SEBA Projections

MLS, USL, and NWSL projections through July 29

Below are the 2018 MLS season forecasts using data from games through July 29.

Some key points up front:

  • Union playoff chances: The Union now have a 56.2% chance of making the MLS playoffs, up from 20.5%.
  • Bethlehem playoff chances: The Steel have a 26.2% chance to make the USL playoffs, down from 32.0%.
  • Largest movers: Philadelphia gained the biggest boost in performance expectations for MLS in the past week, and Houston had the largest drop.
  • Plus/minus: The two top players in the league, in terms of their teams’ goal differentials when they’re on the field, are Atlanta United’s Michael Parkhurst and the New York Red Bulls’ Bradley Wright-Phillips. The top Philadelphia players for this stat are Cory Burke and Jay Simpson.
Interactive Charts

MLS forecasts

MLS forecasts over time

USL forecasts

USL forecasts over time

NWSL forecasts

NWSL forecasts over time

Philadelphia Updated Stats

This following chart shows the Union’s 2018 stats, as compared with typical winning and losing teams.

This following chart shows the Union’s stats in recent MLS matches, as compared with typical winning and losing teams.

Power Rankings

The scores used for these rankings are the expected points-per-game for all league opponents both home and away.

SEBA has Philadelphia increasing from 18th to 12th.

Currently, Philadelphia’s performance expectation is set by the following distribution of minutes:

Philadelphia +/- for 2018

[table id=62 /]

 

MLS +/- for 2018

[table id=65 /]

 

Playoffs probability and more

Philadelphia has increased from a 20.5% to a 56.2% chance of making the playoffs.

Next is the distribution of Philadelphia’s points against the MLS East’s sixth seed in all the simulations.

The Union increased from a 0-ish% to a 0.2% chance of winning the MLS Cup.

 

Philadelphia’s chances of having a U.S. Open Cup win increased from 10.6% to 16.2%.

These odds include the semi-final and final draw scenarios.

Philadelphia increased from a 10.6% chance to a 16.4% chance of qualifying for the 2019 edition of the CONCACAF Champions League.

The following are Philadelphia’s odds at different prizes over time.

The following are the probabilities of each final playoffs placement for Philadelphia.

The following shows the summary of the simulations in an easy table format.

[table id=59 /]

Next is a table showing the difference between this forecast and the last forecast.

[table id=61 /]

The following shows the difficulty for each team’s remaining schedule. Higher values means an easier remaining schedule.

The following shows the expectations for upcoming Philadelphia matches:

Model Validation

The following shows SEBA’s model error compared with random guessing. To do this, we save forecasts for the entire season’s games for every forecast’s data, and update the error charts all the way through the season.

Lower error is better.

USL

Power Rankings

SEBA has Bethlehem declining from 17th to 19th.

Playoffs probability and more

Bethlehem decreased from a 32.0% to a 26.2% chance of making the playoffs.

Bethlehem’s odds of winning the USL championship declined from 0.4% to 0.3%.

The following is Bethlehem’s odds at different prizes over time.

The following shows the probability of each post-playoff ranking finish:

The following are the simulation averages in an easy, combined format.

The following is the strength of remaining schedules for each team. Higher values mean an easier remaining schedule.

The following are the upcoming expectations for Bethlehem matches.

 

NWSL

The following are the projections for the 2018 NWSL season.

The following shows the remaining strength of schedule, as we project. Higher values represent easier remaining schedules.

 

 

11 Comments

  1. Chris, any idea why the playoff % increased so much?
    .
    Looking at the playoff chances over the last few weeks it’s pretty noisy. Same with the Union’s SEBA power rank. The only other team that seems to have similar noise in the power ranking is Chicago. All the other teams seem to have somewhat small week to week variations. Any idea what’s causing this?

    • Andy Muenz says:

      Union got a win on the road while Montreal lost, Chicago lost, and Orlando lost twice.

      • Right, but Montreal and Chicago stayed in the same place in the power rankings. The Union did basically swap places with Montreal this week.
        .
        But looking at the power rankings over time, there’s no other team with as much up and down and by as much at the Union. It’s like the model doesn’t know what to do with them and I was wondering why.
        .
        If it was due only to the games Harris and Dockal missed becoming more and less relevant in the model, I wouldn’t expect that amount of volatility.

      • Chris Sherman says:

        Yeah, primary culprit for Philadelphia’s odds and power-rankings variation are that they’ve won their last 2 road games and lost their last 2 (MLS) home games. All 4 of those outcomes were varying degrees of unexpected due to the power of home field advantage in MLS. For example, I think the model gave the Union only about a 15% chance of winning in Houston, so that gave them a huge boost over expected points in addition to better expectations moving forward.

        Also, teams on the playoff bubble, like Philadelphia, are more likely to have wild swings in playoff odds this late in the season, given the less uncertainty with fewer matches remaining.

        If you look at the power scores over time in the interactive chart, instead of the rankings of those, the scores change a decent amount but not as wildly as the rankings appear. The model somewhat consistently puts Philadelphia as a “Tier 3” club, but has been confused as to where within that Tier they rank when the match results keep being inconsistent with expectations.

      • Chris Sherman says:

        Actually JB, Montreal dropped 5 spots in the Power Ranking. I’m guessing you’ve encountered a bug I’ve recently discovered for the table-plugin we use.

        When I re-upload data to the table, it shows as updated right away when I’m logged into the site, but when I’m logged out it can take up to 2 days to update. I have a work-around for next week where I’ll simply delete the whole table and re-create it with fresh data.

      • I was actually looking at the power rank over time chart, Union from 18 to 12 and Montreal from 12 to 17.
        .
        That’s also where I was comparing the up and down of the Union compared to everyone else. Since mid-June SEBA has had them up and down several spots a few different times. Almost every other team only made one large move over the same time peroid.
        .
        The Union losing games they’re supposed to win and winning games they’re supposed to lose does explain this, but it’s weird that each game has had such a large effect dating back that far.

      • It’s almost like the model is validating the kneejerk fan reactions after every game:
        .
        Union win = we’re great!
        .
        Union lose = WSSM
        .
        And opinions on every other team remain somewhat constant.

  2. Andy Muenz says:

    Given that the Union’s expected points from Saturday is 0.44, I’m wondering how much losing would negatively affect their playoff chances. Theoretically it probably shouldn’t be much, but it seems like individual games can produce big swings. Obviously, outcomes in other games play a major factor as well.

    • Chris Sherman says:

      Yeah, it shouldn’t be much on its own, but as you point out, the results of other clubs could impact the Union’s odds more.

  3. Andy Muenz says:

    Chris, I would be interested in seeing what the expected point values of the other teams are for their upcoming games each week. For example, did Montreal have a high expectation at home against Atlanta which cost them likely points and helped the Union’s playoff expectation?

    • Chris Sherman says:

      This isn’t exactly what you were looking for, but you can get some of that from the table showing the difference in season simulations between this forecast and the last. If you hit ctrl+F and use “Next is a table showing the difference between this forecast and the last forecast.”, you’ll see this (although sometimes there are caching issues with our website where the data takes a day or two to update from last week).

      Montreal’s expectation of average season points went down by 2.6 points, while Atlanta’s went up by 2.2 points, so I’d say Montreal was expected to take the win more than Atlanta was.

Leave a Reply

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

*

%d bloggers like this: