The rest of the League Two season
Back at the turn of the year, I ran some simulations of the remainder of the League Two season to get a sense of how bad things were for my team, Oldham Athletic, who at that point were bottom of League Two, and hence of the entire 92-team Football League and Premier League structure. Back then, Oldham had a 68% chance of being relegated and falling out of the Football League.
In the time since, things got yet worse and by January 22 Oldham were seven points from safety, having accrued a mere 18 points in 25 matches. Only in the 1968/69 season has the club ever had fewer points (16) at this stage of a season (correcting for 3 points for a win). But they then got better: the club appointed legend John Sheridan as manager on January 22nd.
Since then, the team has gone six matches unbeaten, and managed to emerge briefly from the relegation zone. There is actually a fight against relegation now, and so it seemed like a good moment to re-do the simulations. What are the chances now of relegation?
As a reminder, the method is as follows: assume that teams scoring goals in football matches follow a statistical pattern called a Poisson process. This is a rough approximation but a common one. The process assumes a team strength for each team in a match, and that team strength can be a function of past strength, recent form, managerial changes, and so on.
Having worked out current strengths for teams for matches in the remainder of the season, we can then generate match outcomes for all of the matches left. That is, pick numbers from the Poisson distribution that are the goals scored in all of the remaining matches. Hence we pick scorelines for the rest of the season, then tally up the total points that each team wins, and calculate the final league table.
We thus simulate the rest of the season — we generate an alternative reality inside our computer. We then do it again, and again, and again. The more times, the more reliable are the results, in principle.
The upshot is that after a six-match unbeaten run, Oldham’s likelihood of being relegated has fallen from somewhere north of 68% (I calculated that on January 2 hence before the 3–0 defeat at Harrogate on January 22) to 25%.
There’s now apparently a 3-team battle for the second relegation spot. Scunthorpe, eight points from safety, are now at 98% for the drop, but Oldham, Stevenage and Barrow are all around 25–29% likely to take the second relegation spot. Oldham were at 27% before the 1–1 draw at Colchester (6–7% likely to go down).
A related question for fans of teams down there at the bottom is: How many points will be needed to avoid relegation? One way to look at this is to look at past seasons and how many points the team in 22nd got. What we can do though is look at past seasons — since 2002/03 two teams have been relegated, and discounting the 2019/20 season that was incomplete, the lowest points total avoiding relegation was 43 points (2015/16), while the highest was 52 (2005/06). 48 points has achieved safety in 5 seasons, 47 in two, and 46 in two.
That’s only 17 observations though — a small sample. An alternative strategy is to make up data to create more observations and a large sample. By making up, I mean run the simulations as described above, and see how many points the team in 22nd got each season. The advantage of this method is that it factors in special characteristics of the current season — how many points have teams at the bottom got already?
The plot above shows the distribution of points for the team in 22nd in the 2,000 simulations I used. Hence 44 points is most common, at about 15% of the time. Getting 44 points, however, to interpret this properly, means that about 50% of the time that will be enough to avoid relegation. Hence any club needs to aim higher than 44.
One more win, and 47 points, gives a 90% chance that the team getting that many points avoids relegation. Oldham are on 30 points with 15 matches remaining. A point a match, and 45 points, gives a 70% chance of avoiding relegation. 17 points would be 5 wins and 2 draws, or 4 wins and 5 draws, or 3 wins and 8 draws. Minimum targets to aim for…
