Well last week was definitely more like it! We’ve been due one of those.
I mentioned early in this series of posts that my predictive model was developed through the second half of last season and so in the past, it’s always had plenty of data to work with. Starting from the 10th week of this season was a bit of an experiment and has led to more volatile performances than it’s seen before and a couple of disappointing weeks.
Last week finally returned a really strong set of predictions – seven out of ten correct and a profit of 54% if you were following the model’s calls for betting.
Eyebrows were raised at the prediction of Man City to comfortably beat Arsenal, but that proved to be accurate (though Arsenal scored more goals than expected). Newcastle and West Ham were predicted by the model to win rather than draw but both predictions were realistic.
The only one that really missed – and missed by a lot – was Spurs to beat Liverpool. You can’t win them all. It is worth making the point that the model doesn’t currently do yellow and red cards, so it can’t see that the scoreline in a game might be affected by a sending off. Even with this feature incorporated within the model, a straight red is a low probability event and would be very hard to see coming in a specific game. I’m absolutely not saying that Paulinho’s dismissal affected the overall result, but it may well have contributed to the number of goals Liverpool scored and the size of the win.
Here are this week’s percentages:
There’s one very obvious surprise in that list and it goes to show the big disparity between Manchester City’s home and away performances this season. The model doesn’t use home statistics to model away results, or vice versa, so only Manchester City’s away performances are taken into account to generate that prediction. At home, Man City have conceded seven goals for every hundred shots faced this season. Away, it’s more than double that rate, at fifteen. This is a big part of what’s driving a predicted win for Fulham – they’re expected to score more than once. Aguero’s injury is also contributing and making Manchester City less likely to score.
In all honesty, I doubt this forecast as I’m sure you do, but it’s definitely interesting that Man City’s away numbers are poor enough for it to happen. It’s also worth bearing in mind that two weeks ago, the model said Southampton would beat Man City and while the win didn’t happen, that game was a draw (contrasting the bookies’ expectation of an easy away win) and could easily have been three points for Southampton.
As always, we’ll have to wait and see what happens…
Here are the deeper simulation stats on shooting and possession:
And if you’re betting:
Liverpool v Cardiff City – Home win
Crystal Palace v Newcastle United – Away win
Fulham v Manchester City – Home win
Manchester United v West Ham United – Draw
Stoke City v Aston Villa – Draw
Sunderland v Norwich City – Home win
West Bromwich Albion v Hull City – Home win
Southampton v Tottenham Hotspur – Home win
Swansea City v Everton – Away win
Arsenal v Chelsea – Home win
The Manchester United bet needs a quick mention here. Man U are the most likely to win this game, but the likelihood of a draw is high enough to be worth betting on. This tends to happen when expected scoring rates are low and is a combination of Man U’s low shot rates at home and West Ham’s combined decent defence and low chance of scoring themselves.
Good luck and happy Christmas to all and see you in a few days for the Boxing Day round.