Expected minutes for players moving clubs

[nbarlowATI]

This season we were caught out getting players e.g. Lossl, who had played lots of minutes for one club last season, then transferred but were only ever going to be backup for the new club. Rather than taking the average of the minutes in the previous season, maybe we should assume zero for gameweek 1 if the player has changed club?

[jack89roberts]

There's also the related cases where:

• A player is more likely to play after moving (e.g. to a lower club).
• A player who didn't move is less likely to play as the team bought someone else.
• A new player is added to the game.

For new players, before the start of this season we added a guesstimate based on their price relative to the price of teammates in their position. Doing something similar to that could have worked for Lossl (£4.5) as well, since he's priced much cheaper than Pickford (£5.5).

[nbarlowATI]

All good points!

Maybe we can/should do a study looking at: how many minutes each player played in GW1 of each of the last few seasons, as a function of:

• How many minutes they played for that team on average in the previous season (if staying at same team)
• How many minutes they played for another team in the previous season (if moved teams within EPL)
• Ratio of player's price to the most expensive player in that position+team.
• Number of players more expensive than that player that have joined or left the team in same position.

We also need to take into account that it doesn't matter so much if we underestimate a player's minutes (it makes us less likely to pick them but there are other good options we could pick) but if we overestimate them we run the risk of having non-playing players in our squad.

[jack89roberts]

I've made this change (and removed the hack of estimating minutes from FPL price), but not marked this to close with the PR as there are some other ideas/things to investigate above.