Lower league forecasts, January 22–23
Creating forecasts is important, and it can also be really interesting too — especially when it comes to sport. In the Department of Economics at the University of Reading, we’ve been creating forecasts for football matches for a number of years now, using a Poisson regression method, estimated on Elo strengths and predictions, plus some other information on the recent scoring record of the teams involved in a match. This kind of model is often thought of to be in a Dixon-Coles class of models, of which there are many.
In other words, we’re by no means somehow special in doing this. Many others do it. Why do we do it, and especially for the lower leagues? Well, one answer is that I’m a fan of a lower league team, Oldham Athletic. They’ve languished in the lower leagues since the late 1990s now, spending 21 years in League One before being relegated in 2018 to League Two.
To cushion the blow of the likely Saturday afternoon defeat, I like to brace myself by understanding what was likely in that match. The methods I use for that might just as well be used to produce forecasts for all matches in a division, and so that’s what we’ve done for some time now.
We produce tables for each division’s fixtures. Here’s the table for League One:
It’s automatic. So I collect the data from Soccerbase, and the odds from Oddsportal. It means that, in the current Covid age, plus it being the middle of winter, gaps appear. Games at Blackpool, Doncaster and Milton Keynes aren’t happening — at least, according to the prices collected from Oddsportal.
The model is run, and it generates expected goals. So Crewe have an impressive 2.3 expected against AFC Wimbledon, who have just 0.88. How will things turn out though? Lincoln and Oxford are also expected to get above two, and Doncaster would have been close, had their game gone ahead.
What about League Two? League One’s lesser neighbour. I’ve just about started to get used to looking at League Two for Oldham, after those 21 years looking at League One. Messy. Rearranged fixtures mean that Leyton Orient are playing Forest Green tomorrow, but there was a time when they were going to face Crawley away. And our automated code can’t quite cope with that. Sorry, Orient and Forest Green fans. At least the bookies think it’ll be close.
Oldham host Newport County. I can’t deny this is the one I’m most interested in. Oldham have an utterly abysmal home record. The worst in the entire of League Two. Two wins, a draw, and nine defeats. Seven points from a possible 36. And they host second placed Newport County. A Newport County that hasn’t failed to score in its away games this season. Against an Oldham that hasn’t failed to concede in all of its home games.
There will be goals. But hopefully, from my perspective, not too many in the Oldham net. A 1–1 draw is 13% likely, out tables suggest. After the expected goals, it’s the outcome probabilities according to our model. Then there’s the most likely scoreline (usually 1–1), followed by the most likely scoreline given which team is most likely to win. Followed by probabilities of the nine most likely scorelines occurring.
