Simon Gleave posted an article earlier about Arsenal’s finish to the season – and how dominant they were compared to the rest of the league over the final 14 matches. This led to some discussion between myself, Simon, and Sander IJtsma as to just how rare their feat of scoring 35 points over a 14 match span is, and whether choosing 14 matches is a valid cut off point.
Now any cut-off is arbitrary (just as the season being 38 games long is), but Arsenal’s odds of exceeding 72.5 points were obviously greater following their win over Stoke in game number 25 than they were before that happened. So if we cut the data with 13 games remeaining we’d expect to see Arsenal with a greater chacne of hitting their goal than we do with 14 games remaining. So if we track the proportion of times that a span of Premiership games will yield enough point for Arsenal to get past 72.5 after each game has finished then what do we see?
The method I’m using is the same as that used earlier in the season to track the races for the title, a top-four place, and to avoid relegation. Rather than re-hasing the details I’ll link you to here, where the methodology is explained. In this case I’m concentrating only on the ‘spans’ section.
Let’s start from the mid-point of the season – at which point Arsenal had 33 points from 19 matches, and thus need a span of >39.5 points from the remaining matches to hit at least 73 points.
What do we see? Well the pattern is as would be rationally expected, each time Arsenal win a game their odds of reaching the target improve, a draw causes a slight decline, whereas each defeat almost cripples Arsenal’s remaining chances.
Until the final few games of the season Arsenal’s chances were solidly below 10%, reaching a minima of 1.9% with 10 games remaining, and wherever we cut the data they had a pretty good run.
The major drawback of this exercise? Well Arsenal are obviously a better than average team, and this analysis encompasses all Premiership teams. Because of that their odds of exceeding 72.5 points were better than shown here (how much is certainly open to debate) but they still weren’t ever really great.
Edit: It appears we’re not the only ones to be looking at this – Mark Taylor’s piece from May 21st can be found here.