Saturday, March 01, 2014

Young Players and Stats

This post is kind of a train wreck, stream of consciousness, piece of nonsense.  I'll try to make more sense of it after another tournament.

Last weekend Paideia finished 4-2 at Deep Freeze, losing in the semis to Carolina Friends School.  The two losses were both to the finalists (Holy Family Catholic was the other one) and to teams that will be attending Paideia cup.  Probably the biggest struggle for us was how much we had to play young players.  They were mostly sophomores (we have 9 on the team) and they had to take tough assignments and handle most of the weekend.  They all rose to the challenge, but it was difficult for them.  In particular having to at times get upperclassmen to do the "right" thing was hard.  How do you tell a senior that they need to get you the disc on the swing when you can't drive yet?  I think Tiina would solve this through clearly laying out player expectations, but I don't think Paideia is there yet.

One thing that I focused on this past weekend was figuring out what stats I wanted to track.  I broke my stats down into two different categories: in-game and post-game.  The thought was that some stats I needed to be aware of in the middle of a game in order to manage it properly while some stats I didn't really need until the end of a game for trends.  The rest of this post is going to be about stats, so I don't want to get too far into the weeds here where there are more enticing weeds up ahead.

The statistic that I found had the most post-game utility was Unforced Errors.  I decided to ditch turnovers this season for UE since those are the easiest for us to control.  An Unforced Error was counted every time we had a drop or threw the disc away.  Basically a turn over without the defense touching the disc.  Tracking raw UE was pretty interesting and showed a team that uses all of its players and hasn't had much on-field practice time.

But what I am more interested in is thinking about how to track UE in a useful fashion and how to use it as a metric for the over all "quality" of our team.  Presumably lowering UE would be good for a team, but it doesn't even necessarily mean that the total number of turnovers went down.  One could lower their UE score while still  having the same turnovers by simply having the defense get more Ds.  That would feel doubly inaccurate because not only would we not be "playing better" as the metric might suggest, but we are actually playing worse because we are throwing more contested throws.  So a teams decision making ability affects UE, but I don't know how much that really matters.

What I really wanted to do is find a way to get meaning out of UE across disparate games.  This past weekend we beat Birmingham Forge 11-9 and Catholic High School B 13-0.  How do we compare numbers against such a range of opponents?  Perhaps the easiest way is to track UE per point, so the fact that there were 19 points in the Forge game and 13 points against CHS-B will be divided out.  But that doesn't account for the gap in opponent skill.  Without useful rankings in youth ultimate we can't use "strength" as a balancer because we don't know an opponent's strength.

So I started thinking of different ways to approach getting more meaning out of UE.  I came up with some things that I'm confident Sean Childers will be appalled with.  But I'm not a statistician, I'm a coach, and I would like to think that Bill Barnwell would at least be happy that a coach is trying to find more meaning in a number.

I'm stuck between two things to do to UE to give it more meaning.  They both related to score, as the best proxy for "strength" I can find in youth ultimate.  The first would be to take the delta of the score by the end of the game and use that to modify the number of unforced errors.  I think this would take the form of a divisor so that if you are winning by a large margin the value of an unforced error goes down.  But what about when you are losing?

Which brought me to my next idea.  The value of an unforced error (or a D?) is at least dependent on the current ratio of scores (with opposing score in the numerator).  As an example, if we dropped the disc when the score is tied at 6s that UE would have a value of 1.  If the score was 6-3 (so we are winning) it would have a value of 1/2.  If the score was 3-6 then it would have a value of 2.  I'm not convinced that this number is going to more accurately reflect the value of a turnover, and I know that there are some situations that would break this metric (a punt, for example).  But I'm thinking it might do a better job than just raw unforced errors, and there might be a number that we can use from year to year as a metric of "quality."

NOTE: as a side, this metric would highly value unforced errors when you are losing, which might feel counter to common sense.  It also would skew the data early in the game when the ratios could be further away from 1 with similar point differentials.  I guess the big questions (that we all think we know the answer to) is whether or not an unforced error is more devastating when down 3-2 or when down 13-12.


Anonymous said...

Can you explain why you feel the need to account for the quality of the team that you're playing against when counting unforced errors? To me, an unforced error is exactly that, unforced. It shouldn't really matter who you are playing if the player is already wide open and drops the disc. Unforced errors seem to be a measure of your own team's ability to catch and throw consistently. And that consistency is what seems to be what separates the good from great high school teams.

Eric said...

Amit - While not the author, I will weigh in with my opinion. Against an opponent of greater ability, the throws and plays required are more difficult. Thus, unforced errors are more likely. The same cutters aren't getting as open and maybe the defense is doing a better job taking away strengths. A throwaway to a player with a defender trailing closely behind is different than a throwaway to a player who is wide open with no defender around.