Here are the 2013 MLS ELO+ ratings and points predictions for 07/21. Code: Rank Team ELO+ Rank Team Points 1 POR 121 1 SKC 58.0 2 SKC 119 2 POR 57.8 3 RSL 117 3 RSL 56.6 4 VAN 115 4 MTL 53.6 5 LAG 109 5 VAN 53.1 6 NYR 107 6 LAG 52.7 7 PHI 107 7 NYR 52.0 8 HOU 106 8 HOU 51.4 9 COL 106 9 PHI 50.2 10 MTL 104 10 DAL 49.3 11 DAL 102 11 NER 47.2 12 SEA 101 12 SEA 47.1 13 NER 98 13 COL 46.8 14 CHI 94 14 CHI 42.5 15 SJE 91 15 CLB 39.1 16 CLB 88 16 SJE 39.0 17 CHV 78 17 CHV 30.6 18 TOR 73 18 TOR 28.1 19 DCU 64 19 DCU 21.9 The average number of games played per team is now 20.2 (K=9.90), and so far home teams have a winning percentage of 0.622 (H=24). The season is far enough along at this point that it is meaningful to start looking at strength of schedule. ELO+ ratings provide a way to do that, by calculating the average rating of each team's opponents so far at the time when they met, the average rating of their remaining opponents, and the combined average. In all cases, home-field advantage is taken into account. Here are the 2013 MLS ELO+ strength of schedule ratings for 07/21. Code: Rank Team Past Rank Team Future Rank Team Overall 1 SEA 105.5 1 COL 107.7 1 CHV 104.3 2 CHV 104.7 2 RSL 106.9 2 SJE 103.1 3 LAG 103.8 3 VAN 104.3 3 SEA 102.3 4 DCU 103.0 4 SJE 104.2 4 DAL 102.2 5 TOR 102.7 5 CHV 103.7 5 COL 101.8 6 SJE 102.5 6 DAL 103.3 6 LAG 101.6 7 POR 102.3 7 CHI 101.5 7 DCU 101.1 8 DAL 101.5 8 CLB 101.2 8 VAN 101.0 9 NER 100.9 9 PHI 98.6 9 RSL 100.9 10 MTL 98.9 10 SEA 98.6 10 POR 100.5 11 VAN 98.8 11 DCU 98.4 11 TOR 100.0 12 HOU 98.7 12 POR 98.0 12 CLB 99.6 13 COL 98.6 13 LAG 97.9 13 CHI 98.8 14 CLB 98.4 14 MTL 97.7 14 MTL 98.4 15 NYR 98.0 15 SKC 96.7 15 HOU 97.7 16 RSL 97.1 16 HOU 96.4 16 PHI 97.2 17 CHI 96.6 17 TOR 96.2 17 NER 97.1 18 PHI 96.3 18 NYR 94.5 18 NYR 96.6 19 SKC 95.3 19 NER 91.8 19 SKC 95.8
Here's my updated rankings: Code: Team Pois EAP RSL 56.8 61.4 POR 56.7 61.2 SKC 56.4 59.3 MON 53.3 49.3 LAG 53.2 54.5 VAN 52.2 52.5 NYR 51.6 52.2 HOU 50.8 50.1 FCD 49.6 46.5 PHI 49.5 49.0 NER 49.0 55.1 SEA 47.7 47.7 COL 47.2 48.5 CHI 42.4 40.7 CLB 41.1 43.8 SJE 39.4 34.1 CHV 32.0 27.5 TOR 30.7 32.6 DCU 23.3 17.7 It looks like RSL dropped 2 points off their Poisson expected end of season with that result, and SKC picked up nearly that many. Right now it is looking like New England is suffering the most from not getting results when it counts, since their Poisson ranking is about 6 points below their full season expected, and they are currently sitting outside of the playoffs. Strangely enough, my model is showing DC as OVERPERFORMING right now when compared to the expected full season.
Trying a simpler table this week. Teams listed as #1 - #19, but numberical rank plotted instead of positional ranking. i.e. might show the relative rankings a bit better. RSL, Portland, and SKC all within one "rank" of each other, with Vancouver 1.5 ranks behind them. System looks better this week, Colorado and Chicago perhaps a bit high, Dallas, Houston and New York perhaps a bit low, harkening back, again to some weighting given recent results rather than overall season long performance. Then again, Dallas is probably on the worst run of form in the league right now... Snapshot captures most recent performance, so Chicago riding their rout high and Houston sitting at #4 despite not playing last week. Kansas City with the sharp rise, Montreal continues to drop. Graph.
So with the losses Kansas City and LA... rise? Wait, what? -overturns table and storms out- Actually, because all five top teams lost, it was more a matter of who was going to fall faster, and in this case it was RSL and Portland, causing Kansas City and LA to apparently rise. One of the flaws of using a numeric relative ranking system I suppose? I've actually figured out that as a moving average model my spreadsheet tends to lag a game or so when turning out ratings, so I've added a "form" category that looks at the most recent result and compares it to the ranking to "predict" movement for the following week. So if SKC and LA moving up gives you a bit of heartburn while NY seemingly doesn't get any love, I would note the system thinks that at this snapshot in time moment, RBNY is playing the second best in the league and trending up, behind only Colorado, who are tearing it up. DC is overperforming at #19 and trending down...
My ratings did the same thing, actually. Also, my ratings were showing that Kansas City is one of the weakest top rated teams since I started my ratings - meaning that overall the parity in the league is very, very, very high, without anyone really standing out. I'll try to post up my numbers tonight.
That makes me feel a bit better I actually think this has been a pretty crazy season, all things considered. Second half of the season should be really good I think, lots of twists and turns.
Here are the 2013 MLS ELO+ ratings and points predictions for 07/29. Code: Rank Team ELO+ Rank Team Points 1 POR 116 1 MTL 56.2 2 SKC 115 2 SKC 55.8 3 RSL 112 3 POR 55.7 4 PHI 112 4 RSL 54.8 5 NYR 111 5 NYR 54.2 6 MTL 109 6 PHI 53.2 7 LAG 109 7 LAG 51.7 8 COL 109 8 HOU 49.7 9 VAN 107 9 NER 49.6 10 HOU 103 10 VAN 49.5 11 NER 103 11 DAL 49.5 12 DAL 103 12 COL 49.1 13 SEA 103 13 SEA 48.8 14 SJE 97 14 CHI 42.3 15 CHI 94 15 SJE 42.0 16 CLB 84 16 CLB 37.1 17 TOR 78 17 TOR 30.5 18 CHV 76 18 CHV 29.7 19 DCU 59 19 DCU 19.4 The average number of games played per team is now 21.2 (K=9.45), and so far home teams have a winning percentage of 0.627 (H=25). Consistent with what others are saying, these numbers suggest a lot of parity across the league--no team is more than 16% better than average, and only DCU is more than 24% below average. Here are the 2013 MLS ELO+ strength of schedule ratings for 07/29. Code: Rank Team Past Rank Team Future Rank Team Overall 1 CHV 105.7 1 COL 107.0 1 CHV 104.0 2 LAG 104.4 2 VAN 106.9 2 SJE 102.1 3 POR 104.0 3 DAL 103.3 3 DCU 101.9 4 SEA 102.6 4 RSL 103.3 4 DAL 101.6 5 SJE 101.6 5 SJE 103.2 5 SEA 101.6 6 DCU 101.5 6 DCU 102.6 6 VAN 101.3 7 TOR 100.9 7 CLB 102.1 7 COL 101.1 8 DAL 100.6 8 CHI 101.4 8 LAG 101.1 9 NER 99.9 9 CHV 101.3 9 RSL 100.8 10 RSL 99.5 10 SEA 100.3 10 POR 100.5 11 CLB 99.3 11 TOR 99.0 11 CLB 100.4 12 MTL 98.9 12 HOU 98.7 12 TOR 100.1 13 COL 98.3 13 MTL 97.9 13 CHI 99.0 14 NYR 98.2 14 PHI 96.0 14 MTL 98.4 15 PHI 98.2 15 SKC 95.5 15 HOU 98.1 16 HOU 97.8 16 LAG 95.1 16 PHI 97.4 17 VAN 97.8 17 POR 94.8 17 NER 97.1 18 CHI 97.3 18 NYR 94.4 18 NYR 96.9 19 SKC 96.9 19 NER 92.7 19 SKC 96.4
Here are the 2013 MLS ELO+ ratings and points predictions for 08/04. Code: Rank Team ELO+ Rank Team Points 1 NYR 118 1 NYR 57.3 2 POR 114 2 RSL 54.9 3 RSL 112 3 POR 54.4 4 COL 109 4 MTL 53.2 5 VAN 108 5 SKC 52.6 6 SKC 107 6 HOU 51.3 7 SEA 107 7 LAG 51.0 8 HOU 106 8 SEA 50.8 9 LAG 106 9 VAN 50.0 10 PHI 104 10 PHI 49.7 11 MTL 103 11 COL 48.6 12 CHI 102 12 DAL 47.5 13 DAL 99 13 NER 46.2 14 SJE 99 14 CHI 46.1 15 NER 97 15 SJE 43.8 16 TOR 85 16 CLB 36.1 17 CLB 82 17 TOR 33.9 18 CHV 74 18 CHV 28.5 19 DCU 68 19 DCU 23.0 The average number of games played per team is now 22.1 (K=9.05), and so far home teams have a winning percentage of 0.624 (H=25). Here are the 2013 MLS ELO+ strength of schedule ratings for 08/04. Code: Rank Team Past Rank Team Future Rank Team Overall 1 CHV 105.6 1 COL 108.0 1 CHV 103.9 2 LAG 104.0 2 SJE 105.2 2 DAL 101.7 3 POR 102.7 3 DCU 103.4 3 SJE 101.7 4 CLB 101.6 4 VAN 103.3 4 DCU 101.6 5 DAL 101.3 5 DAL 102.5 5 LAG 101.1 6 RSL 101.0 6 SEA 102.5 6 SEA 101.0 7 DCU 100.6 7 HOU 101.5 7 COL 100.9 8 TOR 100.4 8 CHV 100.8 8 VAN 100.9 9 SJE 100.1 9 RSL 99.8 9 CLB 100.8 10 SEA 100.0 10 MTL 99.4 10 RSL 100.6 11 VAN 99.6 11 CLB 99.2 11 POR 100.4 12 NER 99.1 12 PHI 98.7 12 TOR 99.4 13 MTL 99.0 13 CHI 98.6 13 MTL 99.1 14 NYR 99.0 14 TOR 97.7 14 PHI 98.4 15 PHI 98.2 15 NER 96.8 15 NER 98.3 16 COL 97.9 16 SKC 96.6 16 CHI 98.1 17 CHI 97.8 17 POR 96.3 17 HOU 98.1 18 SKC 97.6 18 LAG 95.8 18 SKC 97.3 19 HOU 96.0 19 NYR 91.4 19 NYR 96.5
Here are the 2013 MLS ELO+ ratings and points predictions for 08/12. Parity continues to be evident, with all but three teams within 14% of average. Code: Rank Team ELO+ Rank Team Points 1 RSL 114 1 RSL 56.2 2 POR 114 2 NYR 54.8 3 SEA 112 3 POR 54.1 4 NYR 111 4 SKC 53.9 5 VAN 111 5 SEA 53.4 6 SKC 109 6 MTL 51.9 7 CHI 108 7 VAN 51.6 8 LAG 107 8 LAG 51.1 9 COL 107 9 PHI 51.0 10 PHI 106 10 HOU 50.0 11 HOU 103 11 CHI 48.8 12 MTL 101 12 COL 47.6 13 DAL 98 13 DAL 46.6 14 SJE 96 14 NER 44.7 15 NER 93 15 SJE 42.3 16 CLB 87 16 CLB 38.8 17 TOR 81 17 TOR 31.9 18 CHV 77 18 CHV 28.9 19 DCU 65 19 DCU 22.2 The average number of games played per team is now 23.1 (K=8.68), and so far home teams have a winning percentage of 0.630 (H=26). Here are the 2013 MLS ELO+ strength of schedule ratings for 08/12. At the moment, the Red Bulls appear to have the easiest route to the Supporters Shield. Code: Rank Team Past Rank Team Future Rank Team Overall 1 LAG 105.4 1 COL 110.6 1 CHV 104.1 2 CHV 105.3 2 VAN 106.0 2 SJE 102.5 3 POR 102.6 3 DAL 105.4 3 DAL 102.3 4 SJE 102.1 4 SJE 103.4 4 COL 101.5 5 DCU 101.8 5 RSL 103.0 5 LAG 101.5 6 DAL 100.8 6 SEA 102.5 6 DCU 101.3 7 NER 100.3 7 PHI 102.1 7 SEA 101.1 8 MTL 100.2 8 CHV 101.7 8 VAN 101.1 9 SEA 100.2 9 DCU 100.3 9 POR 100.9 10 CLB 100.0 10 SKC 99.8 10 RSL 100.9 11 RSL 100.0 11 CLB 99.5 11 CLB 99.9 12 NYR 99.3 12 TOR 99.4 12 TOR 99.2 13 TOR 99.2 13 CHI 98.8 13 MTL 99.0 14 VAN 98.8 14 HOU 98.4 14 NER 98.2 15 COL 98.2 15 POR 97.7 15 HOU 98.0 16 HOU 97.8 16 MTL 96.9 16 PHI 97.6 17 CHI 96.6 17 NER 93.8 17 CHI 97.4 18 PHI 95.8 18 LAG 93.4 18 SKC 97.0 19 SKC 95.8 19 NYR 89.4 19 NYR 96.4
Missed last week... oops. Entirely on me. System is in love with Philadelphia right now, and I can't quite explain it. I think I'm coming to terms with the fact that it's lagging a week or so, but I like the predicator column. For Week 23, it correctly "predicted" the movement of teams in Week 24 13 out of 19 times. Dallas and Montreal two of the coldest teams in MLS right now. Who would've thunk it at the start of the season?
RSL is leading everyone in Elo with a rating of... 1600?! The maximum rating continues to decrease as the league compresses itself. Only seven teams are <1500 while the other twelve are between 1500 and 1600. If you check out the End of Season Points Predicted, RSL is on top with a predicted score of 58, and Chicago is sitting in eleventh overall place with a predicted score of 48.6. Less than ten points separating the Supporters Shield winner with non-playoff spots? It's a distinct possibility this year. Chicago and Colorado are on a slow rise, Montreal and Columbus are really starting to sink, and RSL, SKC, SS, and LAG are all potentially one game away from siting on top of the ratings. It's going to be a crazy finish to the year.
Colorado and Philadelphia with the hot hand this week. Kansas City seems generous at #4 and San Jose quickly rising after four wins from their last five. Columbus beats Toronto but can't quite get a head of them, though it looks like they should make a big move next week with a snapshot rating of 5. Wheels are totally gone for Dallas, they're approaching DC territory. Overall the season is incredibly tight, and while it seems like we could say it every week, it should be an exciting end to the season. Prime example, Kansas City could have taken the SS lead on points last night, but instead are now only 4 points above the post season cut, and other teams have games in hand...
Here's the models I occasionally update. Both models use the same underlying concept: a poisson model for each individual game. It predicts the probability of each outcome of a match, and calculates the expected points from that match. For the POS model, it uses the current standings, accounts for home field advantage, and simulates the actual remaining schedule. Basically, it is an end-of-season prediction model. For the EAP model, it puts the team against an "average" opponent, and produces a single-game expected number of points. In order to make differences more apparent, I have multiplied that by the 34 games in an MLS season. Basically, this is a "power" rating of a team so far this season. Code: Team POS EAP POR 55.4 59.2 RSL 55.0 56.9 LAG 53.1 53.4 NYR 53.0 51.3 SKC 52.7 57.2 MON 52.6 47.9 HOU 52.0 52.8 SEA 51.5 50.9 PHI 50.7 50.4 VAN 50.1 50.6 COL 49.5 52.0 NER 48.0 52.5 FCD 46.5 42.8 CHI 45.3 41.6 SJE 43.8 37.6 CLB 41.9 45.3 TOR 32.3 33.7 CHV 30.1 26.9 DCU 23.2 20.9 Even though I have shown it to one decimal place, the data isn't really significant beyond the whole number points. This means that POR/RSL would be expected to land on the same number of points at the end of the season, as would LAG/NYR/SKC/MON. For the west, the teams currently predicted for the playoffs by POS: POR - 55 RSL - 55 LAG - 53 SEA - 52 VAN - 50 If these positions were to hold up, the wild card game would be Vancouver at Seattle. The winner of that game would then play Portland Home/Away. That would make for quite the interesting playoff schedule... For the east: NYR - 53 SKC - 53 MON - 53 HOU - 52 PHI - 51 Interestingly, in the east, changing a single loss into a win could theoretically put you from out of the playoffs to first. This is an extremely tight playoff race, and should be an interesting final stretch, almost no matter who you support.
10 teams within 5 points of each other. I love it. This has been a really good year for soccer fans I feel.
Here are the 2013 MLS ELO+ ratings and points predictions for 08/19. Code: Rank Team ELO+ Rank Team Points 1 POR 118 1 POR 56.2 2 RSL 111 2 RSL 54.7 3 NYR 111 3 NYR 54.2 4 COL 109 4 MTL 52.8 5 LAG 108 5 LAG 52.7 6 SEA 108 6 SKC 52.4 7 VAN 108 7 SEA 51.7 8 SKC 107 8 HOU 51.6 9 PHI 107 9 PHI 51.2 10 HOU 105 10 VAN 50.1 11 MTL 103 11 COL 49.2 12 CHI 103 12 CHI 46.8 13 SJE 100 13 NER 46.8 14 NER 97 14 DAL 45.7 15 DAL 96 15 SJE 44.4 16 CLB 89 16 CLB 40.3 17 TOR 78 17 TOR 30.6 18 CHV 77 18 CHV 28.9 19 DCU 65 19 DCU 21.8 The average number of games played per team is now 24.0 (K=8.33), and so far home teams have a winning percentage of 0.643 (H=29). Here are the 2013 MLS ELO+ strength of schedule ratings for 08/19. Code: Rank Team Past Rank Team Future Rank Team Overall 1 CHV 105.2 1 COL 113.9 1 CHV 104.1 2 LAG 104.5 2 SJE 105.4 2 DAL 102.4 3 DCU 103.0 3 VAN 103.9 3 SJE 102.1 4 DAL 102.8 4 CLB 103.4 4 LAG 101.4 5 SEA 102.2 5 MTL 103.0 5 SEA 101.4 6 POR 101.7 6 CHV 101.7 6 VAN 101.4 7 RSL 101.4 7 DAL 101.6 7 COL 101.3 8 SJE 100.8 8 RSL 100.3 8 DCU 101.3 9 TOR 100.4 9 HOU 100.2 9 RSL 101.1 10 VAN 100.3 10 SEA 99.8 10 POR 100.5 11 NYR 99.3 11 PHI 99.3 11 CLB 99.7 12 NER 99.0 12 POR 97.9 12 TOR 99.6 13 CHI 98.5 13 SKC 97.9 13 MTL 98.8 14 CLB 98.2 14 TOR 97.9 14 CHI 98.0 15 COL 97.4 15 DCU 97.1 15 HOU 97.9 16 HOU 96.9 16 CHI 96.9 16 NER 97.9 17 MTL 96.8 17 NER 95.3 17 PHI 97.5 18 PHI 96.8 18 LAG 93.8 18 SKC 97.1 19 SKC 96.8 19 NYR 89.4 19 NYR 96.7
Here are the 2013 MLS ELO+ ratings and points predictions for 08/26. Code: Rank Team ELO+ Rank Team Points 1 RSL 115 1 RSL 56.3 2 LAG 114 2 LAG 55.6 3 POR 112 3 MTL 54.4 4 SEA 111 4 SEA 53.5 5 COL 109 5 POR 53.2 6 CHI 106 6 NYR 51.9 7 MTL 105 7 SKC 51.3 8 NYR 105 8 HOU 50.2 9 SKC 105 9 PHI 49.7 10 PHI 103 10 COL 49.3 11 VAN 103 11 NER 48.7 12 HOU 101 12 CHI 48.5 13 NER 101 13 VAN 47.6 14 SJE 100 14 DAL 47.1 15 DAL 99 15 SJE 44.3 16 CLB 90 16 CLB 39.7 17 TOR 79 17 TOR 30.7 18 CHV 77 18 CHV 29.3 19 DCU 65 19 DCU 21.3 The average number of games played per team is now 25.2 (K=7.95), and so far home teams have a winning percentage of 0.644 (H=29). Here are the 2013 MLS ELO+ strength of schedule ratings for 08/26. Code: Rank Team Past Rank Team Future Rank Team Overall 1 LAG 105.4 1 COL 113.1 1 CHV 104.4 2 CHV 103.5 2 VAN 107.3 2 DAL 102.4 3 POR 102.9 3 CHV 106.9 3 SJE 102.4 4 RSL 101.9 4 MTL 105.9 4 VAN 102.1 5 SJE 101.8 5 DAL 104.9 5 COL 101.6 6 DAL 101.7 6 SJE 104.3 6 SEA 101.4 7 SEA 101.4 7 DCU 102.2 7 POR 101.3 8 DCU 100.4 8 SEA 101.4 8 LAG 101.1 9 VAN 100.2 9 CLB 100.6 9 RSL 101.0 10 TOR 99.9 10 CHI 99.3 10 DCU 100.9 11 NYR 99.6 11 RSL 97.9 11 CLB 99.5 12 CLB 99.1 12 TOR 97.7 12 TOR 99.3 13 HOU 98.9 13 NER 97.2 13 MTL 98.2 14 SKC 98.4 14 POR 96.8 14 HOU 98.1 15 PHI 98.3 15 HOU 96.2 15 CHI 97.6 16 COL 98.0 16 PHI 95.6 16 PHI 97.6 17 NER 97.3 17 SKC 92.5 17 NER 97.3 18 CHI 96.9 18 LAG 89.2 18 SKC 97.0 19 MTL 95.0 19 NYR 88.1 19 NYR 96.9
Wasn't really sure what to expect after a crazy crazy week, but this actually seems about right. Seattle with a big jump, SKC with a big fall, Chivas not quite out of the cellar yet, but gaining momentum, and... I don't even really know what kind of commentary to add. Colorado not playing was maybe the only thing that kept them safe, given just how bizarre the weekend was. Everyone at the top and bottom seems to be upwardly mobile while everyone in the middle seems to be struggling.
I checked out results really quickly the other day and was surprised to see LA land at #1 for me (although by a slim margin). I was curious as to whether or not anyone else would get anything like that, interesting to see they are a very close second for you as well.
Are you looking mainly at the ratings or the points projections? The interesting thing about the ratings is that yours still have a small lingering effect from last season, which ended with LAG's run to MLS Cup, while mine only incorporate this year's results. I suspect that LAG is also near the top in my points projections because they have the second-easiest remaining schedule based on current ratings.