ASA Podcast: Episode VI

Everyone, here we are with American Soccer Analysis Podcast Episode 6! We talk about Juan Agudelo, shots and finishing (skill vs. "luck"), grass pitches vs. artificial turf, Kei Kamara and his return to KC, and then some about bowel movements. LISTEN NOW!!! [audio http://americansocceranalysis.files.wordpress.com/2013/05/asa-episode-6-a-treatise-on-turf-kamara-and-bowel-movements.mp3]

Possession Confusion

Consider every conversation ever had about soccer tactics. I would bet 99.9% of them touched on one specific subject: possession. Whether it’s the men’s league team you play for, or the club team you cheer for, isn’t more possession always a good thing? I can’t answer that question confidently, but I will explore it. The first obstacle to analyzing and discussing possession in MLS is the data itself. We get our data from Opta, and this is what Opta defines as possession:

During the game, the passes for each team are totaled up, and then each team's total is divided by the game total to produce a percentage figure which shows the percentage of the game that each team has accrued in possession of the ball.

“Possession” in Opta’s data is thus a measure of the proportion of completed passes in a match for each team, not a proportion of time. A lot of short, quick passes will accrue possession for a team that may only have the ball for a matter of seconds. This isn’t necessarily bad or good. It is what it is, and we’ll work with it.

Not all passes are created equally---or better put, not all teams' passes average out to be equally effective---but for a moment let’s suppose that they are. It’s hard to gather data on the value of each pass, and hard to then weight teams’ passes accordingly. So let’s just stick with the assumption that all teams' passes are equally effective. Perhaps someday we can sit around drinking beer and punching holes in that assumption. Today is not that day.

Under that assumption of equal passes, a team that completes a higher proportion of passes than its opponent will likely have strung together effective buildup more often than its opponent. Having created more effective build up, that team will likely have earned more scoring opportunities than its opponent. Having earned more scoring opportunities than its opponent, that team will be more likely to score goals and nab points. So this sort of possession should really imply sunshine and rainbows for the participating team. Seems like fair logic to me, but of course, I’m the one writing.

Looking at the tables—tables that were created with Opta’s version of possession, remember—we don’t see a strong correlation between possession and results. Four of the top five teams (by points per match) have 50% possession or less, but overall there is still a weakly positive correlation. We start to get significant results when we assess the correlations between teams’ possession and Attempt Ratios (0.60*), and again with Shots on Goal Ratios (0.55*). Those positive correlations imply that more possession coincides with more scoring chances. Of course, there is not nececelery a causal link.

Let’s take a look at this from another perspective. If we look at the relationships game-by-game—rather than team-by-team—the correlation between possession and scoring chances is still positive. The team that possesses the ball for a majority of passes (Opta’s definition) during any given match also tends to earn more scoring attempts than its opponent.

So far I’ve bored you with support for conventional wisdom: possession coincides with more scoring opportunities, and thus probably with better results.

But then I control for a few variables and shit goes haywire.

When I control for each individual team and whether or not they were playing at home, the relationship between possession and results is decidedly negative. In fact, a team that possesses the ball an additional 10% in any given match is expected to lose half of a goal on average, equivalent to about half of a point. For example’s sake, consider the Seattle Flounders Sounders. Over Seattle’s top four matches in terms of possession, it has earned just one point. However, during Seattle’s bottom four matches in terms of possession, it has earned eight points. Seattle is an extreme case, but a good example of what my model is picking up. Most teams individually seem to do worse when their possession is higher.

So more possession seems to correlate with more shots, and more shots seems to correlate with more goals, but for some reason more possession does not share a significant relationship with more goals. There is some missing information screwing with me, and I don’t have a definitive explanation for this strange paradox, but I will share a theory.

Each team has a style. Whether or not that style works is probably mostly a product of how well the players fit in, and how good those players are in the first place. Perhaps, in general, a style that focuses more on stringing short passes together tends to produce more shots than a high-risk/high-reward style, but this type of possession is not a necessary condition for success. Once each team develops its style, a certain amount of possession is required to optimize that style. For Montreal, it may be 49% possession, and for Portland, it might be 57%. This would explain the mild positive correlations between possession and shots across teams.

But why is it that, across games, more possession seems to correspond to less goals and worse results?

In a given game, if a team generates more possession—more passing by Opta’s definition—then perhaps that is indicative more of the opponent’s defense than of the desire of the team in question to possess. In other words, an excellent defense may not necessarily kill possession, but rather, push possession to less dangerous parts of the pitch. In this way, more possession is simply indicative of a frustrated team, not a team in control doing what it wants to do.

Without being able to conclude this thought exercise satisfyingly, I will propose a few things. First, that by charting each shot’s point of origin, we can begin to assess the quality of a team’s shots. And second, that possession data should be gathered from the distinct areas on the pitch. Possession in the attacking third is likely more valuable than possession in the defensive third. Some combination of these two measurements could very well help to explain the paradox we’re seeing with passing possession and team success.

*A perfect positive correlation would be 1.0.

New England Revolution acquires Juan Agudelo: What does that mean?

First things first before I make fun of the Revolution (and I will).  Their defense has been---excluding the New York outlier---borderline elite this season. That's possibly one of the few reasons they're still afloat and maybe the only reason to watch them (sorry, Lee Nguyen).

Tempo-free soccer has the Revs ranked 6th in dAG (defensive attempts on goals), which is how many times an opposing team has made any attempt at their goal. Add to that that we have them ranked 2nd (6.2%) in Opposing Finishing%, which is how often a team's opponents successfully convert attempts into actual goals. They're better than every team outside of Montreal in that category.

This has all culminated in only 6 goals allowed in 9 games. Something that would be overlooked if it wasn't for their horrible attack and the need for at least some positive mention.

But now the Revs have added the young (former starlett?) Juan Agudelo, someone who saw time with the US National team only 6 months ago in Russia and didn't look awful by any stretch. To be fair, he's someone that has actually come out looking very strong for Chivas earlier this season, but he's been hampered the last few weeks with hamstring issues.

It was thought that he had mended a brewing off-season situation between himself and Chivas USA head coach, El Chelis. But of late, Chelis has given a lot of credit to his now former striker. He told MLSSoccer:

"I didn’t know what I had in Agudelo, but by having him, what I asked for doesn’t matter because Agudelo is a model. He is the natural on this team. He’s a player that has many technical qualities. He’s very involved in working to improve others."

And now he's shipped off to the greater Boston area and we are with out the full detail of the acquisition being yet to be vented  in exchange the Goats received allocation money. The spice of life and magic dust that no one talks about and everyone wants. Of course for us this isn't about the details at this point.

What Juan Agudelo will bring is spectacular things and then all together frustrating things. He averages about 16 shots on goal per 1500 minutes, a number he has yet to reach in either of his stops in Chivas or New York. A team averages a goal on 9.4% of its attempts this season, and 26.9% of its shots on goal. Using that, there's a possibility that he adds a few additional goals to the line-up. Assuming he is just average at finishing.

That said Agudelo has beat the average ratio over his 3,000 minutes, scoring 11 goals in 36 shots on goal (30.6%). Scoring goals is a skill, and though we don't know how much is luck vs. his ability, I think it's very possible that he will continue to beat the league average conversion rates.

Looking at Chris Wondolowski, Kenny Cooper and Álvaro Saborío--the top 3 scorers for 2012--they all combined to beat the league average by scoring a goal on 43.9% of their shots on target. So, we can safely attribute scoring goals on shots on target as being a skill, the only problem is trying to account for luck. That's a little difficult at this stage, and so for now, we'll just mention it.

But assuming that Agudelo is consistent and continues scoring at a high rate and matches 1500 minutes. I have him for about 6 goals this season. Right now considering their goal conversion and their already abysmal offense, the Revs are on pace for 36 goals total to end the season. Considering their ability to suppress their opponent's talent and ability to score goals I have them for 26 goals allowed, assuming they continue their defensive supremacy.

Using SoccerMetricsPythagorean this comes out at about 51 points... given the asinine goal difference. Add in the additional 6 goals that Juan Agudelo brings and that brings them to a total of 56. Basically almost a full point for each goal.

Now, I'm not about to say that the Revolution have a shot at 60 points, not in your life. But in the last 3 years the only teams to have a plus goal differential AND not make the playoffs were 2011 Chicago Fire and the 2010 Kansas City Wizards. Considering a team-wide return to a league-average ability to score goals AND adding Juan Agudelo, it's very possible that New England just moved themselves within striking distance for the 4th or 5th spot in the East.

Columbus and Philly, beware.

Game of the Week Recap: Dynamo at Galaxy

The Los Angeles Galaxy’s Landon Donovan blew a PK in the 25th minute, and the Houston Dynamo managed a goal in the 56th, stealing three points in LA. Definitely not what our expert panel of misfits projected on Saturday’s podcast.  The Galaxy controlled possession (59.3%), won more duels (55%), and earned more attempts (19 to 14), but earned nothing in the standings for its work. Here’s a chronological summary of Houston’s shots:

Houston Dynamo shot times - May 5 2013

There are actually two shots taken in succession there before Houston’s opponent missed a PK, but the disproportionate bulk of Houston’s shots, including its goal, still came after it almost went down 1-0. That’s probably just random, but interesting nonetheless. Here’s something that’s almost surely not random. Though the Galaxy won the attempts battle, many of those attempts were blocked, and many of those blocked attempts occurred after Houston took the lead. Observe:

LA Galaxy's blocked shot times - May 5 2013

It seems the Galaxy began to get desperate, and this brings up the concept of game states which we discussed in the above podcast. Teams are likely to employ different strategies depending on the score of the game. While the Galaxy out-attempted the Dynamo during the game, many of those attempts appear to have been of the low-probability type.

The Dynamo jumped up to third in points per game, and the updated tables can now be seen here!

*Thanks to Opta and MLS Soccer for the sweet images!

ASA Podcast: Episode V

Hello loyal readers/listeners! My apologies for not having this up yesterday, but it's here now. A 55-min. bit on Markov Chains, Game States, and a preview of the Galaxy vs. Dynamo, while also introducing a new member of the American Soccer Analysis team. Hope you enjoy! [audio http://americansocceranalysis.files.wordpress.com/2013/05/asa5.mp3]

Game States: An Attempt at an Introduction

I'm no mathematician. Matty maybe, but I am not. So when approaching something like Game States, I felt it good to attempt to introduce it with something, though it's rather ominous and a bit intimidating. So most--if not all--of the information provided is taken from a source who is smarter than I am. That's really what this blog is all about, finding people who know and understand the principles we are trying to learn and centralize the material and keep it in tidy location where people that are new---not just to the sport, but also the concept of analytics---can go to find information and grow their knowledge.

The idea of game states is that a match will consist of a sequence of states, where each state is defined by a combination or series of events that culminate in creating a new state. Those events give details and help break down the match. They provide context or meaning to the data that we record. Game states, as I understand it, is based upon the idea of the Markov Chains.

...[A] mathematical system that undergoes transitions from one state to another, between a finite or countable number of possible states. It is a random process usually characterized as memoryless: the next state depends only on the current state and not on the sequence of events that preceded it.

Let's apply this idea to a sport...

In football, if a team is in a certain situation, what happened previously has no effect on what will happen next. For example, if we have a 1st-and-10 from our own 20, it does not matter if the previous play was a kickoff for a touchback or a 10-yard gain for a first down after a 3rd-and-10 from the 10-yardline. Either way, we now have a new situation that will only directly affect the next play.

To put it all into soccer terminology, if a team has possession of the ball and is progressing past the half-line and into the attacking third it doesn't matter if they got it on an interception or a goal kick. Regardless of how it happened they now have the ball going into the oppositions attacking third and will have an opportunity to threaten and score. I struggle with this thought because tactically this may not be true---an individual getting the ball on a break is different than someone participating with a soft building of play in attacking the opposition's net. Coming from a statistical and memory-less position we simply want the facts.

Let's go to another sport and put this into baseball terms because of the advanced progress that the sport in general has made in analytics. Baseball analytics breaks game states down into four basic concepts: the score, inning, base runners and outs. This would be the way you could calculate basic run/win expectancy and ratios.

If you have the average number of runs expected to score in an inning after any game state, you can figure out how many runs a stolen base is worth, or a triple, or a strikeout. The game state essentially allows us to relate everything that happens on the diamond back to the major currencies of baseball: winning and runs.

We're not talking about baseball. We're talking about soccer, or for you euro snobs, 'Fútbol'. However, taking the concepts of an already practically applied matrix, such as baseball and the one that that Tom Tango has already developed(see below), can give us ideas on how we can attempt to create a corresponding one in soccer. Soccer has wins too---though I see more people associate points---but our true currency in this sport is goals. Everything leads back to the price or value of a goal. Whether it be a cross or a tackle, the ultimate result of what we want is to be able to understand how things work together to produce goals.Comparing the possible game states; Soccer just--just like in baseball--has a score line that can help give us an idea of the transitions between states. A team being up one goal, or on the other side of the coin being down one goal, can give us a definition of a game state. It's simplistic, but it works. It also can show us why some times, looking at you Alex Fergueson, teams take less shots than at other times. It would make sense just for the purpose of proper possession.

We also have measurement for the length of the game, in that each team will play against one another for a total of 90 minutes. In soccer, we have time intervals. The largest problem is that because it's constantly moving rather than a set static state, such as an inning, it creates a lot of various probabilities and chains. However, if you wanted to mitigate that to an extent you could instead just revert to the basics of using the first half vs. the second half of a match. While these are two big time intervals, when used in conjunction with other specific game states it could continue to help us develop a better understanding of the game.

The next one listed is base runners, but I'm going to pass on that and move on to the concept of outs. This is something that at very best I can say is a difficult correlation, but if you wanted to attempt one I might try changes in possession. This is a rather poor concept and I have no idea how or if you want to use this... probably not. Baseball limits a team to three attempts to score per their half of the inning rather than having 3 specific possessions or attempts to score. In the first 45 minutes of a half you could have a team with anywhere between 12 and 40 possessions alone depending on who the opponent was and how the team executed its attack.

Back to base runners, and this is one of the easier things to mimic though I have no way of proving how closely they are related. Shots on goal  conceivably could give us a baseline on the probability of goal scored and, more importantly, the points that are associated with winning or drawing.

One additional game state that hasn't been mentioned and doesn't really correlate with anything that baseball has, is yellow and red cards. While, baseball has ejections it doesn't necessarily affect the dimensions of the game. However, in soccer it puts the team down a man, rather than being able to just sub in a replacement.

These are all elements which you would consider context. How do teams perform during these situations? Do their possessions last longer? Do they take more shots? What are the quality of shots? This is the information and really the purpose behind gathering the data points. Are the Seattle Sounders just as likely to score in a "-1" goal situation as they are in a tie-game?

This is all information that we are seeking. The context can provide further details as well as specific entry into how certain aspects and statistics can be properly correlated to goals, and then to points toward the table.

I feel like there is more to write about here... and that's because it's true. There is a lot more to write about. But this is primarily to cover the basics of game states. We'll talk more about it in our podcast tomorrow, and we'll have a follow up post to everything when we start preparing to post MLS game state information.

A thought on Big Data and Club Analysis

I talked a bit about Big vs. Small data last week--in case you missed it, go back and check it out--and we kind of talked about how you don't have to necessarily rely on the revolution of big data. There is a need to make do with what is currently available. However, while that big data is sometimes available, there are other encumbrances to deal with:

Beyond the complexity and time constraints placed on the analysis, another major obstacle faced in the job – like that faced by so many people entrusted with big data within an organisation, football club or otherwise – is to make data useful, accessible and engaging to colleagues who have little interest or experience in dealing with numbers.

This from a recent interview with Ben Smith of the development performance systems at Chelsea FC. A club that is often quoted as one of the "big-4" in the English Premier League. It's important to understand that, while many of these clubs have information at their disposal, few (if any) know or understand the practicality of implementing the information into their planning and preparation phase.

If fact, reading back on the 'Counter Attack' blog by Richard Whittal, some clubs--i.e. most--don't pay their club analysts. That should give you a brief, if not all together insulting, view of how much they respect the value of the service provided. I'm not saying they don't see it as useful in "some capacity", I just think that, in terms of how much they pay the rest of the staff, they could afford to have a full time analyst, especially for what an analyst has the potential to provide.

But it is not, of course, just the coaching and scouting staff that benefit from the big data analytics being carried out at the club, the players are also reaping the rewards of the work across the club. He says: “Every one of Chelsea’s Academy players from the age of nine has a personalised development programme."

There is some interesting stuff here to think about. It sounds a lot like how Ravi Ramineni has started helping out David Tenney, Sounders FC fitness coach, over the past 6-8 months.

By the way, h/t goes to Ravi who linked the article from his twitter.

Game Of The Week Recap: New England Revolution vs. Philadaelphia Union

It's Monday morning! Hope you've had your coffee, been to your morning meetings and are ready to catch up on the MLS weekend. We're not probably at the point to where we're going to review games or teams on a consistent basis. That's not really our thing... at least, not yet. But Saturday on the most recent podcast, Matty and I talked about our game of the week, featuring the New England Revolution at home versus a high powered Philadelphia attack. The Union boys are a club that, in general, doesn't keep possession long, nor win it back. But they rely specifically on creating as many chances as possible and finishing the ones they are supposed too. It's partially why they are an improved club this year and one that has a legit chance to get into the playoffs and do some damage in the East. They were coming up against a New England team that averaged less ball possession, won the ball back less and took a whole lot less shots---and the Revolution was even allowing its opponents twice as many opportunities as it earned for itself.

What it came down to was that Philly was a borderline good team that sat near mid-table, and New England was arguably one of the worst-seeded among those in the bottom. The surprising thing, and partly what made us choose this game as our game of the week, was simply how split the analysts were at MLSsoccer.com. Two were for a draw, two for a Union win, and two had the Rev's with a win. All the way around it was kind of an interesting match-up of a slumping team against a club that seemed to be on it's way up.

If you didn't hear the podcast, I had the Union 2-1, and Matty went even further with a 2-0 Union beating on the road in Foxborough. Both of us look a little silly this morning with the Revolution pulling out the win. So, how exactly did this band of struggling MLS misfits steal 3-points? Simple, they finally took and finished shots.

The first capture below is the Revs and their chances created in the attacking third. This is shots, key passes and corner kicks and successful free-kicks leading to attempts.

RevsAttackingThird-20130427

The next shot is clearances, interceptions, recoveries, tackles, blocks, saves and punches made by the the keeper/back defensive four of the Union inside their own box... it's kind of a mess.

UnionD-20130427

The bottom line is that for the first time, possibly all season, the Revolution were dangerous on the attack and they not only put shots on target, they legitimately tested the opposing defense. Then, once they had a lead, they allowed their strong defense to carry them home for the full 3 points.

Almost as convicting, the Union were only able to create 6 real attempts on goal in the second half and only 4 after the Rev's finally broke the stalemate. While they weren't bad on the attacking end, it just simply didn't match what the home team was accomplishing, and it left Philly out points at the end of the day.

I think this sort of game underlines the amount of parity that is in this league right now and also the fact that you have to consider a team's incredibly poor finishing rate as not being an indication of true talent---at least, not this early in the season. New England basically out-Union'd the Union, by beating them while playing their own style of play on the pitch. Overwhelming them with shots and finishing opportunities. Both teams will face exceptionally tough tasks by playing Northwest clubs (Portland and Seattle, respectively) next week, and it should be interesting to see what comes of it.

Montreal's Paradox

If you have listened to our podcasts or read through our stuff, you will have heard us talk about shot ratios a lot. That's how many shots a team gets divided by how many shots its allows its opponents. A shot ratio of 1.5, for example, means that a team gets one-and-a-half times as many shots as its opponents. When soccer teams create extra opportunities for themselves, it generally leads to more goals and more points in the standings. And then there’s Montreal. The Montreal Impact has been something of a Cinderella story this season, at least statistically. Leading up to its matchup with the Chicago Fire on Saturday, the Impact had recorded the second-worst shot attempt ratio in the entire league. Montreal had earned just 61 shot attempts with 28 on target to its opponents’ 95 shot attempts with 32 on target.  Yet somehow, the Impact had maintained a positive goal differential (+2) and the second-most points per match right behind FC Dallas.

Against Chicago, Montreal not only won on the scoreboard two-nil, it also won the shooting and possession battles. But that is a rare feat this year for the Impact, and it’s worth posing the question: Has Montreal been lucky this season, or does it do things that shot ratios and possession just can’t explain?

Using just shots on goal for now, I regressed goal scoring ratios against shot ratios to see how teams “should do,” as if shots on goal were the only thing that matter. Even this early in the season, the regression was not all that bad (R2 = 0.4). It also said that Montreal’s 0.94 shot ratio should lead to about the same goal ratio.* Well that makes sense. If you generate roughly the same number of shots on target as your opponents, you should score about the same number of goals. The Impact, however, have scored nine goals to its opponents’ five—a 1.8 ratio, or +4 differential, if you prefer.

An obvious thing to consider is finishing rate. Despite being outshot, the Impact players finish their attempts with goals more than twice as efficiently as opponents do. That ratio is the best in the league. My first instinct is that the Impact has been somewhat lucky, and that opponents will start to finish with more frequency. But there are two possible explanations I want to explore first before waving the cliché luck flag: the quality of opportunities for Montreal and the quality of opportunities for its opponents.

Harrison talked a little bit about Montreal’s counter-attacking style during a recent podcast, and there’s a possibility that the Impact’s style allows low-quality opportunities to its opponents, leading to higher-percentage opportunities for itself on the counter attack. (Before we investigate, it should be noted that Montreal’s schedule has featured teams that average out to be, well, league-average when it comes to finishing.)

Let’s take Saturday’s match against the Fire as an example of the tools I’m using. Check out the Opta chalkboard for yourself here, and you can see from where teams are shooting and scoring by clicking the appropriate boxes for team and statistic of interest. During this particular game, I have Montreal down for 16 scoring attempts, nine from outside the box, six inside, and one from right on the edge. Both its goals were scored from inside the box (though you could argue one was one the edge). Chicago, on the other hand, earned 11 attempts, ripping seven of those from outside the box, just two from inside, and two from the edge of the box. Chicago did not score. I did this for each of Montreal's seven games this season.

Obviously things like angle matter, too, but I’m not going to pull out my protractor for this one. Here’s the breakdown for Montreal and its opponents on the season:

Attempts Goals Finishing
Stat Montreal Opponents Montreal Opponents Montreal Opponents
Inside Box

40

45

6

4

15.0%

8.9%

Outside Box

31

56

3

1

9.7%

1.8%

On Edge

6

5

0

0

0.0%

0.0%

Total

77

106

9

5

11.7%

4.7%

 

Montreal earns more shots inside the box than outside, and that might very well be a product of its system and players, rather than just dumb luck. While the Impact is being outshot in total, perhaps that stat is skewed slightly by shot selection. Montreal's system seems to create a greater proportion of opportunities in the box. I would still expect some regression from Montreal this season back toward the middle of the standings—as its shot ratios are not favorable even after adjusting for quality—but perhaps not as far as a simple shot model would suggest.

*One might note that Montreal’s attempts ratio is quite a bit worse than its shots-on-goal ratio, which isn’t even that good to begin with. It is apparently too early in the season for attempts ratios to explain much of anything with certainty, but shots models from past seasons suggests Montreal’s goal scoring ratio should probably be even worse than even-ish. That is, if shots aren't broken down by quality.

ASA Podcast: Episode IV

It's Saturday and thus time to drop yet another American Soccer Analysis Podcast on you all. As we've been previewing this past week, we're going to talk a bit about some of the up-and-coming MLS rookies that have played a big role on their teams to this point in the season. We'll add in our thoughts on using a Plus-minus statistic for measuring individual performance, as well as our selections for game of the week. Hope you enjoy.

[audio http://americansocceranalysis.files.wordpress.com/2013/04/asa4.mp3]