Prediction Model: Round 27

Prediction Model: Round 27

We had a little bit of fun with the FA Cup last week and although I warned against betting on the model’s calls (as it’s untested in the cup), it went on to record three out of four correct results, only missing out on a prediction of Liverpool to beat Arsenal from the all-Premier League fixtures. Nice to record a good performance but I should have trusted in the simulations and backed their calls!

prempred

In the last set of Premier League fixtures, the model bounced back from a couple of disappointing rounds to record a tidy performance. Everton v Crystal Palace and Manchester City v Sunderland were called off due to some crazy winds in the North West, leaving eight mid-week fixtures to play from our original ten predictions.

The model called five of those remaining eight games correctly, returning a 46% profit if you’d bet the same stake on each prediction. None were missed by very much, but three draws tripped us up as Arsenal, Chelsea and Cardiff all failed to pick up their predicted wins. I’ll run updated predictions for the postponed games as and when they’re replayed.

On to predictions for this weekend and thankfully – with a few exceptions – the transfer window changes have settled down so we’re back to running expected starting line ups as they’re listed by Fantasy Football Scout. Crystal Palace are an exception this week and I’ve had to switch a few players including Tom Ince until they build up some more data. The model splits home and away stats, so it can take three or four games before a player is able to be included properly. I’m also putting a health warning on Everton’s percentage as the model doesn’t know anything about Lacina Traoré yet.

You can avoid these two predictions, or cross your fingers that unknown players don’t play well or badly enough to swing the path of a whole game.

Here are this weekend’s percentages:

15-2

The stand-out upset prediction is Sunderland away at Arsenal. Although Sunderland won fractionally more of the two thousand simulated games, over 28% of them were draws, which is how we’ll bet this week using our usual rule of over 27% probability = draw. Liverpool are predicted to have a field day against Swansea and Man City may be at their free scoring best against Stoke.

Here are the model’s deeper passing and shooting stats:

15-2 detailed

And final predictions:

Chelsea v Everton – Home win
Arsenal v Sunderland – Draw
Cardiff City v Hull City – Home win
Manchester City v Stoke City – Home win
West Bromwich Albion v Fulham – Home win
West Ham United v Southampton – Home win
Crystal Palace v Manchester United – Away win
Liverpool v Swansea City – Home win
Newcastle United v Aston Villa – Away win
Norwich City v Tottenham Hotspur – Away win

See you next week!