I tracked every touch of the ball for all USA players in the recent match versus SWE. Each touch was classified by type (pass, shot, etc) and as either successful or unsuccessful according to pre-defined, objective criteria. Based on the data for this single match, I have calculated a simplistic provisional rating for each player that is a function of both the quantity and quality of touches per 90 minutes. This simplistic rating weights all touches equally. Obviously, some types of touches are more valuable than others (e.g., one successful save is more valuable than one successful tackle). Once I have collected sufficient data (4-5 matches worth), I will re-calculate final ratings with the proper weights for each touch type to be determined by the data. Code: rating player 122 Donovan 119 Moor 90 Corrales 88 Altidore 85 Davis 62 Edu 61 White 55 Kljestan 55 Clark 45 Noonan 43 Conrad 41 Twellman 38 Parkhurst 38 Robinson 34 Guzan 32 Rolfe 20 Goodson Given the limitations of the provisional rating discussed above, comparisons between players are probably only useful for players who play the same position. Hence, the forwards rated from best to worst are Altidore, Noonan, Twellman, and Rolfe. For outside mids, Donovan is rated significantly higher than Davis, White, and Kljestan, while for center mids, Edu is rated only slightly higher than Clark. For outside backs Moor is rated higher than Corrales, while the centerbacks rated from best to worst are Conrad, Parkhurst, Robinson, and Goodson. The final (weighted) ratings will enable all players to be rated on a common scale. Thoughts?
Interesting. I've thought of doing this as well but never found the time. I've watched that game three times now and your results surprise me. I think for others to really understand this and provide meaningful comment you need to publish two things: 1 - Your complete rating system with the "predefined objective criteria" and 2 - Your play by play log with minute index so we can review some of your observations and see how your doing it.
Specifically, which ratings surprise you? I have put a lot of thought into the criteria, and a lot of work into collecting the data, so am not yet willing to share all of the details, at least until I have calculated a final set of ratings and see how well they work. However, if people have specific arguments for why the ratings for one player "should" be higher than another player, I would consider citing some specifics to support the ratings.
The high ratings for Moor and Corrales. They both spent lots of time on the ball and were rarely tested, as Sweden's entire attacking scheme through most of the match was to simply try long balls to their forwards. Neither hardly put a foot wrong, and they put forth the work effort moving forward and getting back. But they were frankly crap, regardless.
Why do you think they were crap? Of the 17 players who saw action, Moor and Corrales ranked 4th and 5th, respectively, in terms of quality of touches.
While there may be some merit in trying to quantify performance with a set of objective criteria, great care has to be taken in order to set that criteria to accurately reflect performance. Case in point: Goodson had a very good game. If the criteria slants to favor those who played more minutes then it should be adjusted.
The ratings are calculated based on touches per 90 minutes, so the criteria do not favor players who played more or less minutes. That said, Goodson is a very special case. In part, his rating is (justifiably) low because he had the lowest number of touches per 90 minutes of any USA player. However, in part, his rating may be unfairly low due to the thorny problem of how to treat defensive clearances. Of all the criteria, this is the only one where even I have serious concerns about how accurately the criterion reflects performance. It so happens that Goodson had a relatively large number of (headed) clearances in this match. The question is, are these clearances successful or unsuccessful touches? If you think the answer is obviously "successful", then my question for you is whether you also consider Onyewu's clearance that Rosicky rocketed into the net for the Czech Republic in the 2006 World Cup a "successful" touch? Not so simple, is it? For current purposes, I define a clearance as successful if the clearing team retains possession after the clearance, unsuccessful if they don't. Most of Goodson's clearances were of the latter (unsuccessful) variety, which contributes to his lower rating. Now, if someone has an alternative, but still completely objective criterion, for determining a successful vs unsuccessful clearance, I would be willing to consider changing my definition. One option I have considered is to define all clearances from inside the defending team's penalty area to be "successful", but then to have different success criteria for a touch that is one yard inside the area vs one that is one yard outside the area seems to me somewhat arbitrary and unsatisfying.
A similar venture that sparked a great deal of discussion has been done in the recent past, though the name of the poster escapes me. Illinoi*? Anyway, without publishing your full criteria I am not sure that what you have posted has any more value to us than me assigning numbers from 1-10 based on my review. I too was very surprised at how high Drew Moor was given his atrocious crossing during the match. As for your question on clearances, I would argue that a clearance that leads directly to an opportunity on goal, such as Gooch's is not really a clearance. There is certainly a subjective judgement involved, but many statistics in other sports do as well (for instance errors in baseball).
a thing for ranking defenders is, honestly, touches don't always matter. Was goodson in good position? meaning, if he's in the right place, the midfielder plays the ball elsewhere. Or goodson's positioning and/or shielding of the offensive player means the keeper or teammates makes a play on the ball. Or a defender can be in position such that a fastbreak is halted and the ball played backwards, all without him making a play on the ball. you can play great defense without touching the ball ... thus it's hard to quantify in a system like this. Sometimes just cutting off an offensive player's run is the most important thing a defender can do. also, not saying goodson was all that great, just that his name was mentioned previously. that said, I'm interested to see where you go with this and what the rankings look like with more data.
In this sport, I think opportunity cost is crucial. A player can have a "successful" string of touches and a pass, but he may have missed a great chance to do something much more special, a chance he maybe never even saw was there, a chance that existed and might have been seen and taken by a cleverer player, and would never show up in stats. I think a mediocre player who consistently plays with competence but lacks vision could be overinflated badly by rankings of this type, which makes them dangerous. I don't really see that happening here (except for Edu ranking over Clark, which I feel is a clear example), but it's something I worry about. So I'd ask, in the ratings for "touch type", are categories laid out for genuinely "special" touches, such as the last few before a goal or a shot?
I think this is more of a concern for the provisional ratings, but once the final weighted ratings are available, those players who consistently make important touches will be rated higher. (I agree with you, by the way - my subjective impression is that while both played well, Clark was better - it will be interesting to see if that is borne out in the final ratings.) No, because philosophically I am not convinced that the last few touches are necessarily more important than earlier ones. If you go back to McBride's goal against Portugal in the 2002 World Cup, Donovan made an almost unbelievable trap of a severely overcooked Mastroeni pass, without which the subsequent Sanneh feed and McBride goal wouldn't have happened. My approach is to collect the data for all touches and let the analysis results tell me which are important and which are not.
a good point. somebody did something similiar to this a few years ago and Convey rated higher than people thought. it was because he kept passing the ball backwards and playing scared. so yeah, he kept possession and completed a high percentge of his passes but he never did anything to affect the outcome of the game. 3 dangerous passes and 15 bad ones could often be better than 13 safe passes. not always ... but it's the mark twain thing. three kinds of lies: lies, damn lies and statistics.
I'm rewatching the game again right now. In minute 40, Donovan makes a hash of a manageable pass from Twellman losing possession for a second. Clark however is present and immediately digs the ball back with a deft one-touch pass, moves, receives the ball back and quickly puts Noonan in on the right to create a good attacking chance. Yet Donovan is 122, and Clark is 55. I just don't get it. I want to see how you rate plays like that. Clark covers a ton of space all over the field. How do you handle that versus a less dynamic player? The Gooch clearance and Rosicky goal were mentioned as a negative for Gooch. Was it really a bad clearance? Or did Reyna lose his mark on Rosicky. My memory is that the clearance was not horrible -- but the marking sure was. Reyna got caught ball watching. How do you handle that? Who is right? What are your objective criteria that decides that question? Your asking for comments on the results -- but unwilling to reveal the methods that produce the results. Robinson wins every ball near him, scores a goal, and comes out a 38? Huh? So here is a log of Robinsons half, limited to his time on camera. I have no idea how you get a relatively negative rating for Robinson out of this. After completing the log though -- I'm convinced your objective criteria need a ton of work .... ------ Eddie Robinson's Swedish Adventure, 1-19-2008, Home Despot Center ----- 1:53: Challenges strong for a direct air ball into the defense -- takes an elbow and draws a foul -- POSITIVE. 3:00: Receives back header Twellman, onne-touch pass on to Corrales -- maintain possession -- POSITIVE. 3:10: Cuts out direct ball with a perfect header to Corrales winning posession -- VERY POSITIVE. 3:35: Takes restart in own half. Plays ball into Edu's feet, then receives it back and switches quickly to Conrad's feet -- POSITIVE. 4:57: Wins head challenge directing the ball away to Twellman -- POSITIVE. 5:12: Shephards too deep Sweden pass over the top safely over the endline. If it had been better he had it covered. -- POSITIVE. 6:00: Sweden breaks out of midfield after Conrad loses his mark. Middle is well covered by perfectly positioned Robinson. -- POSITIVE. 7:20: US free kick into the box results in foul call against US. Robinson is in the mixer and could have been the culprit. UNCLEAR. 8:29: Fouls his mark from behind at midfield, stopping counterattack opportunity. -- POSITIVE (Sweden had transition opportunity). 11:40: Loses second ball battle on deep goal kick from Sweden -- NEGATIVE. 11:44: Recovered into good position to block run by Sweden ST into the box, easy collection by Guzon -- POSITIVE. 12:05: He and Conrad trap Sweden ST offside on ball over top -- POSITIVE. 12:27: Receives backpass from Clark, one-touches to Corrales -- POSITIVE. 13:00: Great positioning shields Sweden ST from getting to longball out of back at midfield. Easy collection for Corrales -- POSITIVE. (Robinson missed his header -- but more importantly he focused on his mark and kept him completely away from the ball and it was an easy win for the US. This is the sort of professional play Robinson makes all day long. His touch was poor -- his play was brilliant.) 14:08: US Corner. Wins second ball at the 6' and roofs a shot over two defenders and a well positioned GK on his feet. GOAL. VERY WELL TAKEN. VERY VERY POSITIVE. 17:15: Sweden plays pass toward Robinson mark trying to break out of midfield. Pass is off-target, his man is covered. Easy collection for Guzan. POSITIVE. 18:59: Pass attempt behind the D from midfield to Robinson's man. Cut-out with one-touch out of the air pass to Moor's feet. VERY POSITIVE. 20:26: Poor camera work makes this unclear, but it appears Robinson challenges in the air for a deep goal kick which is not won cleanly, Robinson emerges from melee to touch ball back to Corrales -- POSITIVE 20:50: Air ball from Sweden back into the center circle . Robinson wins it against two Swedish challengers, but Sweden wins second ball -- POSITIVE 21:10: His mark receives ball at the top of the box, left side, then lays it off towards the sideline. That player plays it in to a runner that Clark tackles. Robinson almost caught ballwatching ... ball came in through his general area -- SLIGHT SLIGHT NEGATIVE. 22:42: Perfectly positioned he well clears to Noonan a Sweden cross from the left. -- VERY POSITIVE. 25:12: Cleanly wins airball challenge at halfline, had to come 15 yards -- POSITIVE. 25:45: Sweden free kick to the back post. Robinson perfectly positioned to cover gets fouled from behind. VERY POSITIVE. 26:28: Robinson attempts to cutout through pass out of the air but only knocks it down, giving Sweden chance to win 2nd ball (they wind up fouling Moor). NEGATIVE. 28:25: Robinson's mark shows at halfline, gets ball, Robinson shadows while ball laid off. NUETRAL. 29:03: Robinson's mark shows for ball at defensive third, Robinson fouls him from behind. No need to foul. NEGATIVE. 29:25: Sweden free kick. Near post ball. Robinson has his man covered on far post. NUETRAL. 32:20: Robinson challenges for deep goal kick in defensive half. Coming from behind wins the ball but fouls Swedish player with forearm in the back of the head after winning the ball -- not malicious but reckless. NEGATIVE. 32:48: Sweden free kick to back post. Conrad is marking Swedish (Wernbloom), Robinson has space, middle channel, Conrad lets Wernbloom go, Robinson does not pick up, Wernbloom gets second ball opportunity on near post that Guzon saves. Unclear whether Conrad was at fault for letting Wernbloom go or Robinson for not picking up. UNCLEAR NEGATIVE. 35:36: Robinson, providing defensive cover on the left side, covers a loose ball that falls to a Sweden player in the box, positioned to contain while Clark, Davis and Corrales collapse on the play. POSITIVE. 36:12: Restart with ball to Corrales' feet. NUETRAL 36:26: Covers Sweden direct ball out of the back with perfect header to Edu's feet. POSITIVE. 38:58: Restarts ball in his half to Clark's feet. NUETRAL. 40:03: Robinson's mark shows at midfield for ball to his feet. Robinson beats him to it and clears into touch. VERY POSITIVE. 42:00: Sweden long ball down left -- Robinson has it well covered. POSITIVE. 43:32: Sweden corner. Top of the six toward Robinson's mark. Robinson challenges and prevents anything clean. Ball bounces back out to the left. POSITIVE 43:39: Sweden cross in from left towards Robinson's mark. Robinson does some wierd foot high spinning challenge that was not impressive. Ball over the goal off Sweden. NUETRAL. 44:37: Robinson clears loose ball away from top of penalty box. POSITIVE. 45:07: Sweden corner into top of the six. Crazy melee around the ball. Robinson is in the middle of it... Balls squibs out towards goal where Davis clears. UNCLEAR.
I think a defensive play is different from receiving a pass. If a touch comes from a pass from your team mate, your chances of being able to do something positive are much better than a headed clearance for example. Another example is a defender in the midfield who disrupts a passing sequence with a slide tackle or stab that does not lead to gaining possesion, but does disrupt what could have been a good attack in some instances and give the defending team time to recover. I assume at the points for touches should reflect the possesion arrow. Therefore, with possession a good touch is to keep possesion for the team. A bad touch is to give up possesion. If the team does not have possesion, a good touch is one that leads to regaining possesion. However, a touch that does not result in regaining possesion is NOT a bad touch like the one above. Defensive touches have value based on what they cost the opposing team. Here are some ideas ranked from best to worst defensive touches. Rank-----Cost to opposing team 1-----Stops possesing team from scoring 2-----Steals possesion and scores 3-----Disrupts possesing team in a scoring/dangerous position 4-----Steals possesion and get into a scoring/dangerous position 5.----Steals Possesion 6-----Disrupts passing sequence in own half 7-----Disrupts passing sequence in opponents half In my opnion all of these are higher ranked than receiving a pass from your team and coughing up the ball.
I went back and re-tracked all clearances, defining a clearance forward or to the side out of the penalty area as successful (clearances over the end line conceding a corner are still defined as unsuccessful). With the new definition, the central defenders' ratings become 45 for Conrad, 38 for Parkhurst and Robinson, and 30 for Goodson. So not much difference - Goodson's rating improves relative to the others, but is still the lowest of the four.
As I mentioned in my original post, it probably doesn't make sense to compare provisional ratings for players who play different positions (outside vs central mid). The large magnitudes of the outside mids' and outside defenders' provisional ratings are a direct consequence of Bob Bradley's offensive system - those players see many more offensive touches than the central players. The final ratings, which will be calculated in terms of effective goals for and against, rather than touches, will be far less sensitive to stylistic choices, I believe. Please be patient.
That is an interesting idea. Effectively, you are proposing a system where a touch can be successful, unsuccessful, or neutral. I think I will start with the simplest assumption of only two choices, successful or unsuccessful, and see where that leads, but this could be a potential refinement in the future.
So, as I understand this, if a defender saves the game by clearing a sure goal off the line, over the endline for a corner, it is an unsuccesful touch?
The system produced a relatively low score for the best player on the field during the first half. Robinson. No one contributed more to their team's success. He covered extremely well. He scored a great goal that most other players would not. He shut down Sweden for huge periods in the middle of their attack. Sweden won nothing near him until late in the half. I don't see how more time and statistics will address that problem. I sincerely, respectfull believe it would be very helpful if you went through my log for Robinson and indicated how you scored each of those instances so we can see how you came up with 38. I do like the idea of goals for and against while on the field -- lick Hocket does. I do think that is a meaningful statistic in soccer.
OK. So a cross to the far-post is dangerous. Keeper is stranded at the near post --two attackers are there with one defender who holds one attacker off and out jumps the other to nod the ball over the endline. That is an unsuccesful touch?
Imagine that I have assembled all of the ingredients for baking a cake. You are a bit like a little kid, sticking your fingers in the flour and baking soda and complaining that they don't taste very good! Touches are like the ingredients - they are the inputs to the process that yields the tasty (goals for) or not so tasty (goals against) outputs. In order to understand how different ingredients combine to make cakes, you have to bake a few different cakes with different amounts of ingredients and see how tasty the results are. That's why more matches and data will address the problem.