This is the third and final part of our exercise in humility — evaluating and assessing our league projections from the 2013-14 European season. At some point in the future I’ll post the projected league tables, but I wanted to write my assessments and besides the 2014-15 season has just started.
This was an expected league model that I got horribly, horribly wrong. It’s a funny thing — three teams matched their expected point totals, but none matched their expected positions, and the expectations of goal statistics were a total mess. Eibar, who we predicted to finish relegated, instead finished first. Sporting de Gijón, who we predicted would finish first, instead came in fifth, and then missed out on promotion in the playoffs. We got Jaén right, and I guess we got Murcia right as well, but we don’t count administrative relegations. In contrast to other leagues, the range of expected points in the division was not too low at the top and bottom. We did underestimate the number of points that would bring safety from relegation, but I guess if you factor in the typical error about the estimate, the estimated ‘safe’ point total wouldn’t be too far from the actual. In the end, a point RMSE of 11.9 and position RMSE of 8.6 means that there is a lot more work to be done, starting with acquiring some knowledge of the squads.
France (Ligue 2)
This expected league model was all over the place with respect to the actual table. I did get Lens in the top two — and they had to jump through legal hoops just to get promoted into Ligue 1! — but the only position I got right was CA Bastia at rock bottom. Not to say that the point projections were way off. A points RMSE of 9.07 is decent for a 20-team competition. But being off by a certain number of points has distinct effects at different ends of the table. Caen performed seven points above the expected total and advanced two spots from an expected fifth to third, while Brest also performed seven points above expected and advanced six spots. Promotion was a matter of sides either performed much better defensively than expected (Metz) or much better offensively than expected (Caen). Istres’ wretched season was saved by the French league authorities, as they allowed 37 more goals than expected and finished 18 points worse than expected on their way to a second-from-bottom finish. But from reading some French football sites, Istres was considered a relegation candidate by some, so it could be a case of an over-optimistic forecast on our part.
Germany (2. Bundesliga)
A big fat miss across the board. Our picks for the top two (Kaiserslautern and Fortuna Düsseldorf) finished outside the promotion zone, our candidates for relegation (Ingolstadt and Sandhausen) finished in mid-table, and the top and bottom three performed at least ten points better or worse than expected. But we did get Karlsruhe, 1860, and FC Union Berlin in their expected places, so there’s that.
Germany (3. Liga)
We did a little better here as we projected two of the three promoted sides (Heidenheim and
Red Bull RasenBall Sport Leipzig) and one of the relegated sides (Elversberg). Darmstadt were the big surprise in the 3.Liga as they finished 27 points ahead of expectations to finish third. Preussen Münster had a very disappointing season on both sides of the ball to finish fourth (we had tipped them for relegation). The projection model was more optimistic at the bottom of the table and less so toward the top, which could explain why running ahead of expectations by 7-13 points in the case of Stuttgarter Kickers and Hallescher FC had such a big effect in the league table.
We correctly predicted the league champion Olympiacos, which isn’t too special because they win the league every year. We also predicted five of the top six teams, and three of the top four exactly. The team we tipped for fifth place, Aris, suffered a disastrous season and finished bottom by a wide margin. It looks like the usual problems for a team of this size and with these expectations: heavy debt, unpaid players, forced to play academy/amateur players, etc. Our predictions for relegation candidates completely blew up as OFI, Ergotelis and Panthrakikos finished in the top ten. It seems that Champions League/Europa League positions were predictable and everything else was a crapshoot.
This was one of our better-performing projections in teams of RMSE, but the Ekstraklasa splits into two groups after 30 matches and our prediction powers were quite limited in that respect. We projected correctly Legia, Lech, Wisła, and Górnik in the championship group — the teams whose positions we projected correctly. We projected Cracovia, Jagiellonia, Podbeskidzie, and Widzew in the relegation group, but those projected positions were all over the place. Just another reminder that a lot changes over nine months.
Russia (Premier League)
With the exception of Anzhi Makhachkala, this league projection went very well. We correctly projected the top two teams (CSKA and Zenit) and three of the four relegated teams (Tom Tomsk, Krylia Sovetov, Volga NN). Anzhi probably would not have been a surprise if the squad had been updated in mid-season, but in my opinion, Lokomotiv and Krasnodar were the real surprise outfits of the Premier League. Lokomotiv played well above offensive and defensive expectations to finish third (20 points above expected total), which Krasnodar scored 15 more goals than expected to finish fifth (14 points above expected total). Once again the points RMSE was above ten, which I think would go down with updated squad information and a better understanding of how scoring statistics translate between leagues.
So here’s a statistical summary of the expected league models, sorted by points RMSE. There will be a separate post to summarize all of the league models at once.
|Total Goals||Pred Goals||Pct Change||RMSE Pts||RMSE GF||RMSE GA||RMSE Pos|