SEBA Projections

Updated SEBA Projections through September 4

Welcome to the Statistical Extrapolation Bayesian Analyzer Projection System, which is, of course, a tortured gathering of words used to form the SEBA Projection System.

The system consists of a model projecting the probabilities of game results and then a barrage of simulations which account for the remaining luck. So the probabilities you see below show, essentially, what the odds are of an outcome if the team continues to play like it has been.

Below are the current season forecasts using data from games through September 4, 2016.

Power Rankings

The “Power Rankings” we concoct are the actual “strength” of the team according to competitive expectations. They are computed showing average expected points result if every team in MLS played every team both home and away.

SEBA has the Union remaining at No. 5 despite the 3-0 loss to Chicago. At Soccer America, the Union remain at No. 10. At SI, they drop one spot to No. 8.


Playoffs probability and more

The following charts show the probabilities for making the playoffs, winning the Supporters’ Shield, winning the MLS Cup, winning the USOC Cup, and qualification for CONCACAF Champions League.

The Union went down sightly to 99.4% from 99.6% last week in their probability of making the playoffs.


The odds of Philadelphia claiming the Supporters’ Shield improved from 0.7% to 0.1% with FC Dallas more than seven times as likely to win as any other team.


The Union decreased from 7.1% last week to 5.3% in their odds of capturing MLS Cup. Philly really missed an opportunity to position themselves for a first-round-bye, but that scenario is becoming less likely.


Philadelphia is eliminated from the US Open Cup. In addition to being the best team in MLS, FC Dallas also has home field advantage which gives them a tremendous edge on the USOC (and guaranteed access to the CCL Champions’ League). Then again, New England has less to play for in MLS which the model does not currently account for.


Philadelphia’s probability of qualifying for the CONCACAF Champions League has declined to 12.1% from 23.6%. Philadelphia’s probability of qualification is still largely dependent on topping Toronto and both NY teams for the conference leader.


Philadelphia finishing

The following shows the probability of various playoff categories and Philadelphia’s probability of each of them.


The following is a “density” chart showing the relative probability of Philadelphia finishing at the different after-playoffs rankings from 1 to 20.


The following is the list of Philadelphia’s remaining scheduled matches accompanied by result probabilities as predicted by SEBA.


The charts below show the average results of all the simulations and the probabilities associated with all of the above. The chart is broken up into two parts in order to better fit on your screen.

(Note: CCLAlternate refers to qualifying for CONCACAF Champions League non-traditionally. U.S. teams qualify by winning the MLS Cup, winning the Supporters’ Shield, winning the Conference in the regular season, and by winning the U.S. Open Cup. If a Canadian team wins any of these or a U.S. team wins more than one, the extra slots are filled by top regular season performers.) (Also, “regSeasFinish” refers to finishing the conference in first place AND not winning other CCL slots like MLSCup, Supporters’ Shield, and USOC)



Changes Over Time

This chart shows how Philadelphia’s chances at each of the outcomes has evolved over the course of the season.


This shows the evolution of Philadelphia’s movement within SEBA and other media outlets’ power rankings.


This shows the MLS Eastern Conference’s power rankings’ evolution in SEBA throughout the season.



This shows the odds of breaking MLS single-season records, not of holding them. If more than one team could break the same record this season, they would both be represented below.

(Philadelphia did not break any single-season records in any of the 15,000 simulations.)

The following table shows all single-season records that had a record broken in at least one simulation. However, it limits the records to show only the most probable team for breaking each record.


USL Projections

Bethlehem remained at 25th and the Harrisburg City Islanders declined one spot to 23rd from 22nd.‘s top 10 rankings do not include either team.


Bethlehem’s probability of making the playoffs continues to decline with their winless streak from 7.8% to 4.9%. Harrisburg City’s increased to 4.7% from 3.1%.


Both Bethlehem and Harrisburg City have no chance of winning the USL regular season title. NYRB 2 has almost clinched.


Bethlehem’s probability of winning the USL Championship probably decreased, but in the rounded terms shown here, they remained at 0.1%. Harrisburg City’s chances also remained at 0.1%.




As with MLS, the below density charts show the relative probabilities of Bethlehem and Harrisburg finishing for their final playoff-rank.



The following are the upcoming matches and associated probabilities for Bethlehem and Harrisburg City.



The following tables show the average simulation results. They are divided into two tables to present better to your screen.



Changes over time

Below shows the probabilities of various outcomes for Bethlehem over the course of the season.


Below shows the probabilities of various outcomes for Harrisburg over the course of the season.


Below shows the evolution of SEBA’s power rankings for Bethlehem Steel FC and the Harrisburg City Islanders.


Below shows SEBA’s power rankings evolution for the USL East over the course of the season


NWSL Projections




The following tables are the average simulation results produced by SEBA.







  1. Dumb question. When calculating the likely PPG on Saturday, does the model account for the fact that Montreal has a midweek game and the Union don’t?

    • Chris Sherman says:

      No it doesn’t. In early stages of development, I used rest days as a predictive variable, but it actually slightly favored teams with shorter rest (a result that was actually represented similarly in an NFL model and NCAA-B model).

      I believe this good indicate that teams rest starters and/or players conserve energy when they are expecting shorter rest in future games, but this remains to be confirmed.

Leave a Reply

Your email address will not be published. Required fields are marked *