Assessing the Projections: 2014 Major League Soccer regular season

The 2014 Major League Soccer regular season concluded a week and a half ago.  It’s time to look back at the projections made at the start of the season and assess them.

Here is what we projected at the beginning of the season – first by conference:

MLS_EC2014_Projections

MLS_WC2014_Projections

And then the single table:

MLS_Full2014_Projections

The final league table finished like this at conference level, and I’m including the Pythagorean expectation of the team records:

Eastern Conference

Standard Table Pythagorean
GP W D L GF GA GD Pts W D L Pts Δ
DC United 34 17 8 9 52 37 15 59 16 9 9 57 +2
New England Revolution 34 17 4 13 51 46 5 55 14 9 11 51 +4
Columbus Crew 34 14 10 10 52 42 10 52 15 9 10 54 -2
New York Red Bulls 34 13 11 10 55 50 5 50 14 8 12 50 0
Sporting Kansas City 34 14 7 13 48 41 7 49 14 9 11 51 -2
Philadelphia Union 34 10 12 12 51 51 0 42 13 8 13 47 -5
Toronto FC 34 11 8 15 44 54 -10 41 11 8 15 41 0
Houston Dynamo 34 11 6 17 39 58 -19 39 9 8 17 35 +4
Chicago Fire 34 6 18 10 41 51 -10 36 10 9 15 39 -3
Montreal Impact 34 6 10 18 38 58 -20 28 9 8 17 35 -7

Western Conference

Standard Table Pythagorean
GP W D L GF GA GD Pts W D L Pts Δ
Seattle Sounders 34 20 4 10 65 50 15 64 16 7 11 55 +9
Los Angeles Galaxy 34 17 10 7 69 37 32 61 19 7 8 64 -3
Real Salt Lake 34 15 11 8 54 39 15 56 16 9 9 57 -1
FC Dallas 34 16 6 12 55 45 10 54 15 8 11 53 +1
Vancouver Whitecaps 34 12 14 8 42 40 2 50 13 10 11 49 +1
Portland Timbers 34 12 13 9 61 52 9 49 15 8 11 53 -4
Chivas USA 34 9 6 19 29 61 -32 33 6 8 20 26 +7
Colorado Rapids 34 8 8 18 43 62 -19 32 9 8 17 35 -3
San Jose Earthquakes 34 6 12 16 35 50 -15 30 9 9 16 36 -6

The final single table looks like this:

Standard Table Pythagorean
GP W D L GF GA GD Pts W D L Pts D
Seattle Sounders 34 20 4 10 65 50 15 64 16 7 11 55 +9
Los Angeles Galaxy 34 17 10 7 69 37 32 61 19 7 8 64 -3
DC United 34 17 8 9 52 37 15 59 16 9 9 57 +2
Real Salt Lake 34 15 11 8 54 39 15 56 16 9 9 57 -1
New England Revolution 34 17 4 13 51 46 5 55 14 9 11 51 +4
FC Dallas 34 16 6 12 55 45 10 54 15 8 11 53 +1
Columbus Crew 34 14 10 10 52 42 10 52 15 9 10 54 -2
New York Red Bulls 34 13 11 10 55 50 5 50 14 8 12 50 0
Vancouver Whitecaps 34 12 14 8 42 40 2 50 13 10 11 49 +1
Sporting Kansas City 34 14 7 13 48 41 7 49 14 9 11 51 -2
Portland Timbers 34 12 13 9 61 52 9 49 15 8 11 53 -4
Philadelphia Union 34 10 12 12 51 51 0 42 13 8 13 47 -5
Toronto FC 34 11 8 15 44 54 -10 41 11 8 15 41 0
Houston Dynamo 34 11 6 17 39 58 -19 39 9 8 17 35 +4
Chicago Fire 34 6 18 10 41 51 -10 36 10 9 15 39 -3
Chivas USA 34 9 6 19 29 61 -32 33 6 8 20 26 +7
Colorado Rapids 34 8 8 18 43 62 -19 32 9 8 17 35 -3
San Jose Earthquakes 34 6 12 16 35 50 -15 30 9 9 16 36 -6
Montreal Impact 34 6 10 18 38 58 -20 28 9 8 17 35 -7

I’m writing this post on the night of midterm federal elections in the USA, which serves as a reminder that modeling and forecasting human behavior is a difficult job that can and should humble any analyst.  Overall, I think this model benefits from familiarity with the players and the local characteristics of MLS, but there will always be missed results from a preseason model.   Nevertheless, we got three of the five playoff teams in the East and four out of five playoff sides in the West.  We projected the top two teams in the league and three of the top four, and the difference between Seattle and LA fit within the root-mean-square error of the expected points model.  But when we look at the Pythagorean expectations at end of season — which capture the expected point total for an average team with identical goal statistics — we see a huge difference between the Sounders and the Galaxy.  Seattle’s Pythagorean residual was +9, which meant that they performed nine points better than an average team.  They also performed ten points better than expected for a team with their projected goal offensive and defensive totals.  Both the Sounders and the Galaxy scored 16-18 more goals than expected, but the Sounders were 11 goals worse in their goals allowed expectation.  And yet, the Sounders finished with 64 points and 20 wins — the first team to finish with at least 20 regular-season wins since 1998.  Seattle played much stronger than expected, especially in offense, but in my opinion they were on the receiving end of more luck than previous seasons, which might be what they need to raise the Anschutz Trophy later this month.

The biggest miss, and perhaps everyone else’s big miss, was Toronto FC.  We projected them to finish second and they missed the playoffs altogether.  They can’t blame lack of luck as their Pythagorean residual was 0; this was simply a poor side on both sides of the ball (scored 8 fewer goals than expected and allowed 15 more goals than expected) that ended up running 13 points below expectations.  In a league where the difference between top of the conference and out of the playoffs is 10-15 points, such a drop in performance is fatal.  Of course, this is a top-level observation of TFC’s underperformance; a more thorough examination of the match data is necessary to identify what went wrong and contributed to this debacle.

As one might expect, the playoff teams performed better in terms of goal production than expected — DC United, RSL, FC Dallas, Columbus, New York.  Very few of the teams performed better than expected in terms of goal defense; only three teams allowed fewer goals than expected: DC United, Columbus Crew, and Vancouver Whitecaps.  All three of those teams scored 10-11 points more than their preseason projections and had Pythagorean residuals within ±2 points.  As far as the non-playoff teams are concerned, we were able to capture the cluster of six teams near the bottom quite well.  All of them, except Chivas USA, performed worse than expectations, couldn’t score, and couldn’t defend.  Chivas USA actually had better goal defense than expected and run four points ahead of expectations with some luck thrown in, which kept them off the bottom of the table and instead above Colorado, San Jose, and Montreal.  So some pride was salvaged in their final MLS season.

Statistical Summary:

Pred Goals Actual Goals % Diff Goals RMSE Pts RMSE GF RMSE GA RMSE Pos
813 924 13.65% 8.32 9.73 9.56 4.87

CORRECTION (8 Nov, 20:25): I originally miswrote that I had predicted “two of the five playoff teams in the East”, instead of three.  I have corrected the text above.

Share

Tags: