Greatest Single-Season Goalkeeping Performances

[kenntomasch]

Goals Against Average. Goals Allowed x 90 /minutes played

[tachyon1]

Cheers K.

& apologies to TH,I've stuck the assessment of the 4 from 5 keeper into his figures.

It's actually a zero % chance that Howard is a 50% stopper,
a 0.02% chance he's a 60% 'er
a 15.4% chance he's a 70%'er
a 84.5% chance he's a 80%'er
& a 0.08% chance he's a 90%'er

thought the figures looked odd.

T

[mpruitt]

Cheers K.

& apologies to TH,I've stuck the assessment of the 4 from 5 keeper into his figures.

It's actually a zero % chance that Howard is a 50% stopper,
a 0.02% chance he's a 60% 'er
a 15.4% chance he's a 70%'er
a 84.5% chance he's a 80%'er
& a 0.08% chance he's a 90%'er

thought the figures looked odd.

T
Nice numbers. That seems to be a much more sophisticated way of trying to calculate some kind of Replacement Keeper Value.

[ChrisE]

ChrisE, that forumla really provides an interesting justiposition to the oen that you and beinke came up with. I wonder what you think of its merits vs. yours?

To be honest, Maxim, I think that they're really quite similar. In their formula, they multiply average league shots by the ratio of league shooting percentage to the opponents' shooting percentage.

The league shots average doesn't affect these ratios, it just converts things from a save percentage to an expected goals - the proportions should be the same. As for adjusting for opponent shot percentage, I'm just not convinced it's necessary - I suspect that league schedules are similar enough, and parity is strong enough, that no goalie faces average opponents that much stronger than any other. However, it would make things slightly more accurate, I suspect - unfortunately, we don't have the old lineups and stats for individual games, so calculating these things would involve a lot of guesswork.

I suppose that using the goals allowed to figure an expected win percentage is better than just using raw GAA, but I don't think GAA is a particularly useful stat for evaluating goalkeepers in the first place (it's just above shutouts), so I'd rather just avoid it.

As for NoSix's goalkeeper stat, I think it's a lot like the QB rating stat, or various fantasy basketball scoring systems. It almost certainly does a better job of evaluating goalkeeper ability than simply looking at shooting percentage (or GAA, or win/loss), but I'm not sure it has a lot of potential - it's a good stopgap, but I would hope we could develop something that weights the various aspects of GK performance according to their significance (e.g., if a C/P turns out to be significantly less important than a save, NoSix's rating system wouldn't take this into account [although maybe it could]).

[tachyon1]

Using the GAA stat, I've compiled a list of the best performances as far as the difference between the leading goalkeeper and the second-best goalkeeper, the theory being that a 0.80 GAA doesn't mean much if the runner-up came in at 0.89--obviously, both benefitted from a low-scoring year. However, someone coming in at less than a goal a match while the runner up let in 1.30 a game obviously had a great year, going above-and-beyond the "average" generated by league scoring as a whole.

Whaddya think?

Putting aside for a minute the debate about the best stat to use,if you want to try to compare ratings from different seasons you could do worse than trying to convert them to standard scores.

Take two different seasons with six keepers;

Season one has the keepers rated at 0.8(whatever),0.9,1.4,1.8,1.9 & 2.0. I'm assuming that a low score means a better keeper.The average is 1.466 with a standard dev of 0.475.

Season two has keepers rated 0.8,0.8,1.0,1.0,1.2 & 1.4.This time the average is 1.033 & sd is 0.213.

To compare the 0.8 score in season one with the 0.8 score in season two convert them both to standard scores.

Season one's 0.8 comes out as minus 1.4 while season two's 0.8 equals minus 1.09.

Therefore season one's 0.8 is better than season two's 0.8,because it's standard score is a larger negative number & in this case low ratings equals good keeper,high ratings equals bad keeper.

(You could push it a bit more & say that Season two's 0.8 is equivalent to a score of 0.95 in season one,(by converting its standard score from season two to a rating in season one) & therefore the second best score in season one(0.9) is also a superior score.)

T.

[AussieVamp2]

(e.g., if a C/P turns out to be significantly less important than a save, NoSix's rating system wouldn't take this into account [although maybe it could]).

should be able to look statistically at what the weightings for each might be in his formula, once that data is collected?

[Jabinho]

Lev Yashin ...Moscow Dynamo ...1963.. European Footballer of the Year!
Best season for a keeper in the history of the solar system.. :eek:

[tachyon1]

Just a few (probably self evident)ideas,

if you're going to make an adjustment to goals conceded/on target shots for individual keepers you've got to look at the quality of the shots he faced.

There seems to be pretty good evidence that better teams produce "on target" shots that are harder to save.

From the 2001-2002 EPL there were 3823 shots on target & they produced 997 goals,a strike rate of around 26%.

Actual figures for goals scored per on target shot for each EPL team are significantly different from the figures you would expect if this 26% figure applied to each team.

Also there's a reasonably strong inverse corelation between finishing position and percentage of on target shots converted.

If you started off facing Arsenal,Man U,Lpool & Newcastle,you faced teams who converted 32% of their on target shots.By contrast Derby,Charlton,Sunderland & Ipswich only converted 19% of theirs.

In short,keeper ratings should probably be adjusted for shot quality.

T.

[NoSix]

In short,keeper ratings should probably be adjusted for shot quality.

But "in long", I still doubt it is worth the effort. If you want to compare keepers over a half a season, you probably need to adjust for shot quality. Over the length of a career, even if you play for ManU the whole time, you are going to have enough games against the Arsenal's and Chelsea's that the residual strength of team effects are fairly small.

Speaking of shot quality, my theory (no definitive data to back it up, unfortunately) is that the biggest determinant of shot quality is shot location. If you adjusted for both the distance and the angle of the point from which shots were taken from, say, the point on the end line at the center of the target goal, my guess is that those adjustments would account for more than 90% of shot quality.

[mpruitt]

But "in long", I still doubt it is worth the effort. If you want to compare keepers over a half a season, you probably need to adjust for shot quality. Over the length of a career, even if you play for ManU the whole time, you are going to have enough games against the Arsenal's and Chelsea's that the residual strength of team effects are fairly small.

Speaking of shot quality, my theory (no definitive data to back it up, unfortunately) is that the biggest determinant of shot quality is shot location. If you adjusted for both the distance and the angle of the point from which shots were taken from, say, the point on the end line at the center of the target goal, my guess is that those adjustments would account for more than 90% of shot quality.
Especially in a league with only 10 teams like MLS it's probably not worth the effort. With an unbalanced schedual, which they have maybe it might be worth a go but there's obviously so much parity that it probably doesn't make a wink of difference. In regards to your other theory, that's something that of course we'd like to have as a counting statistic but probably isn't going to happen any time soon unfortunately. Even if they did or could do it however, I imagine the determination between what quaderant on the field vs. another might be nominal. Maybe if it was in thirds or quarters. Though, when you said that it remidned me of the fact that you could alwyas scores sick goals from the side of the pk box in Fifa 94 on Sega Genisis.

[NoSix]

Though, when you said that it remidned me of the fact that you could alwyas scores sick goals from the side of the pk box in Fifa 94 on Sega Genisis.

Well, that settles it. I'll get to work on a new theory immediately... :p

[tachyon1]

But "in long", I still doubt it is worth the effort. .

Hi NS,
parity in the US probably makes it less of an issue in that league.
However I don't know if you're always going to eliminate seasons where some teams stand out.I haven't watched enough mls to make a subjective judgement.

It might be worthwhile to run a goodness of fit type stat on the goals/shots on target numbers to see if the strike rate varies for the mls teams.

It's probably an issue in most other leagues though.I haven't got the figures to hand but iirc the actual EPL figures for the season I looked at would have arisen by chance once every 200 seasons.So there's a strong case for different teams having different shot quality.

Also in England the correlation between current league position and final league position(when most strength of schedule issues should naturally even themselves out) only starts to strengthen when teams have played a dozen or so games.

So teams do get easy or hard rides,short term in a non parity league structure.

On another issue,if you're going to look at a keepers record I think it's vital you preserve the number of shots he's faced.I've just scanned the hockey link posted earlier & they seem to be going down a "convert to average shots per game" route.

Neat feedback.

T.

[NoSix]

Hi NS,
parity in the US probably makes it less of an issue in that league.
However I don't know if you're always going to eliminate seasons where some teams stand out.I haven't watched enough mls to make a subjective judgement.

It might be worthwhile to run a goodness of fit type stat on the goals/shots on target numbers to see if the strike rate varies for the mls teams.

It's probably an issue in most other leagues though.I haven't got the figures to hand but iirc the actual EPL figures for the season I looked at would have arisen by chance once every 200 seasons.So there's a strong case for different teams having different shot quality.

Also in England the correlation between current league position and final league position(when most strength of schedule issues should naturally even themselves out) only starts to strengthen when teams have played a dozen or so games.

So teams do get easy or hard rides,short term in a non parity league structure.

T1, your logic is sound and your argument compelling, as usual. Hard to disagree with the idea that adjusting for team strength is better, but I guess I still feel in some situations unadjusted data is good enough. If you look at my all-time MLS keeper list, I think 90% of long-time MLS fans would agree that Friedel, Cannon, and Busch are among the best to have ever played in the league, and Ammann, Oshoniyi, and Salzwedel among the worst. Frankly I was surprised how well such a simple metric seems to do at separating out the wheat from the chaff.

[tachyon1]

Just a prelim'y shot at rating current season 'keepers.

I've taken percentage of on target shots saved as the main measure of keeper excellence.Number of shots faced has been preserved and so every keeper is rated,even if he's only played half a game.

No account has yet been taken of strength of opposition(I'm still collecting data on who's played against who).I'll factor that in later to see if the figures change much.

The current keeper standings from mls.com are listed for comparison.

..................................(mls rating)
1) J Cannon(Colo).....78.4 (1)
2) H Ring(Chic).........77.4 (3)
3) J Busch(Colu).......76.6 (5)
4) T Meola(KC).........74.9 (2)
5) J Cassar(Dall).......73.8 (Not Rated)
6) A Brown(NE).........73.1 (12)
6) S Garlick(Dall).......73.1 (7)
8) P Onstad(SJ)........72.2 (8)
9) T Perkins(DC).......71.9 (4)
10) J Walker(MS)......70.8 (9)
11) M Reis(NE).........70.4 (10)
12) D Popik(LA).........69.8 (NR)
13) J Conway(SJ)......69.7 (NR)
14) D Warren(DC)......69.5 (NR)
15) K Hartman(LA).....69.4 (11)
16) Z Wells(MS)........66.5 (NR)
17) N Rimando(DC).....66.2 (6)

Therefore for example based on Cassar's current season rating of saving 11 of the 13 on target shots that he's faced(an over optimistic 84.6% save rate) & incorporating the historical save average for the mls you expect him to level out at around a 73.8% keeper.

More later.

T.

[NoSix]

The current keeper standings from mls.com are listed for comparison.

..................................(mls rating)
1) J Cannon(Colo).....78.4 (1)
2) H Ring(Chic).........77.4 (3)
3) J Busch(Colu).......76.6 (5)
4) T Meola(KC).........74.9 (2)
5) J Cassar(Dall).......73.8 (Not Rated)
6) A Brown(NE).........73.1 (12)
6) S Garlick(Dall).......73.1 (7)
8) P Onstad(SJ)........72.2 (8)
9) T Perkins(DC).......71.9 (4)
10) J Walker(MS)......70.8 (9)
11) M Reis(NE).........70.4 (10)
12) D Popik(LA).........69.8 (NR)
13) J Conway(SJ)......69.7 (NR)
14) D Warren(DC)......69.5 (NR)
15) K Hartman(LA).....69.4 (11)
16) Z Wells(MS)........66.5 (NR)
17) N Rimando(DC).....66.2 (6)


Here's the list by ranked by goal keeper rating (a weighted average of save and catch/punch percentage):


RANK GAA PLAYER TEAM S90 SV90 CP90 GR
1 1 Cannon COL 5.00 4.18 3.36 0.902
2 2 Meola KC 4.23 3.31 4.85 0.898
3 3 Ring CHI 5.25 4.25 2.75 0.875
4 4 Perkins DC 4.80 3.60 4.40 0.870
5 5 Busch CLB 6.18 4.91 3.00 0.861
6 8 Onstad SJ 6.11 4.56 5.00 0.860
7 10 Reis NE 5.67 4.00 5.83 0.855
8 7 Garlick DAL 5.56 4.22 3.11 0.846
9 12 Brown NE 7.67 5.83 3.83 0.841
10 9 Walker MET 5.70 4.10 3.00 0.816
11 11 Hartman LA 5.33 3.63 3.33 0.803
12 6 Rimando DC 3.14 1.86 1.71 0.735


GAA is rank by Goals Against Average. S90, SV90, and CP90 are Shots Against, Saves, and Catch/Punch per 90 minutes, respectively. GR is Goalkeeper Rating. Note that MLS average all-time GR is about 0.800.

Interesting to compare Perkins vs Rimando and Reis vs Brown.

[Serie Zed]

Couldn't you compare a keeper's save percentage against each team to that team's average goals/shot?

So if Colorado typically scores on 30% of it's shots (on target) in a season, but Hartman stopped 16 of 20 shots (80%) in his two appearances v Colorado he's effectively +2 v Colorado (giving up four goals instead of six).

Then do that for each team in the league and compare the results.

Although that doesn't really separate his performance from his defense's now does it? Darn.

Though you could do that for a keeper and his backup across multiple seasons and see the marginal effect he has on a match.

This is the only quantitative way that really makes sense though -- otherwise you're really measuring the quality of the defense overall.

If you wanted to reveiw a boatload of video, you could do something qualitative and develop criteria for a "quality shot" and a "high quality shot" to see how many of those each keeper saves.

[kenntomasch]

The other thing to consider is that if Kevin Hartman gives up 2 goals against Colorado and his mates only score 1 goal, the Galaxy loses. :)

[numerista]

I think 90% of long-time MLS fans would agree that Friedel, Cannon, and Busch are among the best to have ever played in the league, and Ammann, Oshoniyi, and Salzwedel among the worst.

FWIW, my impression is that there are some questions about Busch (on crosses), and that Ammann was quite well respected. After the Metros decided to bank on Tim Howard, they were still able to trade Amman for Richie Williams.

[tachyon1]

.

OK I've corrected the keepers percentage of shots on target saved to account for the shooting ability of the teams he has faced.Which,as has been suggested doesn't entirely take out the influence of the defence he has in front of him,but it does improve things.

The adjusted rating should/might/could possibly be interpreted as the percentage of shots a keeper would save when faced with a team of average shooting ability over a reasonable length of time.


...............Unadjusted rating.............Adjusted for SOS.
1) J Cannon(Colo).....78.4.....................78.5 (1)
2) H Ring(Chic).........77.4.....................77.1 (2)
3) J Busch(Colu).......76.6.....................77.0 (3)
4) T Meola(KC).........74.9.....................73.4 (6)
5) J Cassar(Dall).......73.8.....................72.4 (7)
6) A Brown(NE).........73.1.....................75.8 (4)
6) S Garlick(Dall).......73.1.....................75.2 (5)
8) P Onstad(SJ)........72.2.....................70.9 (8)
9) T Perkins(DC).......71.9.....................69.9 (12)
10) J Walker(MS)......70.8.....................70.8 (9)
11) M Reis(NE).........70.4.....................69.4 (14)
12) D Popik(LA).........69.8....................70.0 (10=)
13) J Conway(SJ)......69.7....................69.8 (13)
14) D Warren(DC)......69.5....................70.0 (10=)
15) K Hartman(LA).....69.4....................67.9 (17)
16) Z Wells(MS)........66.5....................69.2 (15)
17) N Rimando(DC).....66.2...................68.4 (16)

T

T.

[NoSix]

...............Unadjusted rating.............Adjusted for SOS.
1) J Cannon(Colo).....78.4.....................78.5 (1)
2) H Ring(Chic).........77.4.....................77.1 (2)
3) J Busch(Colu).......76.6.....................77.0 (3)
4) T Meola(KC).........74.9.....................73.4 (6)
5) J Cassar(Dall).......73.8.....................72.4 (7)
6) A Brown(NE).........73.1.....................75.8 (4)
6) S Garlick(Dall).......73.1.....................75.2 (5)
8) P Onstad(SJ)........72.2.....................70.9 (8)
9) T Perkins(DC).......71.9.....................69.9 (12)
10) J Walker(MS)......70.8.....................70.8 (9)
11) M Reis(NE).........70.4.....................69.4 (14)
12) D Popik(LA).........69.8....................70.0 (10=)
13) J Conway(SJ)......69.7....................69.8 (13)
14) D Warren(DC)......69.5....................70.0 (10=)
15) K Hartman(LA).....69.4....................67.9 (17)
16) Z Wells(MS)........66.5....................69.2 (15)
17) N Rimando(DC).....66.2...................68.4 (16)


So the average adjustment is 1.7% and the maximum 4.1%. Being lazy, I'm not sure I would bother with the adjustments, but I applaud your effort!