Premier League Pythagorean perfectly perplexing
Categories: Soccer Pythagorean: Tables
I didn’t do much updating on the ResultsPage site this season so I didn’t track the Pythagorean table as I did in previous seasons. But now that the 2013-14 English Premier League is over, I want to take a look at the final Pythagorean table to see if the over- and under-performing teams match up with the conventional wisdom. The results seem counter-intuitive but I’m starting to wrap my mind around them after studying them for a while.
Here is the table:
League Table | Pythagorean | ||||||||||||
Team | GP | W | D | L | GF | GA | GD | Pts | W | D | L | Pts | Δ |
Manchester City | 38 | 27 | 5 | 6 | 102 | 37 | 65 | 86 | 26 | 6 | 6 | 84 | +2 |
Liverpool | 38 | 26 | 6 | 6 | 101 | 50 | 51 | 84 | 24 | 7 | 7 | 79 | +5 |
Chelsea | 38 | 25 | 7 | 6 | 71 | 27 | 44 | 82 | 24 | 9 | 5 | 81 | +1 |
Arsenal | 38 | 24 | 7 | 7 | 68 | 41 | 27 | 79 | 20 | 9 | 9 | 69 | +10 |
Everton | 38 | 21 | 9 | 8 | 61 | 39 | 22 | 72 | 19 | 10 | 9 | 67 | +5 |
Tottenham Hotspur | 38 | 21 | 6 | 11 | 55 | 51 | 4 | 69 | 15 | 10 | 13 | 55 | +14 |
Manchester United | 38 | 19 | 7 | 12 | 64 | 43 | 21 | 64 | 19 | 9 | 10 | 66 | -2 |
Southampton | 38 | 15 | 11 | 12 | 54 | 46 | 8 | 56 | 16 | 10 | 12 | 58 | -2 |
Stoke City | 38 | 13 | 11 | 14 | 45 | 52 | -7 | 50 | 12 | 10 | 16 | 46 | +4 |
Newcastle United | 38 | 15 | 4 | 19 | 43 | 59 | -16 | 49 | 11 | 10 | 17 | 43 | +6 |
Crystal Palace | 38 | 13 | 6 | 19 | 33 | 48 | -15 | 45 | 10 | 11 | 17 | 41 | +4 |
Swansea City | 38 | 11 | 9 | 18 | 54 | 54 | 0 | 42 | 14 | 10 | 14 | 52 | -10 |
West Ham United | 38 | 11 | 7 | 20 | 40 | 51 | -11 | 40 | 11 | 11 | 16 | 44 | -4 |
Sunderland | 38 | 10 | 8 | 20 | 41 | 60 | -19 | 38 | 10 | 10 | 18 | 40 | -2 |
Aston Villa | 38 | 10 | 8 | 20 | 39 | 61 | -22 | 38 | 9 | 10 | 19 | 37 | +1 |
Hull City | 38 | 10 | 7 | 21 | 38 | 53 | -15 | 37 | 10 | 10 | 18 | 40 | -3 |
West Bromwich Albion | 38 | 7 | 15 | 16 | 43 | 59 | -16 | 36 | 11 | 10 | 17 | 43 | -7 |
Norwich City | 38 | 8 | 9 | 21 | 28 | 62 | -34 | 33 | 6 | 10 | 22 | 28 | +5 |
Fulham | 38 | 9 | 5 | 24 | 40 | 85 | -45 | 32 | 7 | 8 | 23 | 29 | +3 |
Cardiff City | 38 | 7 | 9 | 22 | 32 | 74 | -42 | 30 | 6 | 8 | 24 | 26 | +4 |
Manchester City are the new champions of England, in the official table and the Pythagorean one as well. They had almost exactly the kind of performance that you would expect from a team with their kind of goal record. Chelsea’s performance was also in line with their statistical expectations. Liverpool’s scoring attack powered them into second place — and put them within touching distance of the title — yet their overall statistics indicate that maybe they deserved to finish third.
The bottom three teams — Norwich City, Fulham, and Cardiff City — are also identical in the official and Pythagorean tables. All three teams scored three to five points better than an average team with their goalscoring records, but all of them suffered at least two defeats by four goals or more.
There were quite a few heavy results which appeared to skew the expectations. Arsenal, who led the league for the longest period, appeared to play way above their station with a Pythagorean residual of +10 but that number is most likely skewed by those three heavy defeats by Manchester City, Liverpool, and Chelsea (aggregate away results 4 goals scored, 17 goals allowed). Tottenham also had an extremely high Pythagorean residual of +14, but they suffered heavy losses to Liverpool (twice!), Manchester City (twice!), and Chelsea (away). Aggregate record from those matches: 1 goal scored, 24 goals allowed. Stoke and Newcastle also had very positive Pythagorean residuals (+4 and +6, respectively), and they allowed four goals or more multiple times in League matches.
That leaves Crystal Palace and Everton and their Pythagorean residuals of +4 each. I’m willing to believe that these residuals aren’t skewed by heavy losses for a couple of reasons: (1) neither side let in three goals very often, and (2) both sides allowed four goals just once — Everton to Liverpool and Palace to Fulham (!). So when it comes down to assessing team overperformance by Pythagorean residuals, the teams whose residuals I give the greatest weight are Liverpool, Everton, and Crystal Palace. As it happens, two of the three managers involved with those teams will receive Manager of the Year awards this week.
This season was characterized by most of the teams overperforming, kind of like most of the kids in Lake Wobegon being just above average. West Brom’s -7 Pythagorean residual is significant in my opinion; maybe Pepe Mel and the board sensed that the team significantly underperformed despite avoiding relegation. The other sides — Manchester United, Southampton, West Ham, Sunderland, Hull City — underperformed a bit, but it’s hard to say how significant those numbers were.
I guess this season shows that there is a difference between a team’s performance relative to expected performance of an average team and a performance relative to expected performance of a team in that specific competition. I know how to capture the former; I need to think more about how to capture the latter.