Gluckonomy: Is it really all about scoring goals?

Chris Gluck writes a regular feature for the Columbian called "Possession with Purpose". After breaking down each of the 600 or so MLS matches played this year, Chris is stopping by to share some of his findings with us.

Here are the top ten goal scoring teams this year in MLS:

1. Real Salt Lake (1.72 goals per game);
2. LA Galaxy (1.65);
3. New York Red Bulls (1.56);
4. Montreal (1.55);
5. Portland (1.53);
6. Vancouver (1.50);
7. New England (1.41);
8. FC Dallas (1.41);
9. Sporting KC (1.38);
10. Chicago Fire (1.31);

Teams hot in the chase for the playoffs but not on this list include Colorado (11th), Philadelphia (14th), Houston (15th), and San Jose (16th).

How else can we evaluate how well a team is performing that has relevance to points on the league table?

Beginning late last year and starting anew this year, I have been collecting and analyzing statistical data points for MLS games to try and find a better way to evaluate team performance. What follows, for your consideration, is my attempt to do this.

All individual data points provided below were collected by reviewing the OPTA Chalkboard and Stats pages for every MLS game to date, excluding the LA Galaxy-Montreal Impact game on Wednesday night and the Sporting Kansas City-DC United game last night.

The Indices offered below are my own, developed based upon my analysis of the data points and how they might have relevance, collectively, to team attacking and defending performance.

Possession with Purpose (Entire Pitch) – Passing accuracy across the entire pitch, passing accuracy within the final third, shots taken, shots on goal and goals scored.

Possession with Purpose (Final Third) – Same as above excluding Passing accuracy across the entire pitch.

Dispossession with Purpose (Entire Pitch) – Defensive activities (tackles won, interceptions, clearances, and block crosses) across the entire pitch, defensive activities within the final third, defensive shots blocked and goal keepers saves.

Dispossession with Purpose (Final Third) – Same as above excluding Defensive activities across the entire pitch.

I also made use of two other common indicators that most statisticians in soccer review: Goals Scored and Goals Against; as well as five other team statistics separate from the indices above, some of which are used by other statisticians: Passing Accuracy in the Final Third, Shots Taken, Shots on Goal, Defensive Activities in the Final Third, and Possession Percentage.

Setting the stage

Correlation results reflected in this evaluation stem from comparing data points (individually) for all 19 teams (separately); for example, San Jose has played 32 games this year; the possession percentage for each game was correlated to the points resulting from each of those games.

17 teams have played 32 games to date, while Montreal and LA have only played 31 (as of the writing of this article). The most recent game between LA and Montreal is not included. I do not look to measure individual statistics relative to this game; it’s not an individual sport, it’s a team sport that requires team play. Teams win trophies; not individual players.

I bought into this philosophy almost 20 years ago as a youth coach in England and I remain stubbornly steadfast and steadfastly stubborn it’s true. The beauty of this year in tracking and trending the Portland Timbers, for me, is that I have witnessed Caleb Porter successfully put this philosophy into practice.

Looking to validate an apples to apples ranking system, I had no preconceived ideas on how any of these indices or statistics would play out, other than if there was little or no correlation/relationship between any of those data points to garnering points in the league table I would throw them out.

This is not to say any of those individual data points don't have value, they do, but the intent here is to find a ranking system that can be used to equally evaluate all teams in MLS without any undue influence by preconceived notions of the game.

There are other ranking systems, predictive models, and in-depth individual statistical reviews out there; the most comprehensive ones I’ve seen have been offered up by some writers for Statsbomb.com; created by Ted Knutson and Ben Pugsley.

Evaluating the relevance of the indices and data points to results...

NOTE: Any data point with less than +/- .4000 is considered to have 'little or no correlation'.

Possession with Purpose (Entire Pitch) = .4856 (average correlation to points in league table)

Possession with Purpose (Final Third) = .5785

Dispossession with Purpose (Entire Pitch) = .2991

Dispossession with Purpose (Final Third) = .4238

Goals Scored = .6206

Goals Against = -.5976

Passing Accuracy (Final Third) = .1789

Shots Taken = .0697

Shots on Goal = .2972

Defensive Activities (Final Third) = -.1949

Possession Percentage per game = -.1319

What categories make the first cut for the apples to apples ranking system comparison?

Possession with Purpose (Entire Pitch), Possession with Purpose (Final Third), Dispossession with Purpose (Final Third), Goals Scored, and Goals Against.

What categories didn’t make the cut?

Dispossession with Purpose (Entire Pitch): There are six teams that have a correlation greater than .4 to points in the league table based upon overall Dispossession with Purpose (Entire Pitch); the team with the highest correlation (.5349) is the Columbus Crew. The other ones are Portland (.4356), Sporting (.4865), Seattle (.4934), LA (.4513), and Dallas (.4560). Oddly enough, 4 of those 6 teams have some of the best point totals this year in MLS; Columbus is probably high given their luck with opponents getting red cards when they visit and Dallas... well... they started out great but died at the end.

Passing Accuracy Final Third: There are two teams that have a correlation greater than .4 to Passing Accuracy in the Final Third; the team with the highest correlation (.6918) is the LA Galaxy with New England (.4344) the other one.

Shots Taken: There are no teams in all of MLS where the correlation to shots taken, as a whole, compared to points in the league table is greater than .4; therefore using shots taken as a comparison between one team (performance wise - versus another team) - really doesn't have much statistical value. This is not to say Shots Taken doesn't have value when evaluating a team performance, but the low correlation indicates to me that it is pretty much a wash when using it as a comparison against other teams and their positions in the league table.

Shots on Goal: There are four teams where the correlation is greater than .4 to Shots on Goal; the team with the highest correlation to points in the league table is Chivas USA (.6229); the other three teams are Chicago, Columbus and New England. I wonder what it says about these teams when comparing them to those who have more than 46 points and the fact that Chivas (with 26) is so far down the table? Again, this doesn't mean Shots on Goal does not have value relative to scoring goals; it does but it's not a balanced statistic that should have value when assessing team to team comparisons like that of Shots Taken above.

Defensive Activities in the Final Third: There are three teams where the correlation is greater than .4 to Defensive Activities in the Final Third; the team with the highest correlation is Colorado (-.5568); the other two teams are LA (-.4318) and Montreal (-.4028).

Possession Percentage: There is one team where the correlation is greater than (-.4) to Possession percentage; that team is Montreal (-.4435). For most that seems reasonable since many feel or think that Montreal is a counterattacking team who usually "turtles-up" in defense.

Even though these statistical data points didn’t make the cut to be part of a ranking system they are still valuable on their own.

The League leaders and the indices and data points for those teams that have the highest correlation to their points in the league table...

By wins New York is the top team; and the single best category that has the highest correlation to their points in the league table is Possession with Purpose (Final Third) (.6405); the next highest category that relates the best to them garnering points in the league table is Goals Scored (.6388).

The current team in second is Portland; the single best category that has the highest correlation to their points in the league table is... Goals Against (-.5918); in other words the fewer goals the opponent scores the more likely Portland takes points. The next highest category that relates the best to them garnering points in the league table is Possession with Purpose (Final Third) (.5322).

The third place is Real Salt Lake; the single best category that has the highest correlation to their points in the league table is Goals Against (-.7326); the next highest category that relates the best to them garnering points in the league table is Goals Scored (.5265).

The fourth place is Sporting KC; the single best category that has the highest correlation to their points in the league table is Goals Scored (.6653); the next highest category that relates the best to them garnering points in the league table is Possession with Purpose (Final Third) (.6262).

The fifth place team is Seattle; the single best category that has the highest correlation to their points in the league table is Possession with Purpose (Final Third) (.7084); the next highest category that relates the best to them garnering points in the league table is Goals Scored (.7022).

The sixth place team is Colorado; the single best category that has the highest correlation to their points in the league table is Possession with Purpose (Final Third) (.6346); the next highest category that relates the best to them garnering points in the league table is Goals Scored (.6329).

The seventh place team is Houston; the single best category that has the highest correlation to their points in the league table is Goals Scored (.7012); the next highest category that relates the best to them garnering points in the league table is Possession with Purpose (Final Third) (.6791).

The eighth place team is LA Galaxy; the single best category that has the highest correlation to their points in the league table is Dispossession with Purpose (Final Third); the next highest category that relates the best to them garnering points in the league table is Possession with Purpose (Final Third). Bottom line(s)...

The index (Possession with Purpose - Final Third) correlates to points in the league table, either first or second for 7 of the 9 top teams in MLS. The Possession with Purpose Index is comprised of passing accuracy within the final third, shots taken, shots on goal, and goals scored. So, while the individual data points don't correlate to points in the league table, the combined consolidation of those points does relative to each other. Overall, 13 out of 19 teams have this Index as being third highest or better.

What is more interesting is that only 3 of the bottom 7 teams have Possession with Purpose (Final Third) as their 1st or 2nd best correlation to points in the league table. For me, that is a great indicator to show that teams who garner more points will, on average, have a better Possession with Purpose (Final Third) indicator; being able to separate the good from the not so good is a worthy outcome for a team statistic.

Goals Scored, as a single data point correlates to Points in the league table, either first or second, for 6 of the 8 top teams in MLS and is third or higher for 17 of the 19 teams in MLS.

It’s not an indicator that shows separation like the Index above but it is a consistent data point that has relationship to the overall majority of teams in MLS.

Goals Against; while this data point is only 1st or 2nd highest in correlation to points in the league table for 8 of the 19 teams this data point had the second highest overall average correlation to points in the league table for all MLS teams across the league (-.5976).

Conclusion...

In considering all the information it seems reasonable to me that three categories, Possession with Purpose (Final Third), Goals Scored, & Goals Against will provide the most value in ranking teams' productivity in an apples to apples comparison.

The others, Possession with Purpose (Entire Pitch) and Dispossession with Purpose (Final Third), have good correlations relative to Points in the League Table, but their frequency as being in the top three, with a stronger correlation, is not enough for me to warrant them as having value in a Ranking System.

Here is how the teams rank as of 16 October, 2013...

Numbers provided in parenthesis equal the total points (low is better) when adding up the separate ranking of those three categories compared to the other teams in MLS. For example, Portland is 3rd best in PWP (Final Third), 2nd best in Goals Against, and 5th best in Goals Scored; therefore PTFC = 10.

1. Portland Timbers (10)
2. Real Salt Lake (12)
3. New York Red Bulls (15)
4. LA Galaxy (15)
5. Sporting KC (20)
6. Montreal Impact (21)
7. New England Revolution (21)
8. Colorado Rapids (22)
9. Vancouver Whitecaps (22)
10. FC Dallas (30)
11. Seattle Sounders (32)
12. Houston Dynamo (33)
13. Columbus Crew (37)
14. Chicago Fire (39)
15. Philadelphia Union (41)
16. San Jose (45)
17. Toronto (48)
18. Chivas USA (51)
19. DC United (56)

Closing thoughts for your consideration...

It's hard to equate something like luck in this game; one could argue that Columbus have been very lucky to have their opponents get red cards at their home games 8 times this year, yet they haven't turned all those red cards into 3 point victories - perhaps that is part of the reason why their head coach was fired.

Notice that Seattle are in 11th place; given their last three games and the score-line for two of them, it shouldn't be surprising that they are drifting in the shallow water as much as they are; four games earlier they were in the top 3.

A surprise for many is how poorly Dallas have closed out the year. Perhaps the higher ranking in this model is a carryover of earlier form, as opposed to how things have progresses lately. Many would probably agree that they are a better team than what their record offers. Does this position in the table lead to a change in managers this off-season? Hard to say, but a disturbing indicator is that Dallas have the second lowest correlation of points in the league table to goals scored (.4306). And, as noted earlier the best category that correlated to points in the league table was Dispossession with Purpose (Final Third); in other words their defense, when clicking, was the indicator that got them their points - and when it wasn't they paid for it.

(It has since been announced that Schellas Hyndman will step down as coach for FC Dallas after this season.)

Houston has been hard to figure this year, but it's clear that when they score goals and possess the ball across the entire pitch they win. This approach has been in place so long now that other teams are finding ways to crack their ability to possess and control. This could be why Moffat was traded - an internal decision to alter their team tactics into a slightly different attacking style. In the last few games Houston have pulled some points back and now sit well in the playoff chase. Having taken 11 out of the last 15 possible points, it seems reasonable that Houston have altered their attacking style.

Last year the bane of New England was scoring goals; this year they seem to have done better in that category. Given that they have back to back games with Columbus, the first in New England there is every chance they could walk away with 6 points just as much as walking away with none. Given the model here, it is my guess that New England wins out and extends their streak to 6 games unbeaten in a row; a nice wave to enter the playoffs.

Hard to figure out are Chicago, Philadelphia and San Jose. Most of this season these three teams have pretty much played under par against most other sides and I am not hopeful any of those three teams make the playoffs. All three are direct attacking teams, but the stronger teams this year (Real, Sporting, Portland, New York and LA) have all averaged over 50% possession the entire year. For me that speaks to quality and quantity as opposed to just quantity. Not much brain power in having the team hoof it forward for 90 minutes. I'm a purist when it comes to possession based soccer and the sooner the direct style leaves MLS, the sooner this league gains more credibility with leagues and players in Europe.

As for the bottom feeders (Chivas, Toronto and DC United); it seems pretty reasonable to me that my ranking system covers both the top and bottom end of the league; it's always a muddle in the middle.