The alt-3 tables are back!
After a long wait, alt-3 is back. For now it’s only England’s Premier League and Championship and Italy’s Serie A that are given the alt-3 treatment for the 2020–21 season, but more leagues will be added in the coming days (before it’s too late!).
The long wait is a consequence of the coronavirus pandemic, which not only caused the old notion of ‘home advantage’ to become hopelessly wrong (in the absence of supporters), but it also stole my own time for developing a better approach — until now.
The ‘better approach’ — which I’ll boldly call alt-3 ultimate, because it’s provably the best that can be done! — is to allow each team in a league to have its own, separate home and away strength parameters. See the updated page The mathematical method explained for details.
A new graph is also added now, for each league — a graph that compares the home and away ‘points per effective match played’ rates of all teams in the league. Here are the current instances of that graph for the Premier League, Championship and Serie A:
Perhaps the most remarkable feature in the three graphs above is that in the Premier League more than half of the teams this season are doing better away from home than they are at home — and there is considerable variation between teams.
This new type of graphical display is helpful (I find) when it comes to making sense of teams’ schedule strength charts — which can as usual be seen by clicking any team’s name in the alt-3 league table. The new graph helps to explain which fixtures are relatively ‘easy’ and which ones are relatively ‘hard’.
I hope to add further graphical displays too, when time permits. For now, though, I am happy to have got things working again, and indeed to have improved alt-3 substantially (in my opinion!) in the process.
This coming Tuesday (April 20th) I’ll be giving a Zoom talk about all this, in the statistics seminar series of Ca’ Foscari University in Venice. In case any reader of this wants to attend, the details are at https://www.unive.it/pag/16818.