clock menu more-arrow no yes

Filed under:

Season-End Table Projection Update: Not Looking Good

New, 17 comments

A few weeks ago Ryan and I shared our competing mathematical models projecting the table as we believe it will look like by the end of the season. Since then a lot of things have happened that have thrown those earlier projections right out the window.

For example, the LA Galaxy have gone on a bit of a run, winning four of their last six and climbing up the table in the process. The Seattle Sounders, meanwhile, have gone the opposite direction, losing ground to the top two Western Conference teams.

And then there's the Portland Timbers. Neither of our projections back in early June expected Portland to finish higher than seventh, and that persists in today's edition. Both of our models predict the Timbers to stay mired in 8th place, where they currently are.

After the jump, we'll recap our methodologies and share our most recent projections.

Ryan's model

For my projections I use a simple calculation. First I separate home and away points and calculate the points per match for home and away games. Once I do that I then take how many home and away games they have left and times that by their points per match for home and away games. I then take those numbers and add them to the points a team currently has.

Andy's model

I use a combination of (a) each team's goals scored and allowed and (b) home and away winning percentages to predict the winner of every game for the remainder of the season. In order to account for changes in team performances, I give a higher weight to recent matches than those that took place earlier the season. In order to make sure the predictions don't get too far out of whack, I make sure that roughly the same percentage of overall draws are predicted going forward as have already happened (about 22% as of today).

Finally, at the suggestion of Stumptown Footy commenter pdb, I've added a feature that prevents a team from winning or losing too many games in a row. Basically, if a team has a recent points-per-match number outside a specified range, a modifier is factored in to decrease the likelihood that their streak, good or bad, continues. That keeps teams closer to the mean.

So, here are the updated projections: