Here are the 2013 MLS ELO+ ratings and points predictions for 06/16. Code: Rank Team ELO+ Rank Team Points 1 POR 128 1 MTL 63.8 2 MTL 121 2 POR 59.6 3 DAL 118 3 DAL 57.8 4 SEA 118 4 SEA 55.4 5 RSL 114 5 RSL 53.2 6 NYR 110 6 NYR 52.9 7 VAN 109 7 HOU 49.9 8 HOU 103 8 PHI 49.0 9 PHI 102 9 VAN 48.5 10 LAG 101 10 SKC 47.7 11 CLB 101 11 NER 46.9 12 COL 100 12 LAG 46.5 13 SKC 99 13 CLB 45.7 14 NER 98 14 COL 44.0 15 SJE 96 15 SJE 41.2 16 CHI 84 16 CHI 35.7 17 TOR 76 17 TOR 31.3 18 CHV 68 18 CHV 27.7 19 DCU 54 19 DCU 19.3 The average number of games played per team is now 14.5 (K=13.8), and so far home teams have a winning percentage of 0.601 (H=20).
My rankings are based on a model which measures the following factors: Offensive and defensive strengths for each team Home advantage A general time dependence within each game (I am finding that the probability of scoring tends to increase as the game progresses) Changing probability of scoring based on the current score of a game. The general trend is that teams that are behind have an enhanced probability of scoring while those ahead tend to have a slightly decreased probability. The rankings are based on Monte Carlo simulations of all remaining games this season. I welcome any comments on these results, mostly because I am still trying to make sure my code is fully debugged.
Okay, so here's last weeks ratings. Glad I'm not gonna fall too far behind, as the games next week are still gonna be part of Week 16. Had another daughter on Monday (second spawn). Home on Friday. Running on a good 10 hours of sleep in the last 72. Weeee! I love late night coffee. Anyway, here's the numbers. Hard to take anything too seriously, as most teams "cruised" on their Week 14 rankings, but still, looks pretty good. The gulf between the top 2/3rds of the league and the bottom third (Chicago, San Jose, Chivas, Toronto and DC) is actually pretty impressive... Make of that what you will. The table. The graph.
I am planning on writing up my methods and results once I have run the code on several more years of data.
Okay, getting caught up again. Numbers through June 19th. No commentary other than whoever said that having your second kid is easier than having your first needs to get punched in the dick until they're infertile. Movement might look a little screwy given the weird scheduling. Next week may not be better... Anyway, rankings... Graph. I'm going to bed.
ESPN Power Rankings: June 24 http://espnfc.com/powerrankings/_/year/2013/week/17?cc=5901 Strong As Oak
They've got Dallas, who managed to come back from 2-0 down this past weekend, behind Seattle, who not only managed to NOT come back from 2-0 down this past weekend, but could barely managed two shots on goal. Never change, ESPN.
Here are the 2013 MLS ELO+ ratings and points predictions for 06/24. Code: Rank Team ELO+ Rank Team Points 1 POR 128 1 MTL 65.6 2 MTL 125 2 POR 60.3 3 RSL 119 3 RSL 56.5 4 DAL 114 4 DAL 55.9 5 SEA 113 5 SEA 53.1 6 VAN 109 6 PHI 51.6 7 PHI 107 7 NYR 50.4 8 LAG 107 8 LAG 50.3 9 NYR 105 9 VAN 49.3 10 SKC 101 10 SKC 48.3 11 CHI 99 11 NER 46.7 12 NER 98 12 HOU 46.4 13 HOU 98 13 CHI 43.6 14 CLB 93 14 CLB 41.4 15 COL 91 15 COL 39.9 16 SJE 87 16 SJE 37.2 17 TOR 80 17 TOR 32.3 18 CHV 63 18 CHV 25.1 19 DCU 63 19 DCU 22.9 The average number of games played per team is now 15.8 (K=12.7), and so far home teams have a winning percentage of 0.610 (H=22).
Shrug. Dallas was at home and eeked out a draw after SKC went down a man for an extended period. Seattle was on the road and lost to an RSL team that's now got the most points in the league. It's close between those two, but FCD has no MLS wins in June.......................[do have that win against Houston in the USOC] I'd put FCD ahead of Seattle as they're got about 8 more points in the table if I remember correctly.
Don't get me wrong. I care not one whit about subjective "power rankings," since the rankings that matter are at the end of the season and at the end of the MLS Cup final. But they do amuse me, since they appear to be a chance for America's soccer writers to show off their own personal biases as if they are somehow an authoritative opinion on who's good and who's not. Fortunately, they have zero impact on the rankings that matter, unlike a popular sporting competition in this country that actually determines its champion based on silliness like "the eyeball test."
Here are my updated rankings. I decided to change how I calculate them slightly. Before, I added the probabilities for each unplayed game to the current records of each team. This is fine if the point is predicting where teams will end up, but I feel like these rankings are more about which team is better. To remove any bias caused by imbalanced schedules, I have decided to simulate the outcomes as if the season was a home-away, 36-game schedule. The first thing I noticed is how high the Revs are in these rankings. I think their 5-0 win over the Galaxy probably makes them look a bit better than they are. My method is based entirely on goals (the model does not explicitly concern itself with who wins or loses games) so a blowout can make a big difference. The next thing I notice is how bad the ranking is for FC Dallas. There is a little bit of noise in these numbers because they were derived from a Monte Carlo simulation, but probably not enough to explain their low numbers. The number of goals predicted is 890.4 for an average of 2.60 per game (compared to 2.64 per game last season). The number of draws is 87.66 (26%) compared to 24% which ended in draws last year. These are both nice checks to make sure my code is behaving reasonably.
Here are the 2013 MLS ELO+ ratings and points predictions for 06/30. Code: Rank Team ELO+ Rank Team Points 1 POR 128 1 MTL 60.1 2 RSL 124 2 POR 59.9 3 DAL 115 3 RSL 58.9 4 VAN 114 4 DAL 55.8 5 MTL 113 5 NYR 52.5 6 SEA 112 6 SEA 52.4 7 NYR 108 7 VAN 51.6 8 PHI 106 8 PHI 50.9 9 SKC 105 9 SKC 50.7 10 LAG 100 10 LAG 46.8 11 CHI 99 11 NER 46.8 12 COL 99 12 HOU 45.0 13 NER 98 13 CHI 44.0 14 HOU 95 14 COL 44.0 15 SJE 93 15 SJE 39.9 16 CLB 89 16 CLB 39.7 17 TOR 76 17 TOR 30.5 18 CHV 67 18 CHV 25.9 19 DCU 59 19 DCU 21.4 The average number of games played per team is now 16.6 (K=12.0), and so far home teams have a winning percentage of 0.604 (H=21).
Way behind now, apologies. Lots of weird movement in the middle this week. I think callups and the like made things screwy, combined with a more dynamic model and a lot of teams seemingly unable to figure out how they want to play this season. Table. Graph.
I'm guessing those rankings are not taking into account this last weekends matches, with Colorado dropping 4 spots after beating Montreal and Montreal climbing back to #1 after losing at home. NY dropping 6 spots after winning, LA climbing 5 spots after losing.
Correct, gonna try to work on those rankings right now. I'm also gonna double check those numbers, as there was a lot more movement than normal. I think maybe the middle of the league really is that tight right now though, so it may not be weird...
Wow, I got busy and/or lazy and it's been awhile since I've posted. I made a change in what I'm displaying here, the change (Chng) is now a full months change, not a week-to-week change. I made that change to try to show a little bit more of the short term form of each team. So, despite Portland not being at the top of my slow moving Elo ratings, you can see that they are up 57 points in the last month, which is way more forward momentum than most other teams. RSL, Portland, Vancouver, and Chicago have had good months and are trending well, Los Angeles, New York, Houston, and DC have had bad months and are trending poorly. Montreal still in the lead for Supporter's Shield according to my numbers, but Real Salt Lake, Seattle, and Portland are within 3.5 points.
Bah. Still more behind than I'd like. All matches current through June 29th. Still a lot of movement in the middle of the table. Still a little wonky due to international stuff and 3 teams off, but looking decent. Look at Chicago go.
Mmmkay, here's all matches up through July 3rd. Gonna try to get the rest done this morning while I've still got coffee in my system and mindless online training to sit through. Seattle the big mover, New England the big loser, still really crowded in the middle but maybe starting to spread a bit. Table. Graph.
Okay, caught up. One thing I'm starting to think is that including the FOTW ranking in the overall ranking is good, but perhaps should not be weighted equally with season long performance, as some screwy results are starting to creep in. Case in point being Houston winning yet dropping two points, while Philadelphia loses and remains in their #5 spot (which is also arguably high). I think there's a bit too much rank skewness going on based on result results and not enough overall big picture trend analysis happening, but I don't plan on changing it again this season. Make a note for next year and move on. The graph is probably a better illustration of relative rankings and movement/trends than the raw table, truth be told.
OK, you're going to have to explain this one to me. Colorado loses to San Jose, Chicago, and Portland and drops 4 spots (makes sense). Beats Montreal in Montreal and doesn't move at all (huh?), then beats NY and draws DC and drops 3 spots (when compared to the 4 drop spot after 3 losses, WTF?)?
Hah. Yes. It seems a little not intuative. The "problem" is that it's a relative ranking system, so as other teams perform better over a five game stretch a team may not jump up or drop down as much after a win or less as they seemingly should. Right now the middle of the league is really tight. Last week Colorado, LA, Chicago, and Kansas City were all tied on identical ratings. Because LA and Kansas City won they were able to overtake Colorado on the draw. Chicago also stayed ahead despite the loss based on their five year form, but Chicago is now trending down. Right now LA, New England, Seattle, Kansas City, Dallas, Chicago, and Colorado are all within 2.5 ranks of each other, but since I'm not doing "ties", it seems like there's 6 positions between LA and Colorado, when performance wise their form is almost idential. The relative movement between those teams is going to be screwy for awhile methinks... Here's a ZOOM-ENHANCE! shot that might make things a little more confusing... I'd be happy to throw the whole thing up on Google docs for anyone who wants to look at it, though I can't promise it would make that much sense.
Maybe this helps. Here's the spreadsheets "season long" rank, last five game rank, and the average rank, which is what I've been reporting/graphing. Over the course of the season, Colorado has been the 9th best team. Over the last 5 weeks, they've been 11th best. Though that gives them an average rank of 10th, an average rank of 10th actually puts them as the 12th best team in the league when considering the season as a whole and the last 5 weeks.