Tuesday, September 28, 2010

On the Playoff Race

One of the problems I've run into with following MLS is figuring out where my team stands in the playoff race. In essence, the top 8 teams make it into the playoffs. In reality, it's possible for a team not in the top 8 to make it into the playoffs, but the likelihood of that happening is very slim. The full rules for qualifying are availabe here. For most soccer leagues, it's easy to figure out who the top 8 teams are. They have a single table format with a balanced schedule. Very few teams have played a different number of matches than the other teams in the league. It's easy to compare apples to apples in this scenario. MLS, on the other hand, has a very bizarre way of scheduling matchs and it's uncommon for teams to have played the same number of matches at any point in the season. How, then, do you figure out where you stand in the table without running through all the scenarios of "Well, if team A beats team B, and then draws against team C, and we beat team D, then...".

I've several approaches to this problem. The first is to use the average points earned per game as an estimate of what a team will do in their games-in-hand. This is a decent approach for the middle of the season when a team's performance will be averaged out over a large number of games. Towards the end of the season, this approach doesn't work as well. For example, if a team is averaged 1.2 PPM, then assuming they will earn 2.4 PPM in their two games-in-hand can be a poor assumption. It's quite reasonable for them to pick up the full 6.

Another approach is to use Monte Carlo simulations to calculate the possible outcomes of future matches. Sports Club Stats is a good example of this approach. The problem is that their estimates are limited to the quality of the model they use for estimating the outcomes. Your outputs are only as good as your inputs and quite frankly, the discipline isn't mature enough to have quality inputs just yet.

Baseball has the concept of games back. It's a beautiful metric that tells you exactly what your team needs to do in order to move into first. Any casual baseball fan can understand it. It's a simple formula that encapsulates where a team stands no matter how many games-in-hand are in play. It got me thinking about whether or not something similar could be applied to a league with trinary outcomes (win, loss, tie). I tweaked the formula a bit to work with ties and the results are decent but take some getting used to. Instead of the simplicity of a half or full game back, there are now weird situations like being 1/6 of a game back. I'm going to update the results each week for the rest of the season to get a feel for how it works. Below are the latest standings as of 9/27.

Team NamePointsGPGB From Playoffs
Los Angeles Galaxy5026-3 2/3
Real Salt Lake4826-3
FC Dallas4626-2 1/3
Columbus Crew4526-2
New York Red Bulls4426-1 2/3
San Jose Earthquakes3924-1
Colorado Rapids3825- 1/6
Seattle Sounders FC39260
Kansas City Wizards33251 1/2
Toronto FC31262 2/3
Chicago Fire26243 1/3
Philadelphia Union27253 1/2
Chivas USA25254 1/6
New England Revolution26264 1/3
Houston Dynamo26264 1/3
D.C. United18267

Both the Sounders and KC are deep in the playoff race after going through some dark times this season. At their worst points during the season the teams were only 2 and 2.5 GB respectively. I think that's a testament to parity in MLS and the fact that half the teams qualify for the playoffs.

No comments:

Post a Comment