You have put your finger on an important and fundamental issue - a completely objective system can never be completely fair, and conversely a completely fair system can never be completely objective. For my current purposes, being objective is more important than being fair. The question is, more often than not, is a clearance over the endline good or bad? If it is bad, say, 90% of the time, then to be completely objective I decide that 100% of clearances over the endline are unsuccessful, and have to accept that 10% of the time that decision is unfair (i.e., a subjectively good clearance is objectively rated as unsuccessful).
Agreed. Positive touches per 90 min provides some useful information, but you need to be able to weight different types of touches properly to get a truly meaningful rating.
Yes, different types of touches in different situations, decision-making, moves and marking and so forth away from the ball, etc. The hard part is that much of that is difficult to objectively rate. I don't blame you for trying, though.
This formula is absolutely unnecessary and suspect to say the least... It reminds me of the movie "A Beautiful Mind" when Russell Crowe is watching the group of pidgeons and charting their walking habits to try and explain it mathematically. Sometimes you just have to watch the game, enjoy it, and then argue with your friends about who sucks and who doesn't. I don't need a formula to be good at that.
Well what sort of chef brings the ingredients out on a tray and asks his guests to sample them and provide their thoughts, then complains that we must wait on the cake after we do? Honestly, without being too much of a pest, I don't understand why we can't see how you reached a 38 rating for Robinson. Your just not addressing the issue -- and you did invite us to raise issues .....
I think I have already addressed this issue. The provisional ratings weight all touches equally, so Robinson's goal counts no more in the ratings than Twellman's initial touch to kick off the match. The final ratings will weight touches according to their importance in scoring and conceding goals, so that Robinson's touch will count more. Since all four USA centerbacks had a similar number of touches in their 45 minutes of play, the small differences in their ratings are primarily due to differences in the quality of their touches, i.e. Conrad had the highest percentage of successful touches, followed by Parkhurst, Robinson, and Goodson.
Obviously, I see the same touches you do. I see two main differences in interpretation - if a player "wins" a header and by doing so concedes possession to the opposition, I consider that an unsuccessful touch. Similarly, if a player commits a foul, I consider that an unsuccessful touch.
Yeah, I was suspecting that. On headers, you have Sweden pumping long balls into the middle of the American defensive half trying to find Wernbloom. Robinson is eating Wernbloom alive, winning everything. Of course, these are passes to Wernbloom. So some of Robinson's headers don't always go to US players -- he is breaking up an attack. When Conrad and other US defenders are on Wernbloom -- the Swede wins the ball much more -- outplaying them more often. Conrad doesn't even touch the ball. So despite the fact that he is playing worse his #s from your system are better. Every touch is equal? Criteria that does not assess context? I think your right. You have nothing but meaningless raw ingredients here. They are not only meaningless in all applications and contexts, they are often misleading. I'm starting to think any cake made with these ingredients will be foul tasting if not poisonous. I do appreciate your doing this. As I have said, I have often wanted to try to comprehensively analyze games and player/team performance. Any early attempts at something are going to be rough. I respect your attempt. I am trying to provide you my honest thoughts out of respect for this sort of work.
As I've said many times in the past - a system like Opta has a weighed formula, which tracks ~ 25 various events. A simple short lateral/backward pass may rate 1 point. A forward pass may rate 2, a completed long one a 3. A shot on goal (save) or a cross may rate 3 as well. A tackle, regardless of whether the ball is actually touched or just steered to safety, or a won header may rate 2-3. A dribble and run may earn a 4. An assist will rate a 5. A won PK will rate a 10. A goal will rate a 15 ... and so on. This rewards both the central mids with a lot of touches, the defenders with a lot of tackles/clearances and goal scorers with more meaningful plays. Now, at the very worst, when you compare players per position, you can say, "A defender has a worst passing percentage but more tackles" or a "forward has several shots on goal but no assists", etc ...