What's the best publicly available defensive metric?

[deepak]

I thought J.E.'s defensive xRAPM was probably the best publicly available one before, but now that its not available any longer what else is out there? I like ASPM, but I'm not sure how reliable dASPM is (oASPM I'm much more confident in).

Would a combination of dASPM, defensive On/Off differential, and maybe PPP-against (Synergy) be worthwhile in lieu of a reliable defensive adjusted +/- stat?

[colts18]

Synergy

[DSMok1]

I thought J.E.'s defensive xRAPM was probably the best publicly available one before, but now that its not available any longer what else is out there? I like ASPM, but I'm not sure how reliable dASPM is (oASPM I'm much more confident in).

Would a combination of dASPM, defensive On/Off differential, and maybe PPP-against (Synergy) be worthwhile in lieu of a reliable defensive adjusted +/- stat?

DASPM is highly limited, I'll agree. About as good as you can do with box scores alone, but that only explains (R^2) 50% of long term RAPM.

RAPM (somebody please do a public RAPM!) is probably the way to go, and look at Synergy too if you can.

[mburke05]

I realize maybe this is naive, but I've been a lurker for a while and still haven't come to a conclusion.

Is there a way to acquire Synergy access now or are people who cite Synergy doing individual searches? How would one go about acquiring Synergy access? Is there a reason that data hasn't been scraped yet for use in modelling?

Again, sorry if this has been discussed somewhere, I just can't find any answers as to how to get Synergy subscription access on their website which looks quite post-dated.

[Kevin Pelton]

Not sure what happened, but it seems like everyone who subscribed to the public Synergy last year was grandfathered in for this season and they did not open it up to new subscribers.

[deepak]

I realize maybe this is naive, but I've been a lurker for a while and still haven't come to a conclusion.

Is there a way to acquire Synergy access now or are people who cite Synergy doing individual searches? How would one go about acquiring Synergy access? Is there a reason that data hasn't been scraped yet for use in modelling?

Again, sorry if this has been discussed somewhere, I just can't find any answers as to how to get Synergy subscription access on their website which looks quite post-dated.

My guess is that they're trying to limit the load on their servers. That's a lot of video data they need to push out if a lot of users are trying to use it simultaneously.

I believe you can still get the stats for the current season and limited video content, but not past seasons and full video access without a subscription.

[balkihowser]

Sorry if this is a basic question, but what does it mean to "use Synergy"? Does it mean having access to their video and actually looking at clips, or do they generate a dataset that one can use for analysis.

Thanks.

[Kevin Pelton]

Sorry if this is a basic question, but what does it mean to "use Synergy"? Does it mean having access to their video and actually looking at clips, or do they generate a dataset that one can use for analysis.

Since they attribute every play to a defender, Synergy calculates a points per play allowed (PPP) for players. Analysts have used the overall figure and also looked at those for specific types of plays (isolations, post-ups for big men, pick-and-rolls, etc.).

For Insiders, I thought these efforts from Bradford Doolittle were an interesting combination of publicly available metrics:
http://insider.espn.go.com/nba/story/_/ ... -defenders
http://insider.espn.go.com/nba/story/_/ ... -defenders

[mburke05]

Where are the current in-season statistics Kevin (or anybody), I'm having trouble navigating their site as is, so I'm guessing I'm missing something obvious.

[DSMok1]

Where are the current in-season statistics Kevin (or anybody), I'm having trouble navigating their site as is, so I'm guessing I'm missing something obvious.

It's rather odd--Synergy is a subscription-only service; the general public does not have access... currently. Last season, they had a MySynergy service, which allowed public access for a fee. This season, it appears that anyone who had access last season still has access, without an additional fee, but no one new can sign up.

Unfortunately, I had not signed up.

Sites like ESPN have organizational access to Synergy stats.

[mburke05]

If anybody currently has Synergy access I'd gladly pay to share it with you if that's a possibility, whatever is reasonable.

[NBAMetrics]

While Synergy is an awesome resource (albeit with a frustrating interface, for their public version at least) I'd caution putting too much stock in the PPP they assign to defenders against different play types. The data doesn't account for match-ups. A lot of players that are "hidden" on D (Novak) are going to have inflated numbers because they guard the worst offensive player on the floor, while the better defenders (Bradley) are going to show worse numbers because they guard the opposing teams best offensive threat. That and Synergy's play type assignments seem somewhat arbitrary, if the team is trying to run a PNR and the D manages to blow it up and force an isolation shot as the shot clock runs out, the play is listed as an isolation when IMO it should still count as a PNR (the D stopped the offense's intended play).

[sideshowbob]

Synergy Stats barely scratch the surface of an individual's defensive impact. They only focus on box score metrics and don't always credit the right player for the defensive play.

Where in the Synergy stats do they keep track of box-outs? If player A does an effective job of boxing out his man, and teammate B gets the rebound, shouldn't player A receive some credit for the rebound, which IS in fact, a defensive play?

If player A is covering his man (opp. A), and teammate B comes over in order to help trap, one of many things could happen. NONE of which, synergy keeps track of. Opp. A will be forced into a bad shot, credit will go to player A and not B. Opp. A will be forced into a turnover, credit will go to player A and not B. Opp. A will be forced to run out the shot clock and perhaps pass to a teammate for a bad shot, which misses; neither player A or player B receives credit.

Player A is slow to rotate, and player B has to compensate, player A's man spots up and hits a shot before Player B can recover to make up for the defensive mistake. Player B is credited with allowing a made field goal, while player A, who made the mistake, doesn't get discredited at all.

Player A mistimes a block on an opposing players shots, causing it to fall right into the hands of an opposing big man, who quickly puts it back in. Player B is credited with allowing a field goal, while player A is credited with preventing one. Once again, Player A's makes a mistake that Player B wrongly gets credit for.

What about ball-denial? Assume player A is guarding a superstar offensive player. If player A does an effective job at simply not allowing that player to get the ball, that effect will never show up in synergy stats. Don't tell me that's not good defense.

And I can go on and on. The Synergy stats are great as a reference, but they offer little to no context, and don't cover anything beyond the box score.

I'd be much more comfortable using an all encompassing +/- defensive stat that tries to capture global defensive impact (defensive RAPM) than using synergy.

[talkingpractice]

After some consideration, we're planning to throw up an open source "RAPM light" (defense only, no prior) over the next week or two.

[v-zero]

That's something at least. Will it be multi-year? Since single-year non-prior-informed RAPM is roughly useless.