Describe the issue
Many a times players will be out of form but the moment they see a specific opponents they somehow find that form, great Example is Steve Smith against India, no matter how his recent form is, he will play well against India, there are many more examples like that, one of the problems with our model is, it doesn't take that into account.
Solution
There's no prototype solution for this, You can collect data from websites(make sure its legal) and form a algorithm or detect patterns, anything that would improve the models losses.
One solution we have in mind to try out is to take into account both recent form and performance of player against that specific team, so if a player has had a rough patch in recent games but has an amazing record against that opponent the model should take that into account so that it predicts a better 11, we think it might help in improving the model.
Describe the issue Many a times players will be out of form but the moment they see a specific opponents they somehow find that form, great Example is Steve Smith against India, no matter how his recent form is, he will play well against India, there are many more examples like that, one of the problems with our model is, it doesn't take that into account.
Solution There's no prototype solution for this, You can collect data from websites(make sure its legal) and form a algorithm or detect patterns, anything that would improve the models losses. One solution we have in mind to try out is to take into account both recent form and performance of player against that specific team, so if a player has had a rough patch in recent games but has an amazing record against that opponent the model should take that into account so that it predicts a better 11, we think it might help in improving the model.
Comment if you would like to work on it