Games have been coming thick and fast and the model’s been holding its own for the past few weeks. No spectacular wins on the betting, but if you’ve been following its calls, then Tuesday night marked three profitable rounds in a row.
Six mid-week results were called correctly, with the games at Arsenal, Liverpool, Stoke, Sunderland, Fulham and West Brom all running as forecast. There were a few odd games, where the dominant team lost: West Ham at Palace, Southampton against Villa and arguably Man U against Everton. The model will rarely get this type of backs-to-the wall result right because unless a team’s shooting or defending is truly appalling, if they dominate possession then across 2000 simulation runs they’ll tend to come out on top.
The final result to round up the mid-week list was Newcastle predicted to beat Swansea, which was a long way off, but overall that wasn’t a bad round at all.
I’m later than usual with the predictions for this weekend because the model’s decided to be a bit controversial and I wanted to make very sure that I hadn’t made a mess of the numbers and take a bit of time to explain why…
Here are the percentages:
Two rows immediately leap out of that table – Newcastle to beat Man Utd and Man City to lose at Southampton. Let’s pick them up one at a time.
Man Utd first and this prediction is being pushed by a lack of strikers and in particular, a lack of Robin van Persie. Wayne Rooney’s suspension isn’t helping either.
This is Fantasy Football Scout’s expected line-up.
One of the fun things we can do with an agent based (individual player-based) model is to fiddle with the starting line-ups and see what happens. Swapping in Welbeck and playing with two strikers doesn’t make much of a difference and neither does pretending that Rooney isn’t suspended. Newcastle would still be forecast by the model as most likely to win under either of those scenarios, because their shooting frequency is higher. Excluding Nani – who is wildly inaccurate – that Man Utd starting line-up has shot fifteen times in total so far this season at home. Will the rest step up without Rooney and RvP? They might but the model can’t see that for the moment.
Mostly, it hangs on van Persie. With him in the team, the percentages would immediately switch to 48% in favour of Man Utd, but without him, there just aren’t enough goals in that starting eleven to predict a win and Man Utd’s defensive stats aren’t good enough to make up the difference.
I’ll be following the model’s call because that’s the discipline I’ve set myself this season, but am definitely waiting until that starting eleven is announced! Moyes says he ‘doesn’t know’ when van Persie will be back but I’m going to wait and see, just in case.
Manchester City’s prediction is being driven by the past away form of the whole team rather than one specific player. It’s well known that their performances home and away have been very different this season and this one hangs on whether you think they’ll sort that away form out on Saturday.
Conversely, Southampton’s performance stats have been dropping after an amazing start, but the model sees averages across the season and their averages are still pretty reasonable. How much weight do you give to form, or believe in luck? The bookmakers have made City firm favourites. My model gives short term form no weight at all (though I’m working to see how optimal that is) and gives past home performances no consideration when a team is playing away (I’m working on that, too). The model thinks Southampton could well edge the game, based on season averages so far.
You pay your money and you take your choice…
Percentages for the rest of the games don’t look too surprising. Here are the model’s deeper stats on possession and goal scoring.
And if you’re betting (and RvP is out)
Manchester United vs Newcastle United – close, but Away win
Crystal Palace vs Cardiff City – Away win
Liverpool vs West Ham United – Home win
Southampton vs Manchester City – Home win (gulp)
Stoke City vs Chelsea – Away win
West Bromwich Albion vs Norwich City – Home win
Sunderland vs Tottenham Hotspur – Away win
Fulham vs Aston Villa – Draw
Arsenal vs Everton – Home win
Swansea City vs Hull City – Home win
I’ve got a feeling that the EPL Model’s either going to gain a whole load of new converts this weekend, or make me look rather foolish! I’ll try to tweet a Man Utd prediction on @data_monkey using the actual starting-line ups for that game once they’re announced. The latest news now has Kagawa potentially unwell too, so God knows who’s playing. Good luck with your bets and I’ll see you next week.