Research topic(s) that could be pursued

[Crow]

Teams like young players at low prices. Teams tend to guard their draft picks pretty hard; and the best teams while they are good usually have lesser draft slots. One way good teams can get additional young players at low prices is to trade for them or pick them up when the first team of slow developing / bad performing players release them before the end of the full extent of the rookie contract process. I therefore wonder which teams have the best recent record of claiming and developing these castoffs. San Antonio might be a leader. It would be interesting to see a full study. Maybe I'll get around to part or all of it later. But, if anyone has looked into this or wants to, it would be welcome. I don't recall seeing anything like this previously. (Does anyone else?) The metric for the evaluation could be winshares or RAPM or other and the time periods looked could be first year, peak and total career with team (and beyond if traded for other value). It would also be interesting to know which stat elements or factors each team enhanced more or less than the other characteristics. That could be based on team style of play, what the needs were as well as what the coaching staff excels at developing. The relative rarity of certain types of productivity and the average cost for obtaining it could be studied as well so that teams are evaluated on what they procured and developed along with the efficiency of that development.

I'd think that this process can be fairly important for "sustained success" but not uniformly. Some good teams will instead use salary cap exceptions or other more expensive methods and some will just fail to do enough of this. It might be good to see if market size or GM / coach characteristics reveal a pattern with regard to these choices or if these choices have a notable impact of degree of team success or length of time in top contention. I am somewhat suspicious that some teams can "overdo" their "sustained success" philosophy and bring in young players to develop (and reduce payroll and taxes) and not do enough / the best possible to maximize their peak year or two opportunities to go from contention to actual title achievement aided by more ready / veteran players (usually at higher prices).

[Crow]

Has anyone done or seen a zone to zone passing turnover rate study? With say 7-15 zones on the court. It was very valuable when there was data developed for shooting by zone and it also would be valuable to see what the average risk of a passing turnover was from zone to zone. Of course it also would be valuable to take this down to team, lineup and player levels.

Likewise, with the data available these days, it would be useful to see player level rebounding rates for shots taken from zones and rebounding position by zone. It would be good to know which shots players do best and worst at rebounding and for which they are getting better or worse position to rebound from.

[bondom343]

Just a thought that came up on RealGM and wanted to see if anyone had looked into it. Is there any sort of trend late in games where a team has a decent sized FT disparity to tend to give FTs to the other team? In other words (and I'm a Thunder fan so I'm not saying it was a poorly officiated game tonight, but the theory applies), tonight OKC had a pretty good sized FT advantage, but in the final minutes SA got a bunch of foul shots to narrow the final margin to 31-16 (still large, but smaller than it was). Is this thing a trend, or just random? A few people over there noticed it and couldn't tell if it was just a mind trick or not. Wasn't sure if anyone ever looked into it.

[Mike G]

I've not heard of such a study, but it's reasonable to suppose that the refs saw something like a 30-5 FT disparity and decided to bring it out of the uncanny range.
Spurs starters ended up with total 4FTA; 12 by subs -- and only 1 between Ginobili and Diaw.
All 31 OKC FTA were by their top 6 guys.

Meanwhile, I'd like to see a study on whether player performances change more from series to series than it does among random opponents (regular season) or among games within a series.
In 3 series, Kawhi has been cold-hot-cold; Manu has been hot-cold-hot.

[Nathan]

One thing I've always wanted to do was to make a player rating system that, instead of only considering games from the current season and weighting them all equally, considers all games in a player's career and weights them according to how long ago they were, while also accounting for the RAPM aging curve. I'm thinking I would have one "mathematically correct" version where the weights are determined by a regression (probably assuming the importance of a game decreases exponentially with time), and one "fun" version where recent games are weighted a bit more heavily and players can shift around significantly on a week-to-week basis depending on how well they're playing at the time.

The main problem is I would need a database of game logs for all players, which I have no idea how to make. I guess I could try to design a macro to scrape them from bball reference, but that sounds like a nightmare. Is anyone else interested in taking up this project, or does anyone have advice on how to make the requisite database?

[Mike G]

Nathan, I'd say all games from previous seasons could be weighted by [years ago]. And just weight this year's games (individually) by recency.
At the beginning of a season, then, you'd have only whole season data, with the previous season weighing most heavily. If a player started last year as a rebounding fiend, only to lose strength as the season wore on -- Which is the more viable number: The one that says he was a pretty good rebounder coming into this year? Or that he is a fading once-great rebounder?

I don't think there's a one-size-fits-all assumption here. Some guys recognize and correct for inconsistency, in the off season. Others fall off the table.

[J.E.]

does anyone have advice on how to make the requisite database?

You could buy PlayByPlay data from nbastuffer

[Mike G]

Has anyone studied PBP to calculate the W% of possessions in a game?
If a possession is defined as one for each team, and both teams score 0, 1, 2, or 3, that's a tie.
If you outscore your opponent by 2-0, or 3-1, that could be called a Win.
A 3-2 possession could be half a win, and 3-0 would be 1.5

If a team wins a game, 100-99, would their Possession Win% will be something like .505 for the game? Or would there be a range of W%, depending on how many ties there are?
A lot of 0-0, 1-1, 2-2, and 3-3 possessions would leave fewer that are won or lost. Conversely, a game in which teams both have major runs would have lots of possessions won and lost by both teams.

Suppose we don't count tied possessions in the W%. Then a game of 94 possessions might have some 40 to 60 which were Won by one team or the other?
Not sure how useful this info is, but it might put perspective on things. What's the pW% in a blowout game?

[Crow]

8 teams had 5 guys in top 20s estimates by RPM by position: ATL, BKN, GSW, HOU, IND, NYK, OKC and SAS (with 8!). Most teams in the east got more DRPM from these guys than ORPM but most noticeably not MIA. In the west it was 2-1 more from ORPM. 4 playoff teams only had 3 such players- CHA, POR, TOR and WAS.

Dividing them by high and low usage, the groups were almost of even size. The high usage players averaged +2.5 on offense and +0.8 on defense for a total impact estimate of +3.3. The lower usage guys were almost a mirror image at +0.6 on offense and +2.4 on defense for a total impact estimate of +3.0. But the high usage guys played 29% more minutes per game and more games and thus had a 39% higher WAR.

20 guys were +1 or better on both sides of the court, with this occurring by far the most often for PFs. The only players +2 or better on both sides of the court were Blake Griffin and Nick Collison. 4 players just barely missed. The best one sided performance was from James on offense and Garnett on defense. Larry Sanders was the worst in the group on offense but 2nd best on defense.


Another thing that could be done with sorting RPM values would be to look at the ORPM, DRPM and overall RPM values for players from the 10-25 colleges that have sent the most players to the pros to see what the surface trends are. For those with access to past factor level RAPM, it could be taken another level deeper.

I may get around to looking at average performance on RPM by certain other measurements and stats later. That might be of some use for draft planning and other player decision-making.

[Crow]

Mike, I think most are satisfied to know team performance by quarter, but with adequate time, data and skills it might be useful for teams to chart the greater detail that you suggest. With lots of resources I'd take it down to possession by possession win rates of lineups. I recall being somewhat dismissive of the utility of Dean Oliver's possession % (% of possessions with some scoring) in the distant past, but for a max analytic effort it might be a part of the analysis used to guide game strategy in general and especially when up or down by a good amount and / or late.

[Crow]

With only 20 players plus one or better on RPM on both O and D, the label of two way strong would seem to apply to only about 5% at most. One way strong, other way neutral or not so bad appears to be the more common case / realistic goal. Maybe this argues for more acceptance and focus on one way specialists.

[Nathan]

Sorry if this has already been done and I missed it, but one thought I had recently was to use box score stats to predict changes in RAPM.

For instance, take all players age 25 and younger, and look the change in their RAPM numbers from last year to this year. Then see if it's possible to predict these changes to any extent using the players' box score stats from last year.

On the other hand, take all players age 30 and older, and look at the change in their RAPM numbers from last year to this year. Is it possible to say using box score stats which vets are more likely to fall off a cliff and which vets are more likely to remain productive players well into their 30's?

[J.E.]

Sorry if this has already been done and I missed it, but one thought I had recently was to use box score stats to predict changes in RAPM.

For instance, take all players age 25 and younger, and look the change in their RAPM numbers from last year to this year. Then see if it's possible to predict these changes to any extent using the players' box score stats from last year.

On the other hand, take all players age 30 and older, and look at the change in their RAPM numbers from last year to this year. Is it possible to say using box score stats which vets are more likely to fall off a cliff and which vets are more likely to remain productive players well into their 30's?

I like this. I'd say that, to some extent, xRAPM/RPM already do this. The weights for my SPM, which serves as a prior in xRAPM/RPM, were set up to give best prediction results the following season, i.e. I regress BoxScore stats of year X-1 onto points-per-possession-data of 10-man-units in year X.

I've toyed with the idea to find weights for BoxScore data of year X-2 to predict year X, but haven't done so, yet

[Nathan]

Interesting thought...I'm finding in making my draft rater that for non-one-and-done players, previous college seasons are nearly as predictive of NBA success as their most recent season, which surprised me a bit.

One other thought I'll throw out there is the following. I remember seeing on here a few months ago a stat showing that teams performed significantly better when facing a deficit than when holding a lead. This suggests that many teams do not play at or even near their maximum potential for long stretches of most games. Could this have to do with coaching? I would expect teams with the worst coaches to have the largest discrepancies between how they perform when holding a lead vs. when facing a deficit, reflecting the fact that the team is playing far below their potential except when they're facing a deficit. I would expect teams with the best coaches to have the smallest discrepancies between how they perform when holding a lead vs. when facing a deficit, reflecting the fact that the team is playing the best basketball they're capable of, or close to it, no matter the situation (note: even if a team's effort is perfectly consistent, they will still fare better when trailing because their average opponent's effort is not perfectly constant). Does this turn out to be the case?

[J.E.]

One other thought I'll throw out there is the following. I remember seeing on here a few months ago a stat showing that teams performed significantly better when facing a deficit than when holding a lead. This suggests that many teams do not play at or even near their maximum potential for long stretches of most games. Could this have to do with coaching? I would expect teams with the worst coaches to have the largest discrepancies between how they perform when holding a lead vs. when facing a deficit, reflecting the fact that the team is playing far below their potential except when they're facing a deficit. I would expect teams with the best coaches to have the smallest discrepancies between how they perform when holding a lead vs. when facing a deficit, reflecting the fact that the team is playing the best basketball they're capable of, or close to it, no matter the situation (note: even if a team's effort is perfectly consistent, they will still fare better when trailing because their average opponent's effort is not perfectly constant). Does this turn out to be the case?

I think the 'effect of leading' is mostly human nature. For one, pacing yourself in these situations decreases unnecessary injury risk. I doubt that coaches have much influence over this. Maybe a new coach will be succesful at keeping motivation high when up 20 in the first ~5 games but I'm sure it will loose its effect and players will start shaking their head at a coach that screams at them to continue to bust their asses even though they're up 25. You can't go full speed all the time, you have to pick your spots, and being up 10+ is probably not such a spot (unless it's a playoff game)