hi there, I really in love with this. I am VERY surprised they dont include the following variables(almost no categorical variables at all yet still so interesting) as determinants of weather a goal is scored or not -Quality of Goal Keeper and their form -Whether the striker is home or away -Strikers performance in say the last three events of a different game(say if a player is on form) -Strikers fitness to the game -strikers favorite leg(right or left) thanks to Data Bricks, Microsoft and Laliga for allowing such shareable insights!
eye opening presentation and well presented especially the bit of how the goal probability model is just for broadcasters and not the same thing as expected goal
May I ask: how come the goalkeeper's quality is not included between the variables that the model uses to calculate the goal probability? Just as the player's quality influences the probability of the same player scoring the goal on that specific opportunity, the goalkeeper's quality certainly influences one's probability to score or not, doesn't it? UPDATE: I noticed someone asked him about that, but still: what if there is no goalkeeper at all in the goal at the time the player shoots? It seems to me as an important variable to consider, since we are talking about a goal probability after all (what if a child plays in goal?).
Yes Goalkeeper quality should be included and also the defender quality if he is close or in infront. They could do it in the next season if not done yet..!
because in itself, it is quite hard to establish that parameter: lots of factors for judging a goalkeeper are assessed subjectively, for example based on the playing style of a specific team
I think the to a certain extent the same dataset can be used to train and predict the defence or keeper side aswell... If its not a goal then either the defence has blocked or keeper saved. So same Logistic Regression will be better for it 🤷♂
This model's accuracy could be improved by incorporating additional variables. Currently, it doesn't consider factors such as shot speed and trajectory (vertical and horizontal arch), nor does it account for the player's position, physical state, or the ball's speed upon reception.
hi there, I really in love with this. I am VERY surprised they dont include the following variables(almost no categorical variables at all yet still so interesting) as determinants of weather a goal is scored or not
-Quality of Goal Keeper and their form
-Whether the striker is home or away
-Strikers performance in say the last three events of a different game(say if a player is on form)
-Strikers fitness to the game
-strikers favorite leg(right or left)
thanks to Data Bricks, Microsoft and Laliga for allowing such shareable insights!
eye opening presentation and well presented especially the bit of how the goal probability model is just for broadcasters and not the same thing as expected goal
really interesting knowing how all these data are processed behind the scenes. Plus, the presenter did a very good lecture about the content. Cool.
May I ask: how come the goalkeeper's quality is not included between the variables that the model uses to calculate the goal probability? Just as the player's quality influences the probability of the same player scoring the goal on that specific opportunity, the goalkeeper's quality certainly influences one's probability to score or not, doesn't it?
UPDATE: I noticed someone asked him about that, but still: what if there is no goalkeeper at all in the goal at the time the player shoots? It seems to me as an important variable to consider, since we are talking about a goal probability after all (what if a child plays in goal?).
Yes Goalkeeper quality should be included and also the defender quality if he is close or in infront. They could do it in the next season if not done yet..!
because in itself, it is quite hard to establish that parameter: lots of factors for judging a goalkeeper are assessed subjectively, for example based on the playing style of a specific team
So well explained more from the back-end!
I think the to a certain extent the same dataset can be used to train and predict the defence or keeper side aswell... If its not a goal then either the defence has blocked or keeper saved. So same Logistic Regression will be better for it 🤷♂
wow , that's really useful.
how can I find more lectures like this ?
@4:17 what time of camera ?
Presentation is really useful, where can I get the dataset ?
This model's accuracy could be improved by incorporating additional variables. Currently, it doesn't consider factors such as shot speed and trajectory (vertical and horizontal arch), nor does it account for the player's position, physical state, or the ball's speed upon reception.
Excelente la presentación, gracias!