New Waldo 10/17

Discussion in 'MLS: News & Analysis' started by alf, Oct 18, 2003.

  1. beineke

    beineke New Member

    Sep 13, 2000
    It's a major stretch to claim Buddle is "better" based on a 0.4% difference when the margin of error is, what, about +/- 6%?

    Anyway, I agree that Twellman is a much better comparison to Razov, in part because he finds almost as many opportunities to shoot (4.23 per 90 vs Razov's 4.43).

    But in general, these numbers are not at all satisfying ... several central defenders have very high shooting percentages -- does this mean they should be moved to forward?
     
  2. voros

    voros Member

    Jun 7, 2002
    Parts Unknown
    Nat'l Team:
    United States
    I'm a bit split on this one. As you'd expect the correlation between shooting percentage and goals per minute on a team by team basis is extremely high (an r around .92) while the correlation between shots per minute and goals per minute is lower (around .60). This would seem to indicate that converting your shots is probably a better way to score more goals than taking extra shots.

    Still, teams don't have a finite number of shots. Just because Razov shoots more than pretty much everybody else, doesn't mean that no on else is getting chances. After all, Damani Ralph is in double digits in goals and DaMarcus now has 7 on the year.

    In baseball one of the things that stat geeks started doing several years back was keeping track of the areas of the field into which batted balls are hit. The biggest use for this has been trying to improve the ways in which we evaluate fielding

    Considering the field markings Soccer has, and the general uniformity (in shape at least) of the field, it might be easier to do this same kind of tracking in soccer. The information you could glean from this both from an individual player standpoint and a league wide basis would probably be unthinkably useful.

    I mention this because I think if we had the above percentages available when the shots are taken from different parts of the field, I think we could learn all sorts of things about Ante Razov (and other players of course) individually as a player, and also information about how the game is being played as a whole. How often do shots taken from the top of the box go in? How often do they go in from 25 yards out? Does a particular keeper get scored upon more often on shots from one side of the field? Do players shoot from outside too often or not enough? Does Taylor Twellman score nothing but tap ins?

    It would seem to me to be useful information both in player evaluation and tactical strategy. Imagine if you knew a particular keeper or team was shaky on shots from distance. You could tailor your gameplan for that week around this weakness, and maybe even get a player into the lineup whose shooting from outside is a strength. Just a thought.
     
  3. beineke

    beineke New Member

    Sep 13, 2000
    Nice post in general, but I don't agree with this inference. Every time you score a goal, your shooting percentage and goals per minute increase by the same proportion (shots per minute increases by a smaller proportion). As a result, even if we had a bunch of forwards who were absolutely identical shooters, shooting percentage and goals per minute would still be highly correlated.

    I ran a simulation where forwards take, on average, 100 shots, and each shot has a 12% chance of scoring. The correlation between shooting percentage and goals per minute was 0.93, even though there was no difference in shooting ability. So if I see a correlation of 0.92, I don't view that as evidence that better finishing helps you -- after all, that's roughly the baseline correlation you get by random chance.

    Incidentally, my simulation assumes that shots are taken with a Poisson distribution. The correlation between shots per minute and goals per minute was 0.36. So a correlation of 0.60 suggests that extra shots take you far above the baseline.

    This is by no means the end of the story, but I hope it suggests that the numbers 0.92 and 0.60 can't be compared side-by-side.
     

Share This Page