APBRmetrics

A plea to all sportswriters and commentators, tasked with discussing the issue/role/current state/future of analytics in basketball (or any other sport for that matter): all such articles and commentary would be vastly improved by explicitly dealing with the implications of the fact that basketball is a zero sum game.

I assume you mean because one team wins and the other team loses? I'd have to think a bit more, but it isn't immediately clear to me that this applies to the NHL, where a losing team can still move up in the standings (provided they lose after regulation).

Also, it isn't so clear that it applies to any 'analytics' aimed at the economics of the game. Decisions can be made that would increase or decrease the income of every team, or every player, or both teams and players. Or they could increase for some and decrease for others, but not equally. But I'm sure that all the sportswriters/commentators that read this plea (if any?) could improve their work by talking about which situation they're in and the subsequent implications (if any).

To be (a little more) clear, the "zero-summedness" I was invoking was both game and season-wise. If all teams become "analytical" (e.g. eventually learn that 1.5*X > X, for non-negative X, but we're not there quite yet!) as a raft of recent articles seem to suggest we are at the cusp of, on average, no one benefits from having invested in such a "competitive edge". This is the general result of all such arms races.

All that ever matters is having a superiority within any competitive margin and maintaining it.

Does Mr. Ranadive have any plausible reason to believe he's got any advantage over Mr. Morey?

http://espn.go.com/blog/truehoop/post/_ ... k-ranadive

Does Mr. Zarren have any reason to believe that he's getting more out of SportsVu than Mr. Cuban? And if so, does he think he will tomorrow?

http://www.bostonglobe.com/sports/2014/ ... story.html

Such a (skeptical) perspective ought to inform every single article written about analytics. Because... that's the reality.

schtevie wrote:To be (a little more) clear, the "zero-summedness" I was invoking was both game and season-wise. If all teams become "analytical" (e.g. eventually learn that 1.5*X > X, for non-negative X, but we're not there quite yet!) as a raft of recent articles seem to suggest we are at the cusp of, on average, no one benefits from having invested in such a "competitive edge". This is the general result of all such arms races.

All that ever matters is having a superiority within any competitive margin and maintaining it.

Does Mr. Ranadive have any plausible reason to believe he's got any advantage over Mr. Morey?

http://espn.go.com/blog/truehoop/post/_ ... k-ranadive

Does Mr. Zarren have any reason to believe that he's getting more out of SportsVu than Mr. Cuban? And if so, does he think he will tomorrow?

http://www.bostonglobe.com/sports/2014/ ... story.html

Such a (skeptical) perspective ought to inform every single article written about analytics. Because... that's the reality.

ALL teams will have or already have all the newer, fancier, more detailed data. Who will have the analysts who best interpret the data for the truest results (which players are better, and in which circumstances).

Let's just say there's nothing anywhere close to a consensus on the best metrics for evaluating players when looking at traditional box score stats after ALL these years. PER? WP? WS? Mike G (sorry, just forgot his metric - EPW?), mine, etc. I'm still working on w/ conventional stats because I think I can do better than what I've seen in terms of historical rankings. Plus, I love the challenge of college metrics, and all one has are the conventional stats.

So, can you imagine how completely differently some analysts will look at all the new data? How biased w/ these analysts already be? If Dave Berri incorporated a ton new data into his WP - he'd probably come up with VASTLY different results than Hollinger or DeanO or Mike G or myself or anyone else I've forgotten to mention.

It'll still be whose analysts are best (when it comes to the analytics of the game) - since everyone has the same data. Since Gms gm & coaches coach - having the best analysts may not matter at all in some cases (Knicks?Kings?), but matter a ton in others (Rockets? Spurs?).

And, I'm not even sure my post even relates to the point that was initially being made in this thread. Oh well.

Statman wrote:And, I'm not even sure my post even relates to the point that was initially being made in this thread. Oh well.

Actually it's a pretty direct response to my point that:

schtevie wrote:All that ever matters is having a superiority within any competitive margin and maintaining it.

What I take from your remarks is that you are of the belief that there is a significant diversity of opinion about relevant empirical issues, such that those teams employing more astute analysts will enjoy a significant competitive advantage.

Regarding diversity of opinion, I expect that there is really rather little of this, especially at the increasingly crowded frontier of best practice.

Think of it this way. When it comes to talent evaluation (surely the most important traditional analytic function?) what range of opinion is "possible" within a group of people who (a) have a common and correct understanding of the basketball "production function" (b) have any and all +/- estimates of key parameters available, and (c) who can now watch every single play of every single game ever to help rationally evaluate any and all reasoned conjectures?

The implication of this view is that the analytics revolution is essentially over (if it ever really began) with whatever advantage realized by the earliest innovators being inexorably eroded by the laggards rapidly bringing up the rear.

(The romantic in me, however, still has high hopes for the meteoric rise of the Houston Vipers.)

First, great discussion point. I agree and disagree with the premise at the same time.

I agree that, as teams increase and equalize their investments in analytics, the returns brought by these investments will diminish.

However, I do not think that the returns will diminish to the point of becoming not worth that investment. What we are basically looking at is a market (or a collection of markets) where everyone is searching for inefficiencies to exploit. As these inefficiencies are discovered (by most teams), they go away, so the argument is that the playing field becomes more level. However, the counter-point is that the market itself evolves and changes, both naturally on it's own and also due to the changes in direction brought about by analytics. So new efficiencies are introduced as old ones disappear.

The easiest example for this is Wall Street. A much older, more mature market, that has had analytics applied to it for a much longer time, yet there are still edges to be had.

I think there's also a question of which strategies are actionable for which teams. Maybe the most agreed-upon 'analytics' results is that having a good team is rarely (if ever) sufficient to win a title; you need to have a top-15 or top-10 player. Since there are more teams in the league than that, not every team can use that strategy or strategies related to building a team around such a player.

You could argue that every team is at least trying to acquire such a player though, so maybe it's a bad example. Another example would be spacing. Having data at a fine enough level is relatively new, but it seems like being able to space the floor, particularly with a big man, is valuable. But there is either disagreement on that concept across teams or not enough quality big men/shooters to do so, because not every team is doing it. In any case, even if analytics were to evolve to the point where everyone agreed on 'best practices', there's no guarantee that every team would be able to follow those practices, and so we would likely still see a spread of strategies across teams.

Since everyone knows that 3>2, defenses protect against it exactly as much as offenses try to produce it.
Yet 3-point shooting is at an all-time high, at 21.4 attempts per game. After a 5-season plateau at 18-18.4, 2 years of record growth. Accompanying the surge in attempts, accuracy from the arc is at .360, better than all but 2 previous NBA seasons.

Two years ago, 20% of all points were from made 3FG. Now it's 23%.
eFG% is .500, just .001 short of the alltime high in 2010.
This year's 100.8 PPG is the highest since 1995.

Among the most prolific 3-pt shooters, 3FG% is more volatile than TS%, typically varying by .100 or more from a player's best to worst season. TS% tends to vary by about half as much.
It's a delicate skill. Some players get good at it after 5 or 10 years in the league; others lose it quickly.