I dont quite understand why people say that DJ has played great this season. Hes made some good saves.. and hes commited his fair share of howlers. His shot stopping on the whole has been average at best. His save percentage reflects that. As for it being due to him being under seige due to a poor Dallas defense... Here are the shots per game and save percentage of DJ Countess and the goalkeepers in the top 10 in save percentage who faced 6 or more shots per game. Name SPG, SV% 2003 Kevin Hartman 6.28 78.3% Tim Howard 7.46 76.3% Jonny Walker 6.40 75.0% Tony Meola 6.08 69.7% DJ Countess 7.33 65.5% 2002 Kevin Hartman 6.00 76.9% Zach Thornton 6.07 75.6% Tim Howard 7.22 71.8% Nick Rimando 6.53 71.6% Tom Presthus 6.86 70.9% Adin Brown 6.37 70.6% 2001 Tim Howard 7.30 76.8% Tom Presthus 7.12 76.4% Nick Rimando 6.20 74.8% Tony Meola 6.53 72.1% Scott Garlick 7.88 67.3% Adin Brown 8.13 65.9% 2000 Joe Cannon 6.96 75.7% Scott Garlick 7.71 74.5% Adin Brown 7.84 70.6% Nick Rimando 6.81 70.6% Mike Ammann 7.36 66.7% Matt Jordan 6.06 66.0% Mark Dougherty 6.80 65.3% I think this shows that Countess's lousy save percentage isnt due to the number of shots he faced. Plenty of players have boasted better numbers than he has facing a similair number of shots. I would also say that Sv% tends to be fairly consistent. There is no random 20% spread that a keeper will fall in, it is at most 6-8%. But draw your own conclusions. Here is the ultimate example I think against the argument that the Dallas defense is solely at fault. Joe Cannon had a 75.7% SV% in 2000, facing 181 shots with a GAA average of 1.49. In the offseason Yallop comes in, the entire team is retooled and improved especially on defense. Joe Cannon faces only 134 shots and lowers his GAA to 1.09. His save percentage that year 75.3%. Food for thought.
Numbers aside the MLS keeper who has impressed me the most is Johnny Walker. He's extremely athletic and makes very intelligent decisions. Brown is good, but seems to be to be a little stiff. I feel that he lacks the necessary athletic ability to become a world class gk.
In my example, Brown had played half a season ... the range was 17.8%. For a full-time starter, the range would be closer to 12%. And, as ChrisE pointed out, that's a 95% interval, cutting off only the bottom 2.5% and the top 2.5%. Now, maybe 12% is still too wide ... but that number comes straight from the laws of mathematics, if shots are independent. To get a more narrow interval, you'd have to assume that shots are inversely correlated. Now, that's only plausible if shooting percentages are based heavily on strategic adjustments, rather than a keeper's shot-stopping ability. That's not a good example at all. Do you remember how the Quakes played in 2000? 90 minutes of bunker defense and long balls to Abdul Thompson Conteh. They even bunkered when they were behind. Any keeper would've had a pretty good save percentage, but he wouldn't have won much. Why not compare Countess to Tom Presthus in 2000? Playing behind several national team defenders, Presthus stopped only 61.3% of opponents' shots (78-127). In 2001, Presthus went to Columbus and stopped 76.3% (136-178). Chance fluctation? Different team strategy? Or maybe Presthus sucked one year and was great the next? (I don't know the answer, but in case it helps, Presthus stopped 68.4% in 1999 and 70.9% in 2002.)
You seem to be using the wrong denominator for Howard ... TH played 13 games, not 15. (Until now, I never checked how many goals the Metros scored in the two games when Grafer was in: four goals in two games.) Anyway, that should read 12.2/14.8 with Howard. This is pretty much what I would've expected. With Howard, they were pressing people forward and looking for a good shot. Now, they're shooting more but scoring less. Less build-up => fewer quality chances.
I truly know nothing about baseball. And keeper is not my speciality in soccer. But yes if I watched every MLS game, I could probably tell you who is a better keeper and why. I appreciate that it's necessary to watch games, but watching games just supports different biases, it doesn't eliminate them. Adin Brown is a very impressive looking keeper, because he's huge, he's intimidating, he commands the area, and he's able to make saves with his size that smaller goalies simply couldn't make. But, statistically, he's not that great looking, and I think that that fact needs to be considered along with how Adin looks during games. [/B][/QUOTE] Absolutely not. Save percentage depends on the quality of shots a player faces as well as what the keeper is doing. It may depend on the style of defense -- for example is the team tries to use an offside trap and crowd the midfield, the other team may not get many shots at the goal -- but when they do beat it, they are going to have lots of space and the shots they get may be very high percentage one. A keeper for that type of team might be excellent, but have a lower save percentage because of the type of shots faced. And the random factor is huge as others have pointed out. And GAA is a laughable statistic from a goalie standpoint and and catches/punches depends on how many crosses and long balls are served up. No it's hard, but the only way to evaluate players, including goalies is to watch the games. That's why not everybody can do Bruce Arena's job.
This is a terrific point, because it can be examined with data. Let's compare each team's save percentage to the number of times it draws its opponent offside. (These include defender blocks as well as keeper saves; I've subtracted out PK's.) NE 128 --72% KC 99 -- 77% LA 72 -- 83% Col 65 -- 77% Chi 63 -- 79% SJ 59 -- 80% DC 53 -- 81% Dal 51 -- 74% Clb 48 -- 79% Met 47 -- 83% Even though there is only a minimal amount of data, the correlation between offsides and save percentage is borderline significant (p-val 0.07). The keepers who seem to buck the trend are Hartman (all-star) and Countess (rookie). It would be interesting to see the numbers from other seasons.
I'm confused about how you got your numbers (I'm not disagreeing with you, really). Was it (Saves+blocks-PKm)/(SOG+blocks-PKa)? I'm getting slightly different results with that formula.
I was just using your number for how many games Howard had played (you didn't expect me to browse through 25 games did you). I guess I should have checked first. I guess your conclusion is one way that you could look at the figures, but it's far from convincing. It looks to me like they're playing pretty close to the same with both goalies. I don't see how you can say they're playing more defensively because Walker is in goal when they have the same shots/game against for both goalies.
Well, if you seriously don't know anything about baseball, I'll try to explain that point a little bit better. I'm sure voros, if he were kind enough to make an appearance, could far out-do me. My point was, basically, that the difference between a .250 and a .280 hitter is a pretty significant difference. .250 is borderline major-league ability, while .280 is a pretty solid hitter. Now there are some things you can tell about a hitter; if a guy is patient at the plate, taking a lot of pitches, while another guy swings at most of the balls thrown at him, you can tell that by watching him (being patient is better). Facts like that would obviously be used in evaluating the player. But the difference between hitting .250 and hitting .280, over the course of a season, figuring 400 at bats, is 12 hits. That comes out to one extra hit per 10 games (or less). It's not reasonable to expect that, watching 10 or 20 games, you're going to be able to tell who is the better hitter, despite the relatively large sample size. The same, I'm saying, goes for goalies. Yes, save percentage depends on a lot of things. But one of them is shot-stopping ability. There's a reason that Kevin Hartman's career save percentage is .744, and Matt Jordan's is .694, and it's not just that the Burn defense hasn't been as good as LA's. Obviously, this isn't a situation as clear-cut as batting averages, but I think it's still important to consider statistics when evaluating a goalie. The random factor is not as huge as beineke made it out to be. Hartman's % in any one year has never been more than .04 off from his career average. Brown's has never been more than .02 off. Joe Cannon's has never been more than .012 off. Yes, some guys have higher variabilities, but it's not like there are massive fluctuations every year. Save percentage obviously has a good deal to do with shot stopping ability. But your points about playing style are pretty good, and I think we should look at those. But, hopefully, they'll yield to some kind of statistical analysis. Catches/punches is, again, a pretty consistent statistic. I'd bet that goalies see about the same amount of crosses and long balls, seeing as, over the course of a season, they play the same teams about the same number of times. Here are some career numbers for C/P's: Howard: (3.00) 4.37, 4.31, 3.87, 4.20, 4.79 Meola: 3.41, 3.92, 3.55, 2.59, 2.64, 2.29, 2.55, 2.78 Hartman: (3.73) 3.01, 2.67, 2.45, 2.68, 2.68, 2.80 Rimando: 3.37, 4.19, 2.96, 2.65 Cannon: 3.17, 3.94, 3.40, 3.75 I included Howard and Hartman's rookie years in parentheses, because I don't really think they're representative. Howard only played 90 minutes, and Hartman's stats are totally different from the other 6 years of his career. I'm not saying you shouldn't evaluate players by watching games, I'm just saying that it's not always going to tell the full story. And it's possible that a player who performs well, but doesn't look as impressive (Ralston, Twellman) might get overlooked for a tools guy (e.g. Albright), and that paying closer attention to statistics would perhaps be a way to remedy this.
Real simple: "shots allowed per game" does not measure whether a team is playing conservatively. Heck, I can't even figure out whether a conservative team is supposed to allow more or less shots. Which is it? But anyway ... the Metros' offensive production has dropped in half. Its midfield scored six goals with Howard; zero with Walker. When you watch a game, you see the midfield taking up very defensive positions. The evidence is clear. So I'm not sure why you think that "shots per game" calls that into question. Also note that I never said that Howard played all 15 games prior to Walker's arrival.
I wasn't even serious about that, man, but you sure implied it by including Grafer's stats with Howard's. If the Metros were going to play more defensively with Walker, you'd expect them to go into a shell with Pants in goal.
I was assuming that more shots per game would be less conservative, but admittedly I haven't really thought it through or looked at any kind of evidence. I looked at the 2002 team numbers for shots and goals, and, to my surprise, it didn't really look like they have much to do with each other. So I'm going to concede this point. I'm not sure I agree with you yet, but I'm not really sure what I'm arguing any more, either. Now would you tell me how you were getting those shot percentages?
I should have said I don't want to know anything about baseball. I respect your apparent affection for that game, but baseball is a game of discrete plays that occur repeatedly making it susceptible to statistical analysis [which is actually the point of baseball, I guess]. In a game like soccer, where there are no discrete events, but continuous actions involving more actors, statistical analysis is much less useful. I agree a tool guy will always get an extra break, but with a good evaluator of talent, that won't last long. It's better than relying on statistics that won't tell you that guy like Twellman may be seriously missing some necessary tools to suceed interntionally. Only your eyes can tell you that. There is no way a player can perform well but not look impressive as a soccer player (as opposed to a male model or track star or something)
I would be trying to compare this to baseball if it weren't goalies we were talking about. The goalie's performance is not, to the degree that a midfielder or forward's performance is, dependent on the team around him. Or at least, it will take a deal of convincing for me to believe that it is. And a player can't perform well and not look impressive in soccer? I'm not sure I can agree with you there either. For the last two years, Taylor Twellman has been, in terms of strike rate, one of the top two strikers in MLS. And yet there a host of people who do not think he's worth anything. Mike Magee, at 19 years old, has six goals. It looked like he was even going to break the rookie scoring record this year (before he cooled off and Damani went crazy). And yet is there anyone who thinks he's looked impressive? No, statistics can't tell you whether or not Twellman is lacking some essential tool that is necessary in the international game, but I don't think watching him play in MLS can tell you that either. I think he needs to be given the opportunity to play (an extended opportunity) before he is written off, but that is another topic entirely. Statistical analysis may be less useful in soccer than it is in baseball. I'd be willing to say that it is very probably less useful. But, I don't know of any attempt to analyze soccer with anything resembling the rigor that sabermetricians have analyzed baseball. Until it's been tried, I'll hold out the hope that soccer can be analyzed a lot more meaningfully than most proponents of 'the beautiful game' would like.
Of course you can hand-pick a couple of players who have only played four seasons and appear to be consistent. When you only have four data points, that kind of thing is bound to happen. But let's look at Hartman: 98 71% 99 79% 00 72% 01 73% 02 77% 03 79% One year, he improved by 8%. The next year he dropped by 7%. Seems like plenty of random variation to me. There are only two keepers who have faced at least 80 shots in 7 different seasons: Scott Garlick and Tony Meola. Garlick's save percentage ranges from 66% to 79%; Meola's ranges from 69% to 80%. You're trying to say that some years Garlick and Meola were disasters, other years they were all-world. Seems to me that they've been pretty much the same keepers all along.
With Grafer, they fell behind 3-0 in one game, so they were forced to push forward (and scored two late goals). In the other game, they won 1-0 against a severely understrength Galaxy team -- no Ruiz, Califf, Vagenas, or Elliott.
Oops, didn't see that first post... TOTAL SAVES - Opp PK MISSES --------------------------------------------- TOTAL SHOTS - Opp PK ATTEMPTS By throwing out PK's, it's possible to eliminate some of the variability in save percentages. Total saves already includes blocks.
What are you trying to prove to me here? I was just saying that using two standard deviations, and so saying that we might expect Adin Brown to range between 60 and 82% is not very realistic. Yes it is possible that that could happen, but throughout the course of most careers you're not going to see that much variation. These numbers don't swing wildly from season to season, it's pretty clear that they are based to a good degree on player ability (I am not denying that defenses play a significant part in any individual season save percentage). With Hartman's stats: yes, obviously, there is a good deal of either random variation and/or defensive contribution. But His numbers are nevertheless very consistent. Is this because he has played for the same team his entire career? Possibly. Is it because he is an inherently more consistent keeper than Tony Meola? I doubt it. Obviously, random variation plays a part in any of these numbers (I'm not saying Jeff Cassar is dramatically improved over the beginnign of the season because he saved 100% of his shots last saturday). However, despite the random variation, it's still clear that Hartman's numbers, and even's Meola's, are varying around a pretty constant point. And so, although it's not the be-all and end-all of goalie performance, an evaluation of their ability should involve looking at their save percentage.
Goalies get credit for saves made by their defenders? Doesn't exactly seem fair. These are the numbers I got, by your formula. 2002: Chi - 84 - 73.8 Cbs - 58 - 75.3 DC - 70 - 71.2 Met - 107 - 71.1 NE - 124 - 66.0 Col. - 83 - 65.6 Dal. - 51 - 69.2 KC - 97 - 72.6 LA - 54 - 74.8 SJ - 101 - 72.1 2001 DC - 95 - 66.4 Met - 89 - 76.5 Mia - 95 - 74.2 NE - 87 - 67.8 Chi - 86 - 76.1 Cbs - 80 - 76.8 Dal - 60 - 66.4 TB - 120 - 62.4 Col - 110 - 70.9 KC - 86 - 69.2 LA - 46 - 70.5 SJ - 77 - 74.8
Please be careful about what I've said (corrected above). I gave a range of 17.4% for Brown in half a season, and in a later post I said 12% for a starting keeper. Garlick's seven-year range is 13% and Meola's is 11%. Thus, my numbers aren't out in left field. But he's also played for a consistently strong team that hasn't employed the offsides trap. Put him behind New England's defense, and it's tough to say what would happen. After all, Matt Reis put up respectable numbers for LA, too.
Thanks for the numbers, although I'm afraid that the PK thing didn't get communicated right. AFAICT, you substracted PK GOALS from the numerator instead of PK MISSES. (The first time I ran the 2003 numbers, I made the same mistake.) But anyway, the results are going to be pretty satisfying. More offsides trap => lower percentage of saves. If you can get me the corrected numbers, I'll do a fuller analysis.
Yeah, that's exactly what I did. Here are the corrected numbers. 75.6 74.7 73.4 71.6 67.0 66.8 71.7 71.3 76.2 74.8 67.8 78.0 74.8 67.8 78.2 77.3 64.9 65.6 73.0 69.7 71.2 76.2
Is he taller than Rimando? I swear there were oompa loompa men at Willy Wonka's that are taller than Nick.
You know what? MLS is making it hard on us ... turns out that in past seasons, they didn't include blocks in the TOTAL SAVES category. So for those years, the right formula is: SHOTS FACED - GA - PK MISSES -------------------------------- SHOTS FACED - PK ATTEMPTS This could have a substantial effect on the results.