MLS Playoff Predictions

In case you are unfamiliar, here is the format. The first round is the semifinals for each conference. Each pair of teams plays two games, one in each team's stadium. At the end of the two matches they add the scores up and the team with the most total goals wins and advances to the next round. If the scores remain tied (note that away goals don't matter as they do in some other competitions) then they play two 15-minute periods of extra time and penalties if necessary. That's the format for the conference semifinals. For the conference finals, it is only one match played at the higher seed's (smaller number) stadium. The MLS Cup final is a one-off that will take place November 22nd at Qwest Field in Seattle.

Personally, I like the one-game final at a site decided before the playoffs. I think the conference finals should follow the same format as the semifinals. Especially with the unbalanced schedules during the regular season, I think it gives too much of an advantage to the higher-seeded teams. If it's due to scheduling I think they should flip it so that the first round is just one match and the conference finals have the two-leg format.

Here are the matchups:

East
Real Salt Lake (4 seed) - Columbus Crew (1)
New England Revolution (3) - Chicago Fire (2)

West
Chivas USA (4) - Los Angeles Galaxy (1)
Seattle Sounders FC (3) - Houston Dynamo (2)

The first team listed plays at home the first leg. This gives an advantage to the higher seed because it's better to play at home the second leg since extra time could happen in the second leg.

Predictions

I'm actually unhappy with the version of the Poisson-Logit Model that predicts scorelines. The problem is that it doesn't seem to be sensitive enough to the attacking and defending qualities of the two teams. Therefore, I'm reverting back to the Poisson Model. Though it has problems as well, I think it will even out somewhat since each team plays at home making the predictions reasonably close. If it's tied after two legs, I'm giving the team that plays at home last a 55% chance of winning.

Here are the predictions:

Real Salt Lake - Columbus Crew

Columbus got a tough draw in my opinion. Despite finishing best in the league, their reward is the best scoring team in the playoffs. Real Salt Lake were unlucky to get just 40 points and sneak into the playoffs and I have them fourth overall. In addition, Utah is never an easy place to play due to the altitude. I think this tie will be closer than most people think. The first leg should be huge. If Columbus can hold Salt Lake scoreless, they will be in great shape heading home for the second leg. It will be tough sledding if they are down at that point.

Poisson predictions:
Columbus to advance: 52.4% (42.1% outright, 10.2% in extra time or shootout)
Salt Lake to advance: 47.6% (39.2%, 8.4%)

New England Revolution - Chicago Fire

As unlucky as Columbus were in the matchup, Chicago were just as lucky. New England are by far the worst team in the playoffs. They only got in due to above-expectation results in close matches which, again, is mostly, if not fully, due to luck. The Revs are the worst team in the playoffs at both scoring and conceding. Despite all that, there is a lot of variance in this format and since the MLS has so much parity, it won't necessarily be as lopsided as I'm implying. Chicago should win but New England certainly have a chance.

Poisson Prediction:
Chicago to advance: 57.3% (46.7% outright, 10.3% in extra time or penalties)
New England to advance: 42.7% (34.3%, 8.4%)

Chivas USA - L.A. Galaxy

Both these teams were near the bottom at scoring but pretty good at keeping opponents from scoring. I wouldn't expect a high-scoring affair. I'm sure others think the Galaxy have a big edge, but I have these two teams as very even. I think this will be very tight tie.

Poisson Prediction:
Galaxy to advance: 51.9% (40.2%, 11.7%)
Chivas to advance: 48.1% (38.5%, 9.6%)

Seattle Sounders FC - Houston Dynamo

This is a great matchup that should be better than the conference final will be no matter which of these teams advance. In my rankings these teams are right in the middle offensively and the two stingiest teams in the league. Houston edge out Seattle by one spot in each category. I expect a good, hard-fought tie that will still be up in the air in the second half of the second leg.

Poisson Prediction:
Dynamo to advance: 52.1% (40.5%, 11.6%)
Sounders to advance: 47.9% (38.3, 9.5%)

Championship Predictions

As I mention above, there is a lot of parity in the league. While I think New England is the worst team in the playoffs, the gap between the Revs and the top teams isn't nearly as large as you would see in cup competitions in Europe or South America even in later rounds. That combined with the format, which features a single match in the conference finals and the final, mean things are wide open. Here are the title percentages according to the Poisson model:

Columbus - 17.4%
Seattle - 15.8%
Houston - 15.1%
Galaxy - 14.5%
Chicago - 12.2%
Salt Lake - 9.9%
Chivas - 8.5%
New England - 6.6%

Needless to say, there is no clear favorite.

MLS End of Regular Season Rankings

The season has wrapped up and the playoffs begin on Thursday. I'll shortly post an article with playoff discussion. I wanted to separate the ranking discussion because it turned out to be a lot longer than I thought it would be.

End-of-Season Rankings


O Rank - goal-scoring rank
D Rank - goal-preventing rank
EGD - expected goal difference if each team played two matches against every other team at level shown by season results

A feature of my rankings is that at the end of the season for most leagues the rankings converge to goal differential. They don't for the MLS and Scottish leagues due to the unbalanced schedule. Each team plays every other team twice with two additional matches against a conference opponent. They stay pretty close, but teams with harder or easier opponents those extra two games will have their expected goal differential adjust based on how hard they were. For example, Houston and Columbus finished even on goal differential but Houston's extra games were against tougher opponents - Salt Lake and Dallas compared to Chicago and Toronto.

For a comparison to other rankings, here are Elo ratings. There is an explanation link near the top of that page. Elo ratings are commonly used for chess. There is a formula that gives or takes away points based on the result, including the score, of each individual match. The Elo formula emphasizes the match result which I ignore. I just take goals scored and allowed into account giving no extra bonus for winning or penalty for losing. The other big difference between that and what I have here is that there is some carryover from last year in those ELO ratings and I just started from scratch at the beginning of this season. To be honest, I think including previous results is probably better for stating which teams are the best. It's one of many things I plan to work on in the future as I'm not sure how to best use old results.

For a subjective rankings, check out Soccernet's Power Rankings. They "consist of a consolidation of votes from Jeff Carlisle, Jen Chang, Steve Davis, Frank Dell'Apa, Kristian Dyer, Ives Galarcep, Allen Hopkins and Andrew Hush." In my view they seriously overrate New England and Chivas while underrating Salt Lake and Colorado. This doesn't surprise me as these were teams that over or underperformed in close matches, which I've shown is mostly (and probably all) luck.

Looking at the season as a whole, I think it's fair to say that Houston and Columbus are the top two teams in the league. I think you can argue either way which of the two is better. Columbus won the Supporters Shield given to the team with the most points in league play. Houston finished just one point back and played a slightly tougher schedule. Really there's not much separating them and a bounce here or there could have flipped things. The Galaxy finished the league even on points with Houston but again I think this was because they had good luck in close games.

MLS - Analysis and Playoff Predictions (Oct 1)

While the European leagues are just getting going, the MLS is heading into the tense weeks at the end of the season.

The Format

The MLS uses a kind of hybrid between typical league and cup formats to decide its champion. There is a regular season in which every team plays every other twice plus two additional matches, one at home and one away. That is where we are now, teams have played 25-27 matches so there are 3 to 5 left. When the regular season is done, the top 8 teams make the playoffs, with the caveat that at least two teams from each conference must be the in the playoffs. Like other sports in the US, the playoffs are split by conference as much as possible. I say as much as possible because a team, or even two though that's very unlikely, from, for example, the West could play in the Eastern Conference playoff.

The format in the playoffs is similar to that of cup competitions, like the knockout rounds of the UEFA Champions League. They play two matches, one in each team's stadium. After the two matches the scores are added. If teams are tied in aggregate goals, then they go to extra time and penalties if necessary. This is different than UEFA competitions where the away-goals rule is used.

For now, the thing you need to know is that the top 8 make the playoffs assuming two from each conference are in that top 8.

Ranking

Here are the rankings:



The right-most column, which is what they're ranked by, is what the average goal differential would be for each team if the teams played a million balanced seasons at the level indicated by the results thus far. I say balanced seasons because I'm assuming each team would only play each other team twice, not play the additional two matches. O rank is their scoring rank, D rank is their defensive ranking.

Compared to the rankings for other leagues, the most striking thing about this is how much more parity there is in the MLS compared to other leagues. The obvious reason for this is the salary cap, which keeps some teams from spending drastically more than others. Probably for this reason, the best team in the MLS is, at least according to my model, not the best offensively or defensively. For comparison, Manchester United rates best at both, and both Inter and Real Madrid rate as the best teams in their leagues at scoring and second best defensively.

Let's look at the second column. That's the difference between my rankings and the team's current spot in the standings. For example, I rate Colorado 3rd but the Rapids are 5th in the table right now. Some of these difference are simply because teams have played a different number of games so far. Another difference is the schedule - teams don't play a balanced schedule. I think that part is minimal. The biggest difference in my view is that some teams have, if you'll excuse the gambling term, "run good" in terms of how many points they've picked up given their goal differential. My rankings are based on goal differential, obviously the table uses points. As I argued in a previous article, how many points a team gets for a given goal differential is apparently all luck. In my view, New England and Los Angeles are not as good as their records suggest while Real Salt Lake and Dallas are significantly better than their records.

The Playoff Race

To come up with playoff predictions, I simulated the league playing out 10,000 times using the Poisson-Logit Model, which I believe to be very accurate at predicting match outcomes this far into the season.

For a more in-depth description of the playoff race, see Ives's soccernet article which goes over the schedules and recent results for each team still in the playoff race. I don't like his odds, they underrate every contender, probably because when you convert his odds to percentages they add up to just over 5 spots instead of 8. Otherwise the article is solid.

Here is the breakdown:

Already in:
Columbus Crew

Almost certain:
Houston Dynamo - 99.99% (they missed the playoffs in exactly 1 out of the 10,000 simulations)
Chicago Fire - 99.88%
Los Angeles Galaxy - 98.95%
Chivas USA - 94.5%

Probably in, but with some work to do:
Colorado Rapids - 80.9%
Seattle Sounders - 75.6%
New England Revolution - 63.6%

Uphill climb:
Real Salt Lake - 39.0%
Toronto FC - 25.4%

Not looking good:
DC United - 13.9%
FC Dallas - 7.3%
Kansas City Wizards - 0.9%
San Jose Earthquakes - 0.02% (2 out of 10,000 simulations)

Out:
New York Red Bulls