It’s Always In The Numbers

I’ve finally put my finger on why I wasn’t such a huge fan of Soccernomics. I had the wrong expectations of the book.

Number-crunchin'

Beyond the chapter that TSG reader Patrick pointed out on the Champ’s League Final where Edwin Van Der Saar plays a little game strategy with the Chelsea penalty kick-takers, I think one reason I wasn’t enamored is that I wanted more on the in-game play.

What metrics can we put on in-game play, like Billy Beane does with the Athletics, like Daryl Morey does with Houston Rockets, to determine if a player has played a good game or a poor game?

I’m not looking for tendencies, I’m looking for evaluations.

In terms of statistics, we’ve seen a few individual statistics start to creep into the broadcast of the game. The one that pops into mind is showing a player’s “pass completition” when they get subbed out.

I’m not sure how that lone statistic really gives the audience a read on just how that player contributed, or deducted for that matter, from his team’s play.

So in that vain, I’m going to take a shot at some individual statistics that may or may not give you a better idea of how player made an impact on the game:

Goalie:

Traditional: Saves

Saves is a tough one for goalie, a necessary one, but it doesn’t communicate the quality of those saves.

Additional:

Paries: A judgement call here–much like an error in baseball. A parry is a save where it is ruled that the goalie’s was unable to keep possession on the save because of the quality of the shot. Tim Howard would pretty much lead international competition in these.

Service Possession Maintained/Service Possession Lost: Not sure this one works. You tell me. How would we be able credit a player like Pepe Reina who’s service is a weapon. Let’s remove outlets to defenders from this one; so no defender pass completion in the stats.

When a goalie winds up and punts the ball, does his team maintain possession? I like this one also, because if you’re a goalie you are going to start punting it away from someone, like a Robbie Findley, who’s not going to win that reception.

I'm betting these defensive numbers would show that Simon Kjaer (right) is closing in on top defender status...

Defender:

Traditional: N/A

Additional:

Tackles Made/Tackles Failed: Make a tackle and earn possession or disrupt the play? +1. Miss a tackle that concedes an offensive chance? -1. That simple.

Attacks Defended/Attacks Conceded: This likely the best wingfull defensive statistic. A winger possesses the ball and initiates an attack. Initiating the attack is important because it’s putting pressure on the defense. If a defender concedes a shot on goal, a cross or a completed pass still in attack to another offender, then it is a Run Conceded. No dice, Run defended.

Offsides Called/Errant Runs/Errant Runs Scored: Love this one. It’s a team statistic, but much like a penalty in football, you can label the guilty party in this one.

First, the definitions. Offsides Called, again self-explanatory. Errant Runs? This is when an offender, like a Charlie Davies, get behind the defense, another offender finds them with the ball and the defender has to track back and make a play or the goalie does. If the player on the run scores, Errant Run Scored.

The box score for a match might look this might look like this:

OC: 9

ER: 3 (Bornstein (2), Spector)

ERS: 1 (Spector)

Now for an offensive one:

Passes Completed Not Under Duress: I’ve been trying for over 6 months to work this “frustration” into a column. One of my biggest frustrations in game play is when a defender who is not under any sort of duress, makes a simple square or just a check down to the goalie that is imprecise and puts the receiver under duress. I don’t understand how players can’t be more precise at the highest level of soccer. It irks me to no end when a team may create an advantage through quick, precise ball play and the simple first pass is off-the-mark and the chance is gone.

Now I have a rather lengthy definition to take care of it for me. Look out Carlos Bocanegra and you too Jonathan Bornstein, you’re under my microscope.

Midfield:

Traditional: Pass Completion?, Assists, Shots, Goals

Additional:

Xavi would be off the charts with these metrics...

Chances Created: I don’t know, you tell me who wins this one Xavi? Sneijder? Cesc? A chance created is simply when a touch, a pass, or even a tackle allows for the offense to have a scoring chance. If we need a definition of a scoring chance, it can be any chance where a defender has to make a play to prevent a goal or the offender makes or misses a shot.

Another judgment call here, but players like Xavi who so often “unlocks” the defense would finally have a statistic to be proud of.

Tackles Made/Tackles Failed: See above.

Attacks Initiated/Attacks Completed: The inverse of the defensive definition of Attacks Conceded/Attacks Defensed. Note: An attack completed does not necessarily mean a score.

Striker/Forward:

Traditional: Pass Completion?, Assists, Shots, Goals, Fouls Suffered

Additional:

Crouch: "I finally have a stat for me!"

Service Received/Service Lost: You would like you this statistic this year to showcase Jozy Altidore’s improvement this year. What about the guys like Altidore, Heskey, Crouch whose primary role isn’t to score goals. Is it really fair to evaluate them exclusively on scoring statistics outright?

Here’s the one, SR/SL,  that matters for those guys.

Chances Created: See above.

Attacks Initiated/Attacks Completed: See above.

Miscellaneous: Team:

Possession Time Vs. Shots On Goal Index: There has been much dialogue lately about whether possessing the ball for the majority of the game possesses any advantage itself for the team which does it.

TSG’s contention is that possession is really a means of defense for many teams.

Here’s a simple statistic that should aid in review of the game. How about an index of time of possession vs. shots on goal or chances created.

Easy enough? Too easy. A team that maintains possession in a non-threatening way by knocking it around the back would be very low on this index; where as the United States against Spain in the Confederation Cup would score very high. This one is still a little rough I confess.

Miscellaneous: Individual:

Positive Service Percentage: Landon Donovan would likely win this stat in MLS; Sacha Kljestan would not. Essentially a player’s free kick or cross arrived in a location that created a chance. Obviously balls that went directly out of bounds or balls that overshot the peloton of players in the box would be negative.

Stupid Stupid Fouls (SSF): Fouls you shouldn’t have made. I’ve added these in to help evaluate Michael Bradley, Rico Clark and Jonathan Bornstein.

About these ads

23 responses to this post.

  1. Just a quick idea before I go to bed, but I would love to see a statistic on first touch. Perhaps a +1 for a touch that continues the run of play, a zero for a touch that ruins the run of play, and -1 for a touch that loses possession, then divide the total by number of touches. It’s subtle stats like these that I think are critical in building a successful USMNT, as we lack players with the speed and on ball ability to break down defenders, so positive linking play will be key for a potent attack.

    Reply

  2. Posted by robbie on 2010/03/29 at 7:18 AM

    I think this is a great idea, in principle.

    My only issue is that many of the line items are judged arbitrarily. Basketball and baseball are so statistics based that it is extremely easy for them to judge based on their set parameters. With this system, like an error in baseball, you would have differing opinions on almost every criteria if you ask a bunch of different people. In other words, only one person would be able to evaluate players because someone else would have a totally different score on them. And even then who would be right?

    It’s too hard to judge. It’d take probably 5 hours to do one game by the time you rewind and record all your notes. It’d be impossible for clubs to use this system but a cute way for websites to grade players after a game.

    Even then I have a feeling that the eyeball test would be just as accurate in rating a player….they’re essentially the same thing.

    Reply

  3. Posted by matthewsf on 2010/03/29 at 7:39 AM

    Robbie:

    Thanks for commenting and thanks for asking those questions. I was hoping I would get similar questions; so thank you. All fair points I might add.

    In regards to your first question, let me answer it from the team perspective. If I’m a team with as big budget like Manchester United, I likely employ an odd number of folks to review games of my team and the opponents.

    Over time, each person’s bias will become apparent, maybe you take the average over time or throw out the high/low on a game or come up with an index. But accounting for a “bias” should be relatively simple over time with multiple observations of the same game/experiment.

    Also, an important point that I was considering making in the post, is that this system would be used to augment human eyes. I’m a big fan as well of making the following statement (in business) “In the absence of empirical data, the expert with the most observations should make the call.”

    Same here both you and I would probably agree that Fernando Torres’s impact on the game, for example, goes beyond the numbers. But what if–and this is not true–I told you, say, Carlton Cole had nearly identical stats (the ones above, not just shots/goals). You might appreciate Carlton Cole just a tad more.

    I don’t think you need to rewind the game at all. I’m actually going to try it for a game at some point in the next few weeks.

    Thanks for the comments.

    Reply

    • Posted by matthewsf on 2010/03/29 at 7:51 AM

      By the way, I thought I would add this — I always have a sense if people like a column at TSG or not. When I look at this numbers, this usually backs it up. Things I discount in my columns’ number are “day of the week” and “competition” — for example, I didn’t expect to get a lot of readership on Super Bowl Sunday — the number bore that out.

      But I have been surprised at least twice by the numbers not reflecting my thoughts.

      In one case, I “overrode” the numerical data and decided to do a follow-up piece on the post.

      Don’t mind sharing — that was this one (http://theshinguardian.com/2010/03/15/ten-national-team-trades-wed-like-to-see/) — which I likely will follow-up on again.

      Reply

  4. Matt: Your attempt at creating a statistical evaluation system is commendable. But, don’t reinvent the wheel. Check the French publications, they are sticklers for football statistics, and have a lot of these categories, but in more, eh-hem, sophisticated terminology.

    With that in mind, these statistics will mean nothing unless a governing body produces a study of what’s important, how to point it out, and how it is valued. Only then will it be of any importance to anyone in the states.

    You’re closer to achieving this goal than most just by posting about it. No one in the American soccer community touches on this regularly. Everyone starts somewhere, I mean, no way they had a Passer Rating statistic (you know, the one NFL QBs are judged on, the 160.7 point scale… ridiculous stat given the complexity of formulating it, but worth its weight in gold when evaluating a free agent) even as early as the 80s without first starting the conversation that traditional statistics don’t tell the whole story.

    On an unrelated note, Jen Chang siting…

    http://sportsillustrated.cnn.com/2010/writers/jen_chang/03/22/mls.liverpool/index.html

    Reply

  5. Posted by George on 2010/03/29 at 10:01 AM

    MLS used to keep a stat for keepers that i thought was pretty useful but i think they stopped keeping track it was the amount of catches or punches a keeper would make. this stat gives you an idea of how active a keeper is coming for crosses and chips in to the box. maybe you could add some type of stat for cutting off through balls and it could be almost like OPS (on base + slugging) you could have CPT (catches+punches+throughballs).
    CPTS(catches punches throughballs and saves) would give you a more accurate account of how effective a gk is during a game

    Reply

    • Posted by Matt Mathai on 2010/03/29 at 10:29 AM

      The only problem is that tracking the number of catches/punches only makes sense if you also have the context of knowing how many crosses the opposition sent in. It’s possible (although not likely) that a team that lives off the ground game might have no crosses at all, in which case you could look at the GK’s stats and say “He sucked. He didn’t punch or catch a single cross all day!” And that wouldn’t reflect what happened in the game.

      Reply

  6. Posted by Matt Mathai on 2010/03/29 at 10:24 AM

    Hey,

    Just a couple of thoughts.

    1) I understand the desire to be able to evaluate the game more objectively, but I’m not sure that soccer lends itself to that kind of analysis. Stats like goals scored, minutes played, fouls committed/suffered, even miles run by a player, are pretty objective and easily obtained. There are other similarly obvious stats out there. As soon as go past those, however, it gets murky quickly. We use ‘assists’ all the time. What does that mean? What is a ‘Shot on Goal’?

    What I’ve found is that years of following and playing this game have given me the ability to watch a game and to be able to tell who is playing well and why. I have often looked at stats from that game and been surprised because they appear to tell me things that counter my own judgement. So, being arrogant, I dismiss the stats because I know what I saw. :)

    2) Maybe I misunderstood your point about “Passes Completed Not Under Duress”, but it has more to do with the passer than the ‘passee’. Presumably putting the receiver under duress with a pass would result in fewer of those passes completed. I believe they’re referring to someone who can’t send the ball to another teammate even if he’s not being harrassed by the opposition. This stat breaks down too. We’ve all see predominantly negative players, the kings of the square ball, who might complete all kinds of worthless passes, but miss the chance to make a killer through pass that cuts out defenders. A brilliant through ball (think Donovan to Davies at the Azteca) is thrilling, but a much lower-percentage pass, and having the confidence to attempt such a pass will inevitably result in a lower “Passes Completed Not Under Duress” stat.

    Bottom line, I don’t read box scores for soccer other than to tell how many goals were scored, by whom, in which minutes, and how many cards were shown. Oh, and who the subs were. I definitely don’t look to any stats to tell me which team played better.

    Reply

    • Posted by matthewsf on 2010/03/29 at 12:08 PM

      Matt, thanks for weighing in.

      On 1) I have to disagree, they said this about basketball for awhile after baseball starting operating more in the Moneyball manner and now the Houston Rockets — read the article that is linked above — are forcing people to change their mind.

      In both business and in sport, it’s about unveiling the correct metrics. Which are the ones that the game hinges on? Be it team and/or individual stats…

      I think in soccer, there is not enough data stored and processed, but also incorrect stats. For example, Shots On Goal…what does that give me.

      A certain degree is a labeling a moment, an “event” as an inflection point, defining it, and then looking for that “event” on more occasions. Part of that I tried above.

      On 2) My passer under duress is specfically focused on the defender who has to make many passes under pressure. If we wanted to get more technical, we could create an exception that excludes “risky” or “chance-creating” passes.

      All subjective, but in reality it’s more of a process:

      Step 1: Learn what metrics matter Step 2: Observe metrics over time

      Reply

  7. Posted by CJ on 2010/03/29 at 11:03 AM

    Diggin the concept, can’t wait to see you implement the “beta” test.

    Reply

  8. How about tracking the distance of the pass and the space between the passer and the nearest defending player on each play.

    If you take the sum of the distance of completed passes and divide it by the sum of the spaces and divide completed pass you get a statistic that sums up a players technical and creative abilities. If they’re Xavi, pinging 40 yard passes in a yard of space, their number will be very high, if they’re a centerback passing the ball to your defensive partner in acres of space, the number would be quite low.

    There’s no human judgement involved besides determining what is a complete and an incomplete pass.

    You could also look at how frequently a players completed passes lead to another player completing a pass, a cross or taking a shot.

    Reply

    • Posted by matthewsf on 2010/03/29 at 12:00 PM

      That’s a great idea — only challenge is building the software for it (which I imagine someone has already) or I’m going to have to keep holding a ruler up to the television and I don’t have a flat screen yet. :>

      Reply

  9. Posted by Antonio H. on 2010/03/29 at 12:32 PM

    Favorite piece so far :). I’m a man of remembering stats, and I guess this relates the beautiful game to the other sports that I play/watch like basketball, football, volleyball, ping pong, etc.

    Also, I can relate to your frustration of getting really good ideas down in stone. Especially when I have to write essays on crappy books like Heart of Darkness

    Reply

  10. Posted by Rhodie on 2010/03/29 at 10:30 PM

    Interesting piece, one of the things that I enjoy about the game of soccer is that it is 50% intangables, therefore defying statistics. That said I think that statistics are still wonderful.

    I would say that there are two different uses of statistics, one is to determine the type of player, the other is to evaluate players. Some that I would like to see:

    Positive passes/Negative passes and the completion percentage of each
    Passes to feet vs. passes to space (this used to determine the type of player)
    Attacks percentage (team statistic), this would be calculated as the number of attacks with a positive result (corner, shot on target, near miss?), divided by number of attacks.
    Crosses come for vs. Crosses disrupted (keepers)
    Average touches before passing (type of player)
    Time spent defending vs. Time spent attacking

    That was everything that I was thinking, I would love to hear your opinion on these.

    Reply

  11. Posted by GeorgeCross on 2010/03/30 at 11:39 AM

    Matthew – great idea for a post. And this is a topic that gets me hot under the collar! I also tend to agree with Matt Mathai’s post.

    When talking to my friends about sport, I have always been amazed about the statistical element in US sport. I am not suggesting that performance measurement analytics and statistics are irrelavant – far from it, but more to the person watching live sport when they don’t have the paper to pore over yesterday’s game – I find it irritating to be honest. If you’re watching the bloody game and know what you’re talking about, you should be able to tell who’s on top without the plethora of data to regurgitate in order to sound knowledgeable.

    The other thing is that football is a free flowing game, a continuum if you like; where MLB and NFL are basically a series of individual events, especially baseball (I am guessing many people have read Moneyball?).

    Tuesday mentioned Xavi pinging a 40 yard pass – and I agree, that shouldn’t have the same weighting as a sideways pass from a ball-winner to a teammate – but what about the element of Henry getting onto that 40 yard pass, sometimes the fact that Henry is so quick makes it look like the pass was perfect, whereas a slower player like Ibrahimovic might not have gotten on the end of the same pass. Would you measure that ‘identical’ 40 yard pass the same?

    It’s a little like the possession statistic – why does that matter if you you cannot get behind the opponent? If the ball is going from side to side, in front of you, in the middle third, why should that be a statistic? What is the relevance? This should be of interest to USA fans as you say that you’re a counter-attacking team…

    I sort of like the chances/conversion ratio for strikers, but then not all chances are the same…

    So while I do like statistics, I also know to tread carefully with them – and nothing beats watching the game and appreciating something that you cannot fat-finger into SPSS and ‘measure’…

    Reply

    • Matt, while I like your ideas and think you’re doing some important work with statistics, I honestly don’t feel we need to obsess over them like baseball fanatics. Exhibit A: the NHL.

      When I compare two sports I try to take into consideration the flow of the game, whether or not the game is broken down in little parts (football, baseball) or flows continuously (like soccer, basketball, and hockey). Since hockey and soccer are similar, in terms of the flow of the game, I went to ESPN’s NHL site and poked around. My findings revealed that hockey fans are not given the plethora of statistics that baseball and football are accustomed to. Actually, the statistical categories are very similar to that of soccer’s: goal-tending stats like saves, shots, save percentage, and clean sheets to name a few.

      What stuck out most was, that aside from the penalty stats, hockey doesn’t have an overwhelming about of stats like that of pointyball and stickball.

      My conclusion: we should enjoy the simplicity and mystique of the game “as is” and judge a team on its chemistry, momentum, and good ol’ luck of the bounce.

      Basically I’m agreeing with George (first time perhaps?). But, I would love to have more stats, I’m just not sure I have the mathematical ability to dream up all of the scenarios necessary to do it accurately. Because if it’s not done scientifically, the stats don’t mean anything. And I’m writer, I don’t do science (that well).

      Cheers George, you swayed me on this one brother!

      Reply

      • Oh, one more thing George, stats give some people the notion they know what they’re talking about, but sometimes it clouds their judgement–like when my St. Louis Cardinals backed into the playoffs, losing some 15 out of their last 20, and were given no chance of competing because statistics showed teams that slumped towards the end of the season never do well… we went on to win our 10th championship.

        Reply

  12. Posted by GeorgeCross on 2010/03/30 at 11:45 AM

    Matthew – great idea for a post. And this is a topic that gets me hot under the collar! I also tend to agree with Matt Mathai’s post.
    When talking to my friends about sport, I have always been amazed about the statistical element in US sport. I am not suggesting that performance measurement analytics and statistics are irrelevant – far from it, but more to the person watching live sport when they don’t have the paper to pore over yesterday’s game – I find it irritating to be honest. If you’re watching the bloody game and know what you’re talking about, you should be able to tell who’s on top without the plethora of data to regurgitate in order to sound knowledgeable.
    The other thing is that football is a free flowing game, a continuum if you like; where MLB and NFL are basically a series of individual events, especially baseball (I am guessing many people have read Moneyball?).
    Tuesday mentioned Xavi pinging a 40 yard pass – and I agree, that shouldn’t have the same weighting as a sideways pass from a ball-winner to a teammate – but what about the element of Henry getting onto that 40 yard pass, sometimes the fact that Henry is so quick makes it look like the pass was perfect, whereas a slower player like Ibrahimovic might not have gotten on the end of the same pass. Would you measure that ‘identical’ 40 yard pass the same?
    It’s a little like the possession statistic – why does the possession statistic matter so much if you cannot get behind the opponent? If the ball is going from side to side, in front of you, in the middle third, why should that be a statistic? What is the relevance? This should be of interest to USA fans as you say that you’re a counter-attacking team…
    I sort of like the chances/conversion ratio for strikers, but then not all chances are the same.
    So while I do like statistics, I also know to tread carefully with them – and nothing beats watching the game and appreciating something that you cannot fat-finger into SPSS and ‘measure’…

    Reply

  13. [...] thing that is bound to happen – more stats will become prevalent for the fans to consume and the teams to leverage.  I feel like the EPL is [...]

    Reply

  14. [...] stat rant done till next time. If anyone wants to give us a hand cataloging the stats here, let us [...]

    Reply

  15. [...] • Unrelated: Simon Kuper looks at a statistical revolution underway in football. (TSG says it’s always in the numbers also.) [...]

    Reply

  16. [...] What new metrics are important or would you create? Do any of these (link) make [...]

    Reply

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

Follow

Get every new post delivered to your Inbox.

Join 220 other followers

%d bloggers like this: