Here's the link. Includes the data for every player in the past season of Major League Soccer. Only regular season games are included, and only full strength, 11 on 11 play was considered. All minutes and goals after red cards occurred were removed. I've broken the data down by team, so you can easily see the 11v11 minutes played, GF, GA, +/-. and +/- per 90 minutes for each player. I've also included lists of the league leaders in plus/minus per 90 minutes, both good and bad. You can find those top 25s in both unadjusted and adjusted forms. The latter accounts for the team's overall plus/minus per 90, and was possibly the most interesting part of the post to me as it allows you to see who stood out from their teammates. There were some fascinating results like Jacob Peterson and Mac Kandji in the top 25 and Guillermo Barros Schelotto and Steve Zakuani in the bottom 25, check it out.
Why wouldn't you include goals from 10v10 situations? I'm pretty sure in hockey that all even strength goals go into consideration. Considering how wary most experts are of this statistic in ice hockey, I was prepared to look at your list and be underwhelmed. But I'll be damned if the stats for LA Galaxy don't line up with the way I would have expected them to. Maybe this is a stat MLS should consider tracking and going back retroactively and doing it historically as well. Would be interesting data.
Wonderful information. From a Sounders' perspective it kinda highlights what many of us suspect about Zakuani -- that he's not much help on defense. You have to wonder if someday he is moved up to striker to allow more of a two-way player on the left flank.
Why statistically penalize players that are on a full 11-man squad when someone on the opposite team makes a poor play and gets sent off? I'm sorry, but I just have a hard time thinking this stat means anything when it ranks Facundo Diz and Jacob Peterson above Omar Cummings and Conor Casey, and the factor that ultimately ends up placing players in this order is someone on the other team screwing up. It's like car insurance rates.
Because most of the time, 10v10 comes after a 11v10 situation and the game's already been affected by that. I feel it's easier just to ignore all non-11v11 situations. It wouldn't change much, there was only 126 minutes of 10v10 soccer (and 3 goals) this season. I'm already working on 2008.
I don't think it's fair to include situations where one team has a clear advantage. Casey and Cummings played almost every minute, so of course they're right around the team's average (-3). This stat is more worthwhile when it comes to the guys who aren't on the field every minute. It doesn't mean that Diz and Peterson are better players. Diz only played 77 minutes, you can't take anything from that. But Peterson played over 1,000 minutes, so I think you can look at that and be impressed.
Surprisingly when you factor in non-even goals Cummings stays right at the team average of +4, along with Diz, but Casey is still at -1 and Peterson goes to +9!
Sebastian LeToux (1300 minutes, .554) is apparently better than: Stu Holden (2020 minutes, .490) Landon Donovan (1889, .382) Kyle Beckerman (1928 .327) Robbie Findlay (1559 .231) Dwayne DeRosario (2390, -.339) How is this a useful stat again?
Why not filter out breakaway goals too? C'mon. There is a fundamental difference between 11v10 and PKs.
An Lenhart is better than all of them. It's primarily useful in comparing teammates, rather than between teams, more as a relative measure than absolute, and loses some relevance if you play the overwhelming majority of the available minutes.
I haven't done all the math on it yet, but PK takers this season scored over 75% of the time (49/63 in the regular season, to be consistent with how scaryice kept his records). I would bet (but again, haven't done the math) that when a team gets a man advantage, the percentage of times that team capitalizes on that situation is less than or equal to 75%. So yeah, the man advantage is different, but if we're just ruling things out on clear advantages, PKs shouldn't count either. And unless they changed this rule after the lockout, the NHL doesn't count penalty shots in its +/- stats for this same reason. Speaking of which, this statistic has much more value in hockey and even somewhat in basketball because of the constant substitutions. With the limited number of substitutions, this stat penalizes starters far too much and rewards players just for standing on the pitch in garbage time of games (ie., Diz's stats in particular). And I'm not saying this as a guy who is anti-stats: I tried keeping track of the same thing last year for the Rapids until I got disillusioned with the results.
There's a difference between PKs and hockey penalty shots too; PKs are just DFKs inside the box. It restarts play, and rebounds are fair game. There is a certain logic to excluding "power play" goals. You may not agree with the decision, but the reasoning is pretty sound, IMO. Correct. This is not nearly as useful in soccer than in sports with free substitutions. But it does have some marginal value. But even in hockey it has issues; it's a metric tied largely to the quality of the team that tries to quantify what is ultimately a qualitative assessment: of what value is Player X to the team's success? Obviously, virtually every player on Columbus has a positive +/- number, but it would be ridiculous to use that to say that Jed Zayner is better than Juan Pablo Angel. Adjusted numbers (like at the bottom of the blog post) are better for that, but they still don't tell the entire story. After all, I don't think anybody believes that Justin Braun is a better player than GBS. Or maybe even that Braun had a bigger positive impact on his team's performance than GBS. I think the only really useful cross-team comparison that can be made is that Braun stood out for Chivas, and GBS did not stand out for Columbus.
If you're going to adjust +/- for the team's aggregate goal diff, you might also want to adjust the value of each player's individual points to account for the expected goal diff in each game For example, for each player, for each game: +/- = GF - GA - (Expected GD between teams)*(fraction of game played) And the difference between a player's +/- using this method (expected goal diff), and the adjustment method you posted before should be a good way to tell who steps up for big games.
I thought about doing something like that, and also about playing time in home vs away games, but it's enough work as it is. It's very simple to adjust for the team GD rather than to go through every single game like that, again. I'm more interested in looking at the other years first.
Thanks for putting this together. I think +/- is probably a more useful stat than many that are kept. Oviously the small sample of goals will introduce some error due to noise. But in general you get some pretty good data. The overall is not as important as the comparison within a team. This stat captures all the little things and dirty work that help a team win but are not obvious unless you breakdown film and review everything each player does. That run pulls apart a defense, the pick that frees a teammate for an open header, the tracking back to keep shape when a teammate is drawn out of position etc. etc. A lot of these things are sexy or exciting, but are critical to a team success. For example watching the Galaxy I was not surprised to see that Miglioranzi was so much better than Dema as I thought his absence hurt the Galaxy in the final. I was also not surprised to see Magee so low and AJ so high. About the only big surprise to me was that Klien was so low. I also bet Beckham was nowhere near as good last year as this year.
You could also do a simplified form of +/-, where you just measure + vs expected team + for offensive players and - vs expected - for defensive. The main problem is subs. Coaches will sub tactically (probably why Le Toux had such a good performance-- if we're winning, we bring on Ianni in place of a mid, but drawn or losing we bring on Le Toux == more attacking chances per minute played vs less pressure). Between infrequent subs and infrequent goals, there's a lot of outlier noise.
It's useful because it adds relevant knowledge. Anyone trying to find a single stat that is The Answer is always going to be disappointed.
Still working on the 2008 data. Should have it completed sometime in the next few days. In the meantime, feel free to make predictions. Who will be an outlier for their team?
Wow. I would not have expected those numbers for Brad Evans. I definitely would have expected them for Padula. I think Donovan's numbers are a function of being on a horrible, horrible team. You'll note that his team-adjusted numbers aren't quite as shockingly bad.