:format(webp)/cdn.vox-cdn.com/uploads/chorus_image/image/69777686/1335905272.0.jpg)
I made the grievous error of telling Charlie via DM that this prediction model was suggesting scorelines that were way to low- no team was predicted to score more than a single goal- and then all but one team in the league go and score one goal or less.
Last Week’s Predictions:
This week's model went 3/10 on both results and scorelines. Relatively the same as last week's model- at least there's consistency! pic.twitter.com/UVxEyUz6LP
— Charlie Tuley (@analyticsLaLiga) August 23, 2021
As I said last week, I find myself in the weeds a little bit evaluating these predictions because when I look at them, I see a lot of results come up draw that were relatively close in the prediction model to begin with. For instance, per the model there was a 70% chance that Granada-Valencia finished either with a Granada win or a draw, and it ended in a draw with Granada winning the xG battle 1.9 to 0.8 (most of that being a given penalty). The model correctly predicted how that match would look on the field even if Valencia happened to nick a goal on a penalty right at the end of the game.
This week’s predictions:
:format(webp):no_upscale()/cdn.vox-cdn.com/uploads/chorus_asset/file/22806542/matchday_3._1.PNG)
You’ll see that Charlie and Mitchel have started including the betting odds for comparison next to their predictions so that you can get a feel for how their model is matching up to other predictions. The model smiles on us again, predicting a 0-0 draw even though Atleti winning is still more likely than any other result. I think Mallorca/Espanyol is the first time in three weeks I’ve seen a draw be listed as the most likely result, though I expect as they continue to build on data from this season we will see more of that. The prediction of theirs I am most dubious about this week is the La Real prediction. I think they should be favored to win, but given their performances this year that prediction seems too strong in their favor over a Levante side with real attacking potential. All the scorelines are still predicted to be very low.
Thoughts?
What do you think of this model so far? How can it be improved?
As always, give Charlie a follow @analyticsLaLiga and Mitchel @msocanalytics as well. This model is always growing and over time it will start to include more data from this season so it should get even better!