APBRmetrics

I was recently speaking with a college athlete whose coach uses TENDEX (or some variation that adds additional contributions such as deflections) to evaluate the players' production. Those unfamiliar can find a description of the metric here

http://wagesofwins.com/2011/02/01/by-re ... n-the-nba/

The player appreciated the use of the metric (as an aggregation as opposed to just listing the typical box score stats). It served as a motivator to work hard in all aspects recorded in the statistics.

On the one hand, TENDEX is probably not the best indicator of a player's performance. On the other hand, it is simple. All players can understand the positive and negative contributions.

There may be a happy medium between accuracy and simplicity of a model for the coach to player communication. And the metric that works best in that context is not likely to be the one most appreciated by an analytics team or general manager. As an analytics community, should we be looking for two sets of player/team evaluation tools?

I am looking for any types of studies or theories pertaining to the use of metrics in coach to player communication specifically.

Thanks.

You might look at the materials reference here.
viewtopic.php?f=2&t=53&p=66&hilit=jose+martinez#p66
I'd think talking to Dr Jose Martinez could be productive.

steveshea wrote:I was recently speaking with a college athlete whose coach uses TENDEX (or some variation that adds additional contributions such as deflections) to evaluate the players' production. Those unfamiliar can find a description of the metric here

http://wagesofwins.com/2011/02/01/by-re ... n-the-nba/

The player appreciated the use of the metric (as an aggregation as opposed to just listing the typical box score stats). It served as a motivator to work hard in all aspects recorded in the statistics.

On the one hand, TENDEX is probably not the best indicator of a player's performance. On the other hand, it is simple. All players can understand the positive and negative contributions.

There may be a happy medium between accuracy and simplicity of a model for the coach to player communication. And the metric that works best in that context is not likely to be the one most appreciated by an analytics team or general manager. As an analytics community, should we be looking for two sets of player/team evaluation tools?

I am looking for any types of studies or theories pertaining to the use of metrics in coach to player communication specifically.

Thanks.

Dick Mays has been posting recently - and he devised his Magic Metric years ago as a way to try to evaluate players, adjusting for level. I believe he just posted some you tube videos on this board - should be one of the more recent posts.

Perhaps a discordant note.

At the abstract level, the game of basketball is really very simple and easily understood (and modeled). All the complexity enters in in application: the design of plays, spontaneous decision-making, blah, blah, blah.

When one speaks of "coach to player" communication, this implies the conflation of two, distinct subject matters: first, the request that players do things better in the context of the coach's specific designs (what may or may not be "optimal") and, second, that players improve their generic skills. In both cases (and certainly in the latter) I am dubious that any "advanced" metrics are of any significant importance. Furthermore, I would wager that the potential value-added, in the team context, of the second category (where very simple statistical constructs should guide a player's efforts and development) is far more important than the first (which would also demand only simple, "non-analytic-originating" metrics).

schtevie wrote:
When one speaks of "coach to player" communication, this implies the conflation of two, distinct subject matters: first, the request that players do things better in the context of the coach's specific designs (what may or may not be "optimal") and, second, that players improve their generic skills.

To expand a bit on what I'm after. A coach might want his players to play more efficiently, to (for example) not take the contested jumper early in the shot clock. If the coach is consistently celebrating player scoring with no regard to how efficiently they produce those points, the stats probably aren't going to be of much help in motivating players to be more efficient. However, if the coach was regularly recapping player efficiency (with some appropriate metric), and making known that efficient players would get more playing time, the stat could be a motivational tool. Is the result-a player that chooses better shots, or generally makes more efficient decisions- an improvement of generic skills or doing things better in the context of the coach's specific designs?

There is a "moneyball" analogy. There is a difference between a coach chastising a player for not swinging at more first pitches, and a coach criticizing that player for swinging at too many pitches out of the strike zone.

steveshea wrote:To expand a bit on what I'm after. A coach might want his players to play more efficiently, to (for example) not take the contested jumper early in the shot clock. If the coach is consistently celebrating player scoring with no regard to how efficiently they produce those points, the stats probably aren't going to be of much help in motivating players to be more efficient. However, if the coach was regularly recapping player efficiency (with some appropriate metric), and making known that efficient players would get more playing time, the stat could be a motivational tool. Is the result-a player that chooses better shots, or generally makes more efficient decisions- an improvement of generic skills or doing things better in the context of the coach's specific designs?

This perhaps identifies a tweener category, but if push came to shove I would put optimizing shot selection with respect to the shot clock in the "generic skills" box because the basics of this are general: shot quality thresholds decline over the possession and, on average, shots neither at the rim or behind the three point line are suboptimal except as a final option.

But another point should be emphasized about the particular issue raised: there are no advanced metrics required to communicate the underlying point. At the most, recourse could be had to realized averages as a function of time and location, but even that is unnecessary. Necessary communication: do not take any contested mid/long-range 2s except if it happens to be the last possible shot on the clock, and (depending on whom the coach is speaking to) only possibly take open ones as third or final options.

Precisely what I was going to say. Surely good coach-to-player communication in these cases would be what Schtevie said: "don't shoot a long 2-pointer unless there's less than 7 seconds left on the clock" or whatever. Unless the player is someone like Shane Battier, they're not going to ask for the theoretical, stats-based justification. At most they'll ask "why" and the coach simply has to say: they only give us two points, and they are a low percentage shot compared to lay-ins. Which any 5th grade basketball player can understand, and then nod and either decide to play the way the coach says or else be a knucklehead and keep gunning, and I very much doubt that advanced stats are going to change the knucklehead's mind.

Of course if the player in question is Ray Allen or LaMarcus Aldridge, you tell them something different: don't shoot the 2 unless you've got some space.


Defense is probably a lot harder, there are a lot more choices with a lot more gradations than shoot-or-don't-shoot. But once again the coach should probably emphasize what to do, rather than trying to give a lot of theory explaining why to do it.

If the Toronto Raptors have reportedly produced algorithms that show the best defensive positioning for every type of offensive movement (I read something recently where Alex Rucker I believe said this), it would seem theoretically possible that one could put an electronic device (with GPS and wireless connection to a computer) on players in practice where the device could be programmed to beep when a player moved near to the optimal way or was way off and give the player instantaneous coaching. It could possibly be programmed to shout off instructions (left, back, faster, jump, switch, etc.) as the play proceeds or maybe give him the feedback immediately after the play.