have started to get like francis ?

50:06

francis played the entire tournament . Like this , like

50:08

we don't want to have a

50:11

against goal , you don't want to start

50:13

by making actions at the midline

50:15

, because that's not the tactical

50:17

plan you have in mind . You don't want to lose possession

50:19

at your own half , because it leads to , but at some point

50:22

, if you're 2-0 behind , then it's perfectly fine

50:24

to start making actions at half court .

50:26

Yeah , no , I agree , I agree , I agree

50:28

. Maybe

50:31

, yeah , no , I agree , I agree , I agree , maybe , um , sorry

50:33

, there we go , um

50:35

, maybe . One other thing I wanted to mention and I think this

50:37

is going to be controversial talking to you , tim , because

50:39

you're very much into it but eventually , like the people

50:42

that perform well under pressure , I also think of , like

50:44

the buzzer shots in basketball

50:46

. I also think of , like michael jordan and all these

50:48

things . You know , usually there are players who say this is the guy that's going

50:50

to take the last shot and usually I think , because

50:53

basketball , I think , like

50:55

the , the execution time is very

50:57

short , you know like you need to be the guy that gets the ball and

50:59

shoots . You know it's not like you get the ball

51:01

, you dribble and this and this . So I feel like the margin

51:04

for error , like one small mistake , also costs

51:06

a lot , and

51:12

I costs a lot , right , um , and I mean you know more about basketball than me , but I also get the impression

51:14

that , like michael jordan kind of guys , they were the people that perform really well under

51:16

pressure , better than

51:18

without pressure . Or

51:22

maybe I'll rephrase it a bit the

51:24

impression I have is that when

51:26

the games were very high pressure , when the stakes

51:28

were high , those are the people that really

51:31

stood out . I don't know if it's because they

51:33

play better or everyone plays worse , and they just kept

51:35

the same level .

51:36

But when you look at

51:38

, yeah

51:40

, when the stakes are high , like what's going to happen , they're

51:43

the ones that there's

51:46

a lot of research that's done

51:48

into that and what

51:51

you see in terms of like

51:53

, if you look at basketball , you see and you look

51:55

at the sort

51:58

of the metrics

52:00

of people in what they call

52:02

clutch situations , you

52:04

see that there are players that perform better but

52:08

ironically it's not like the like

52:10

. I think Michaelael jordan is is

52:12

good , but it's not disproportionately

52:15

a lot of a lot better than he usually is

52:17

, better than other players in the court but then I guess

52:19

he just shows up more because yeah

52:21

there's an attention part to it . Yeah , there's

52:23

a , there's a . You remember the situations

52:26

where he's like game six against the jazz and he makes

52:28

the game winning shot , and that's what you

52:30

remember . Those are the moments but I also wonder .

52:32

I also wonder if there's a like

52:36

they can be on autopilot and kind of like

52:38

rest there , you know

52:40

, like he's like yeah , I'm just going with the

52:42

flow , I'm just going with it because everything's under control

52:44

. But oh , things are under control , but I can .

52:46

I can switch it on yeah , but if

52:49

you look towards the percentage

52:51

, the field goal percentage on

52:53

average , and

52:55

in basketball they go really far in that they

52:57

have not just your field goal percentage

52:59

, but on open , wide open or

53:02

contested shots , they

53:04

track everything , everything For

53:06

every shot , where everybody is . They have so many tracking

53:08

metrics For example , kobe

53:11

Bryant , michael Jordan they don't have the best

53:13

field goal percentage on these kind of shots . For

53:15

starters , you have a super long tail because there's a lot

53:17

of players that one time in their career

53:19

they make this shot and then they are the best

53:21

in this because they have a hundred percent . Um

53:24

. But stephen

53:26

curry , for example , he's the best

53:28

shooter of all time . I don't even think that's arguable

53:30

anymore . He also has the best percentage

53:32

at that point . Um , um

53:36

. There's another like Joe Johnson

53:38

, um , he's a way less known player

53:40

. Yeah , I think he has at this

53:42

point , the most , the most , the

53:45

most buzzer beaters , which means that your

53:47

shot leaves before the buzzer goes and the

53:49

buzzer goes before it goes through the rim . That's

53:52

the definition of buzzer . He has the most of them

53:54

, like I think he has nine of them , which

53:56

is crazy yeah and that

53:58

might be an example of a player that shows up

54:00

when it

54:02

that . That is , I think , one of the only

54:05

players . Yeah , there's a number , there's a number of them

54:07

that are known for that , and it's like joe johnson

54:09

, rob , or so . There are players that

54:11

tend to show up yeah but

54:14

at some point I think it's a self-fulfilling prophecy

54:16

.

54:16

Yeah it's true . It's true . I think a lot of it

54:18

is perception as well . Right so , but that's

54:20

why I think , looking into the numbers and really

54:22

trying like I think it's a very good way to just try to remain

54:24

objective . Yeah , I don't know in the sports

54:26

that , in an area that is very opinionated

54:29

yeah , I have a .

54:31

I have a very like , I'm very opinionated on

54:34

on these kinds of things . Um , and

54:36

and I I recently

54:38

like , I recently gave a presentation

54:40

on what businesses

54:42

can learn from sports analytics and one of the one

54:44

of the easiest lessons you can draw is that

54:46

the statistics alone are not going

54:49

to bring you there . Um , we saw that

54:51

a number of times in the nba , where people sort of

54:53

just try to look at oh , the best team

54:55

is doing this , we're going to emulate the statistics

54:57

, we're going to build our team according to the statistics doesn't

55:00

work yeah the entire moribund era

55:02

in houston . Um didn't

55:04

work , yeah , um , but you

55:06

always have to have , like , the philosophy and the

55:09

and the statistics work hand in hand and there's

55:11

then there's just a huge survivor bias

55:13

yeah it's always sports super prominently

55:15

present yes

55:18

, yes , and maybe uh

55:20

, now to switch gears a bit .

55:22

Uh , in basketball

55:25

, I think there's more tech in

55:27

like , more

55:29

in the actual like in the game itself

55:32

, like in the . Well , I'm saying

55:34

this because I think now in the euro , one thing that stood

55:36

out a lot is the amount of technology

55:38

, yeah , that we're starting to see in the sport right

55:41

, the var , but also the offside

55:43

. I think offsides in the zero was crazy . There

55:45

were some things there's like the guy's foot

55:47

is five centimeters above and it's

55:49

like man hell , come on , um

55:52

, I think it was denmark that scored a goal . That was

55:54

, uh , against yeah germany . I want to say

55:57

, um , the handball

55:59

stuff . So now there are chips inside the ball

56:01

and there's the goal line technology and there's

56:03

this and there's that . Um , I

56:05

had a mini discussion . Well , maybe we can uh

56:08

to share a bit

56:10

some things for people that may be following

56:12

us on video . This

56:14

is a bit the um spain

56:16

, england offside right so actually not offside

56:19

, it was inside right but , uh , they're

56:21

trying to show a bit the simulations

56:23

that they did . They drew a line , I think . If you go

56:25

down , yeah , they have this like 3d

56:27

rendering thing here and you can see

56:29

the guy's arm is up in front and his knee

56:32

is actually in front and that's why he's on side right

56:34

. And then you see some other examples . Let's

56:37

see this one uh

56:39

, this is lukaku , lukaku , right

56:41

and uh , so this was against romania he's

56:44

offside right because yeah

56:46

, yeah , but like he doesn't count , no

56:49

, the arm doesn't count only the body parts

56:51

that you can score with to me . It's

56:53

crazy . It's like come on , man . The guy's like

56:55

look at this man look at his

56:57

foot . Look at his feet . If

57:00

he had cut his nail , he still nails

57:02

that day , you know his knees will

57:04

his knees will .

57:05

Yeah , I

57:07

said that yesterday about the goal of Oyarzabal

57:09

accident

57:16

. Like if you happen to be like one

57:18

leg extended , one leg behind , you

57:20

probably will not be offside because the other

57:22

person is with the knee forward

57:24

.

57:24

Like that's the point where you know it's

57:26

crazy that it just depends on and

57:29

you cannot time it if you try to kick

57:31

across because I feel like in

57:33

american football sometimes , because you like the rules

57:35

are very clear as well , like for , so

57:37

for touchdowns or whatever , you need to have both feet on

57:39

the ground , but then you see like players catching

57:42

, and you can see they're clearly dragging their feet

57:44

on the ground so they really like move

57:46

like they're almost like a dance . You know they do this thing , but

57:48

I think in football it's so dynamic that it's like you can't

57:51

it's impossible , it's impossible .

57:53

The tippy toes in football is one of the most beautiful

57:55

things .

57:56

The fact that they drag their toes . But

57:58

, um , maybe this . Uh

58:00

, I had a mini discussion with italy . Do you like this

58:03

technology in football ? Maybe we can just do a quick . Do

58:05

you think this adds ?

58:06

to the sport . This one I think

58:08

I do because , like for me , offside is black and white

58:10

okay , so I think everything that helps in

58:12

making it okay , we're gonna move to some other ones

58:14

.

58:14

But uh , and what

58:16

about you ?

58:16

tim , I think , yeah

58:19

, I , I do love these technologies , um

58:21

, but in the case that you adjust the

58:24

, the rule book alongside with it , as

58:26

in um , you have technology to

58:29

very much black and white , decide

58:31

if something is offside , make

58:33

it that there's also no room for interpretation . Either

58:36

you do that or you leave it up to interpretation

58:38

, but then you leave the technology out of it as well , because

58:40

if you don't have the combination of the two , like , it

58:42

leads to frustrating and and one of the examples

58:45

is the I don't know if you have it the goal

58:47

of a panda , where he like touches the ball

58:49

, but like very slightly with the hand and

58:52

there there was a discussion when

58:54

, um , when the goal happened

58:56

. There was a discussion obviously in belgium

58:58

journalism afterwards where I

59:01

think some people said , like you have to , it

59:04

has to be a voluntary movement , which I think obviously

59:06

is not a voluntary movement . You have

59:08

to drastically change the trajectory

59:11

, like you have to impactfully change the trajectory

59:13

of the ball , which obviously doesn't like

59:15

. And then it's the question like yeah

59:18

?

59:18

I think there was just a wrong call right , but

59:20

I think I think .

59:21

But I see what tim's point is like . I guess what he's saying

59:23

is like either you say a touch is

59:25

a touch and that's it , and if you happen

59:27

to touch your hand inside the box it sucks

59:29

, yeah . But other , if you say like

59:32

intentional , not intentional , but then you have this technology

59:34

, that kind of supports saying yeah , it did touch , but now

59:36

you have to say it's intentional , I think it's a bit weird

59:39

.

59:40

The goal of the Netherlands

59:42

against England , where

59:44

Xavier Simons was

59:46

against England , right , xavier Simons kicks

59:49

it like super hard .

59:51

It's against France , the one Against France .

59:53

Yeah , they disallowed it . And like

59:55

there's a guy standing there and everybody

59:58

knows that the keeper is not going to get it . Like

1:00:06

the ball was , I think , 120 kilometers an hour from inside the box like

1:00:08

nobody's gonna like . The keeper didn't even react , yeah . But then there was a guy standing next

1:00:10

to him and in some interpretation of the rule book you

1:00:12

could say that if he dives like

1:00:15

, he jumps against the player , so he's influencing

1:00:18

.

1:00:18

And then you're like , yeah , okay but that's because of offside or

1:00:20

what because of offside yeah , because I guess in offside

1:00:23

there is a bit of subjectivity in , uh , if

1:00:25

the guy interferes with the play , right . So , for example , if

1:00:27

you , I'm clearly passing the ball to you , tim

1:00:29

, and you're offside , but

1:00:31

then you didn't touch the ball because maybe

1:00:34

you couldn't , you weren't fast enough , which , tim

1:00:36

, is a bit hard think , but anyway . And

1:00:39

then let's imagine the ball goes in and it's a goal , right

1:00:41

. Somehow , because the ball was a pass

1:00:44

to you , you interfere

1:00:46

in that play and therefore that's offside . But

1:00:49

it's not like . This is a very hypothetical

1:00:52

black and white example , but it doesn't have to be as

1:00:54

objective , right . Like maybe I do a through ball and

1:00:56

maybe there are two players and one is offside , the other one is not

1:00:58

, and maybe the ball goes through .

1:01:00

Or maybe , like you know how , do you , but that's also

1:01:02

, I think , you . You can't like . I

1:01:04

see how you see the rules have to follow if

1:01:06

you implement stuff like this , but you can't . I mean you

1:01:08

can make a handball black and white , you

1:01:10

just can't .

1:01:11

Yeah , I yeah

1:01:14

. Hockey is like and

1:01:16

I think there's way less discussion in hockey it's

1:01:18

like it hits your foot . It hits your foot

1:01:21

Like you have to . It's your responsibility to

1:01:24

make sure it doesn't hit your foot .

1:01:25

But I think in football there wasn't a . Isn't there a rule that , like

1:01:27

, if your hands is near your body , it's

1:01:30

not a handball ?

1:01:30

But then it's gray .

1:01:32

No , like

1:01:41

then it's a gray area again . But how it's touching ? Is it touching , not touching ? Look , it's

1:01:43

touching . It's not touching this . Oh , it's even clicked or something

1:01:45

. I , I see what you're saying . Yeah , I

1:01:48

think it's . Uh , I think I don't know . I feel

1:01:50

like the . To me , the , the accuracy of these

1:01:52

devices is a bit questionable . You know , when I see the guy with

1:01:54

like a little foot and even like in brazil

1:01:56

there was a joke that like they zoomed in and you can see the cells

1:01:58

and like two cells are above them , it's

1:02:01

outside , you know , but to me that's where ?

1:02:04

yeah , that's that's .

1:02:05

That's where I have a bit of a like . Come on , man

1:02:07

, like , just let it play you know , the thing

1:02:09

is also like .

1:02:11

The thing I dislike

1:02:13

the most is like far was intended to

1:02:16

solve , like clear and obvious errors , maybe

1:02:19

if you have to use a microchip

1:02:21

inside a ball to check whether someone

1:02:23

touched it don't bother

1:02:25

right , I mean yeah

1:02:27

, that's true or you completely

1:02:29

automate it away , but make a choice this

1:02:32

is the chip you're saying , right , sander ?

1:02:33

I think you said that it's from .

1:02:34

Adidas there's

1:02:36

literally a chip inside that

1:02:39

every touch of the ball , like 500

1:02:41

times a second , you get data

1:02:43

. So

1:02:45

every touch of the ball you record and I

1:02:47

think this is also the example right , Because I wanted

1:02:49

to show specifically .

1:02:51

So basically , Ronaldo said that he scored

1:02:53

a goal , but the signals in the ball says that Ronaldo

1:02:55

did not touch at all .

1:02:56

Yeah , so close not close .

1:02:57

So he's a liar in the ball . Says that Rinaldo did not touch

1:02:59

at all .

1:03:00

Yeah , so close , not close

1:03:02

. So so he's a liar

1:03:04

. Cristiano Ronaldo , we're calling you out , you're

1:03:06

first you're first , well

1:03:09

, I'm not your second , right , because we're this is the third year but

1:03:11

anyways we got the scoop over Sky

1:03:14

Sports .

1:03:14

Yeah , but yeah .

1:03:17

I do think it's . Yeah , I think those .

1:03:18

Yeah , I see what you're saying I think it's interesting

1:03:21

and you for stuff like this . It's fun

1:03:23

to use and to put a sensor in the ball and stuff , but

1:03:25

it shouldn't be used for decision

1:03:27

making of the referee like a sensor inside

1:03:30

the ball yeah , maybe

1:03:32

, I don't know the only thing which is nice about the sensor

1:03:34

is you can identify the exact point of

1:03:36

context for like offside decisions , because

1:03:38

that's always like also did the

1:03:40

ball already leave the field or not ?

1:03:42

that's always a point of discussion . We

1:03:44

didn't see that in this . We didn't see that data use

1:03:46

like that right , but that's what the semi-automated

1:03:49

it also uses the

1:03:51

. So tell us about it , tell us more about it . What is the semi-automated

1:03:53

thing I'm going to put on the screen here ?

1:03:55

well , so basically it's , there's

1:03:57

like 12-ish cameras across the stadium

1:03:59

. They hang from the roof and

1:04:01

they track every player with like

1:04:04

29 of these body

1:04:06

points at any point in time . So

1:04:09

basically , they just track you at

1:04:11

any point in time and these body points

1:04:13

are across your body and are the ones which are relevant

1:04:15

for scoring goals .

1:04:17

Okay , so this is a video from FIFA for people following

1:04:19

the live stream From

1:04:25

FIFA , so it's not like a research group or anything . No , no , it's actually . Yeah

1:04:27

, yeah , so that .

1:04:27

so this exists , but they're trying right , so now they have the sensor to see where

1:04:29

the so yeah , the sensor of the ball then makes it . You can

1:04:31

identify the exact point of contact

1:04:33

when the ball leaves

1:04:35

or the first contact is made okay

1:04:38

, and I also wanted like these 3d simulations

1:04:40

.

1:04:40

I feel like that's what people think programmers do you know

1:04:42

, like you have like three screens and like this , and

1:04:45

then you have the bug going through the wire , you know , and

1:04:47

then it goes to the internet or something um

1:04:49

isn't that what you do man

1:04:51

. Well , yeah , I do it I keep

1:04:53

selling the wrong thing um

1:04:57

, it's okay , really cool , and then I guess you

1:04:59

have all the cameras with .

1:05:00

Oh , wow , like the thing here is like . What I always

1:05:02

wonder is like you see these body points right

1:05:04

?

1:05:04

yeah some people , some

1:05:06

players are bigger than other players well

1:05:09

, excuse me , more muscular , he has thicker

1:05:11

bones no

1:05:18

, but do they adjust the model based on the player in question

1:05:20

yeah , I think so that's

1:05:22

cool um , yeah , I was also thinking

1:05:25

about that right , like today we have a referee running

1:05:27

around , but I was thinking like , oh , yeah , okay

1:05:29

, there's some outside examples what if the

1:05:33

actually was this used in the euro or no , this one

1:05:35

, yeah , semi-automated , okay . Um

1:05:37

, I'm also wondering like we have referees that run around

1:05:40

, basically , and they kind of there's actually a lot of training

1:05:42

where to position yourselves and whatnot . Do

1:05:44

you think in the future we'll have a pure VAR ?

1:05:48

I don't think so but

1:05:50

you could .

1:05:51

You could because you have , like , a camera .

1:05:52

I mean , you still have someone to blame , but just that they're not on the pitch

1:05:54

but then like , what do you do if they're things

1:05:57

get heated like two players start pushing

1:05:59

each other ?

1:06:00

let it happen , bro , I mean

1:06:02

what's the game ?

1:06:03

but ?

1:06:03

what does our first dude they're

1:06:05

like they take their ears off and just like come on , I'm

1:06:11

from brazil , man . I've seen things like too much

1:06:13

like if , if dust comes . How do

1:06:15

you say dust comes ?

1:06:16

the shovel or something , that's it shovel yeah , it is okay , push

1:06:18

comes the shovel thank you , I

1:06:22

looked at it .

1:06:22

I was like Alex , help me , it's okay , when

1:06:25

the time comes , they don't do much either .

1:06:27

So just , you didn't use it the second time around

1:06:29

, did you ? No , no .

1:06:30

I was too embarrassed . My

1:06:32

brain you know , my mouth was ahead

1:06:34

of my brain . I was like , yeah , I shouldn't have started with that

1:06:37

.

1:06:37

But there's times at which I do also think

1:06:39

, like what's the guy still on the pitch for ? Like

1:06:42

, literally , like with the VR , for

1:06:44

example . It's like people 50

1:06:46

kilometers away that say go watch this screen

1:06:49

, yeah .

1:06:50

Or even like the linesman as well . Yeah , they're

1:06:52

just going sideways . If it's just a camera , if

1:06:55

it's just a camera and there's so much really just controlling with a stick

1:06:57

that you can move faster than whatever that guy is doing

1:06:59

.

1:07:02

You could automate the whole thing .

1:07:03

Yeah , right , oh yeah

1:07:07

, maybe one last final

1:07:09

question . We're

1:07:11

talking about like predicting stuff with AI

1:07:13

and the impact of technology in sports

1:07:16

. I

1:07:18

had a discussion with my friends a while ago that we're talking about

1:07:20

the point system versus knockouts right

1:07:22

, knockout stages , right . So in like

1:07:24

in the in the uk or in england you have

1:07:26

the premier league and you have the fa cup , which

1:07:29

I always thought was funny that everyone says fa cup

1:07:31

. But the first time I heard it was like fuck up . I

1:07:34

was like anyways

1:07:36

, um , and

1:07:38

I remember we're talking about like which one do you prefer

1:07:41

? That's kind of what the discussion was about . And then in the

1:07:43

end we kind of agree that point systems

1:07:45

are more fair , the

1:07:48

best team usually wins in

1:07:50

point systems , but knockouts

1:07:52

are more exciting because

1:07:54

it's unpredictable . So

1:07:57

then I guess it kind of to bring this back to this

1:07:59

discussion is even

1:08:01

if we could like is it a good thing ? Like

1:08:03

I think guess it kind of to bring this back to this discussion is even if we

1:08:05

could like is it a good thing ? Like I think usually we say we're predicting and

1:08:07

this and this and it's not fair

1:08:09

and this usually seems maybe a bad thing , but

1:08:12

isn't that part of what is exciting about sports , the fact that unpredictability of it , like

1:08:15

if everything was really like

1:08:22

, if all the predictions that we did from either Snowflake or the research center from KU Leuven

1:08:24

were correct , would that be good ? Is that a good thing or is

1:08:26

that a bad thing ?

1:08:28

So that's . That's where the conversation

1:08:30

, like we had we had this conversation I

1:08:34

mentioned to Sando before that . So

1:08:37

basketball has become

1:08:39

rather predictable , as you've

1:08:41

shown before . Like eight of the ten last

1:08:44

nba championship winners

1:08:46

could have been predicted . Like could have been predicted

1:08:48

based on data available at that point . Um

1:08:50

and the uh

1:08:53

. Popularity of basketball has been doing like

1:08:55

it's . It's going down . Like

1:08:58

there's a less people watching . There's more

1:09:00

people . Like more people watch the caitlyn

1:09:02

clark ncaa final in

1:09:04

women's basketball than they did the

1:09:06

final in nba basketball really

1:09:08

yeah , wow so , um

1:09:10

, it

1:09:13

is for , like , a lot of people are criticizing

1:09:15

that it is because of the fact that it's been

1:09:17

very analytics driven , um

1:09:19

, and it's analytics driven in two ways

1:09:22

. Like , on the one hand , what the nba tried to

1:09:24

do was make it really popular by

1:09:26

introducing all these rules where people scored

1:09:28

more , which is why , like

1:09:30

, at some point you can like there

1:09:32

there was at some point they introduced a hand checking rule

1:09:35

, like if you touch somebody

1:09:37

in front of you with your hand , it's a foul

1:09:39

which very much limits

1:09:41

you as a defender in what you can do in terms

1:09:43

of steering somebody . And

1:09:45

then they introduced sort of the they're

1:09:48

like James Harden went with with fouls at

1:09:50

some point all the way , like he always went

1:09:52

in , jumped into the defender who was standing

1:09:54

still and was a foul , like you had to jump

1:09:57

backwards . Like a lot of rules

1:09:59

were introduced where , um

1:10:01

, it made it more difficult on the defense

1:10:03

and better on the offense . people

1:10:05

really went away from it , like they really didn't

1:10:07

like it , because everything became

1:10:09

a scoring festival , like it wasn't

1:10:11

special anymore . Um , and

1:10:13

then this year they introduced some rules where defense

1:10:16

again got some power back . Um

1:10:18

, and and you saw that people that

1:10:20

again the , the viewing numbers went up , but

1:10:23

still the playoffs were very , were

1:10:26

rather predictable . Everybody

1:10:28

thought celtics couldn't win . Celtics won , voila

1:10:30

. There's more to it , obviously , but and

1:10:33

the popularity is still not picked

1:10:36

up from . You know , yeah , a

1:10:38

couple of years ago . So the unpredictability

1:10:41

, the , what is it ? The Cinderella , the

1:10:43

Cinderella story . Like the , the small

1:10:46

team comes up and wins big

1:10:48

. Those kind of stories sport needs

1:10:50

it like it's obvious like you need

1:10:52

to have the Leicester story once in a while the

1:10:54

. Union story , although Union is not even close

1:10:58

, it's not even .

1:10:59

But like , you need these stories

1:11:01

because otherwise yeah

1:11:03

, teams are , but if you

1:11:05

have too many of these stories , but

1:11:07

that won't happen right if

1:11:10

you have .

1:11:11

Yet , there's the economies of scale part . Manchester

1:11:14

City proves it year and year again , real Madrid proves

1:11:16

it year and year again . Uh , but if

1:11:18

you have too many of these stories , yeah , then

1:11:20

you also lose out .

1:11:21

That's , that's also because I feel like I guess the thing

1:11:23

is like if you have too many of these stories , then maybe the rules are not fair

1:11:26

.

1:11:26

The rules are not like you know , because you still want

1:11:28

somehow that the best team will win nba

1:11:31

was more popular when , when golden state

1:11:33

warriors won like four times in

1:11:35

five years , than it is now . Now

1:11:37

do we have five times a different champion the

1:11:39

last five years in nba . So yeah , maybe

1:11:41

there is a like we . We do

1:11:43

like and that's what they say in the us as well

1:11:45

we do like dynasties yeah we do

1:11:47

like real madrid and the legendary

1:11:50

galacticos . And true , what ?

1:11:51

do you think something ? What's your thoughts on that ?

1:11:54

I do agree those man

1:11:56

, not the nba part

1:11:59

no , but I think like it's nice

1:12:02

to have these unpredictable games or

1:12:04

champions once in a while , but I

1:12:07

do think , yeah , you would also not feel

1:12:09

fair if you are the best

1:12:11

team and that's not somehow represented

1:12:14

in you getting rewarded for it at the end of

1:12:16

the road so yeah , cool

1:12:19

guys .

1:12:20

Thanks a lot , this was fun

1:12:22

. Thanks

1:12:24

, tim , that

1:12:27

wasn't a good clap , that

1:12:30

was man , it was just closer to the mic thousand

1:12:32

excuses . Thanks y'all .

1:12:34

Thanks alex you

1:12:39

have taste in a way

1:12:41

that's meaningful to software people hello

1:12:44

, I'm bill gates I

1:12:49

would , I would recommend .

1:12:50

Uh , yeah , it writes

1:12:53

a lot of code .

1:12:54

For me , usually it's slightly wrong

1:12:56

. I'm reminded , incidentally , of

1:12:58

Rust here Rust .

1:13:03

This almost makes me happy that I didn't become a supermodel . You said that , cooper

1:13:05

and Ness . Boy

1:13:07

.

1:13:07

I'm sorry , guys , I don't know

1:13:09

what's going on . Thank you for the opportunity

1:13:12

to speak to you today about large neural networks

1:13:14

. It's really an honor to be here . Rust

1:13:16

Data topics . Welcome to the data

1:13:18

, welcome to the data topics podcast To

1:13:21

you .