APBRmetrics

[Ben F.]

I ask this because I've gotten some interesting results from my simulation. I want to know how you think this team would perform in the NBA. The team is made up of a collection of good players, but doesn't really have any stars. They all shoot well, play good defense, don't really turn the ball over, and rebound well.

MPG for each player in parenthesis.

PG: Antonio Daniels (28 ), Damon Jones (16), Devin Harris (4)
SG: Manu Ginobili (32), Damon Jones (16)
SF: Shane Battier (32), Anthony Peeler (16)
PF: Jeff Foster (16), Mehmet Okur (16), Mike Sweetney (16)
C: Marcus Camby (32), Jeff Foster (16)

The question is, does a team need a star to do well? Is Ginobili really that good, or is it just because he plays with Duncan? Or using language from Dean Oliver's book, would the increased usage of possessions on this team for role players decrease their overall efficiency? Or would they still perform well?

I've just been VERY surprised by the results I've gotten from them, so was wondering if this was an error, a flaw in the design, or could actually happen.

[drebelx]

I too am interested in how a team like that would perform. It sounds like you may be using parameters that I have thought would be ideal for a basketball team, high %'s, blocks and steals indicating good defense, low TO's and good rebounding. By the way are you neglecting assists and points scored in you simulation?

[bchaikin]

simulation shows 60+ wins in an average 82 game season, playing each nba team 273/274 games in a simulated 8200 game season (remember this is against the other nba teams and not a "draft" league where only the best 150/200 players in the league are used rather than the full 360 players in the nba)...

this is a very good defensive team (daniels, ginobili, and battier are all excellent defenders) that would have a very high Scoring FG% (.563) as a team versus their opponents (.495). the only team in the nba right now with a Scoring FG% that high is phoenix, and no one else is even close (no one else in the nba right now is even at 55%, the current league average is 51.9%)...

this team would have off/def FG%s of .465 vs. .427, shoot almost 10 more FTA/g and get 6 more reb/g than their opponents, commit less than 14 TO/g, and commit only 21 fouls/g versus 26 fouls/g by their opponents...

camby, foster, and battier would get 11 pts/g each, ginobili 20 pts/g, and daniels 15 pts/g. camby and foster would get 10 reb/g each, ginobili, daniels, and d.jones 5 ast/g each...

daniels, ginobili, and battier are easily three of the most underated players in the league, and 82games.com shows foster, battier, and ginobili having excellent +/- numbers...

as for "...would the increased usage of possessions on this team for role players decrease their overall efficiency?...", i'd like to see if there is any statistical evidence for this. for the above team the following players touches would each increase over what they are currently getting in real life in 04-05 by:

8% - camby
12% - foster
10% - battier
15% - ginobili
20% - daniels

but - camby had just as high touches/min in real life in 96-97 as he gets in the simulation, foster did in 01-02, battier did in 01-02, and daniels almost as much in 03-04 (ginobili's high in touches/min is this season), so the increased touches/min are right in line with what these players have gotten at some points in their careers...

[Dan Rosenbaum]

Bob, your comments got me thinking. Does your simulator assume an inverse relationship between touches and efficiency, or as in BoP, possession usage and efficiency?

I would think that you would have to build in an assumption that as a player is forced to use more possessions (or more touches), his efficiency falls. Otherwise, I would suspect that your simulator would say that surrounding Fred Hoiberg with Michael Ruffin, Trent Hassell, Brevin Knight, and Tyson Chandler would likely lead to Hoiberg challenging for the scoring lead in the league without a decrease in his efficiency. Such a prediction almost surely would be false, as it would pretty much turn Hoiberg into an MVP candidate.

But building in such an assumption is tough, because the relationship between touches and efficiency surely is something that differs from player to player and is one of the more difficult relationships to estimate without bias.

[bchaikin]

Bob, your comments got me thinking. Does your simulator assume an inverse relationship between touches and efficiency, or as in BoP, possession usage and efficiency? ...

i have deano's book, so before i answer this, please tell me where in his book does he talk about this? i'd like to see the statistical proof of this before i respond...

[Kevin Pelton]

Just for the record, here are the estimated percentage of possessions and shots these players are using/taking so far this season:

[Dan Rosenbaum]

[Mike G]

I'm not sure if we're using this year's stats or last year's; the inclusion of a rookie suggests it's 2004-05.

Assuming we're using this year, and assuming none of these players misses many games -- a big assumption, I should say -- my eWins for the group projects to 60.3, almost exactly what BobC came up with.

The 9 non-rookies averaged 72 games last year, and only Foster and Peeler have missed significant time this year. That might be pretty close to league average "availability", so you could suppose these estimates are OK.

The lack of a "pure" point guard might also affect some of these players' effective scoring. Damon Jones might revert to that role.

[bchaikin]

Bob, your comments got me thinking. Does your simulator assume an inverse relationship between touches and efficiency, or as in BoP, possession usage and efficiency?
...

nope...

I would think that you would have to build in an assumption that as a player is forced to use more possessions (or more touches), his efficiency falls.
...

uh... now why would you assume that?... i wouldn't assume that at all...

you may think that intuitively this is a correct statement, but i would need to see the data proving it...

Otherwise, I would suspect that your simulator would say that surrounding Fred Hoiberg with Michael Ruffin, Trent Hassell, Brevin Knight, and Tyson Chandler would likely lead to Hoiberg challenging for the scoring lead in the league without a decrease in his efficiency.
...

is this you saying this, or the simulator? in either case you would suspect wrong...

Such a prediction almost surely would be false, as it would pretty much turn Hoiberg into an MVP candidate.
...

false, huh? uh... ok... again, who's making the predictions here, you or the simulator?...

when i simulate the above starting five (chandler, ruffin, hassell, hoiberg, and knight) against the rest of the league they win only about 40 games per average 82 game season. hoiberg gets about 20 pts/g (shooting the same very high Scoring FG% he is in real life now), chandler 15 pts/g, ruffin 4 pts/g, hassell 13 pts/g, and knight 15 pts/g, with each playing 36 min/g. all have their touches/min increased significantly, but lets look at this realistically...

coming into this season brevin knight had been in the league for 7 years. because of the current makeup of the charlotte bobcats, he is now getting his highest touches/min of his career. if i had told you prior to this season that a journeyman PG who had played over 2000 minutes in a season just once in 7 years was going be on an expansion team - and be 2nd or 3rd in the league in assists per game playing just 28 min/g, would you have believed me?...

knight is getting 8.2 astg playing just 28 min/g, stephon marbury is getting 8.3 ast/g playing 40 min/g. if nash wasn't having such an outrageous season in phoenix, we could realistically be seeing a journeyman player play less than 30 min/g leading the league in ast/g on one of the worst teams in the league....

except for possibly his rookie season, knight is having probably his best season statistically - so tell me, does the assumption for decreasing a player's efficiency with increased touches cover brevin knight? or is there a qualifier just for low touches/min SGs like fred hoiberg?...

and again this is real life, not simulation...

my point is that before making assumptions check the historical records. brevin knight is just one current example, there are more...

fred hoiberg may not shoot a very high Scoring FG% of 66% with increased touches, but in 00-01 he played 2200+ minutes for the bulls with touches per minute of 1.0 and he shot a Scoring FG% of 59%, 6th best in the league of all players playing at least 2000 minutes, so there is historical precedence for him shooting very well with increased touches and increased minutes...

But building in such an assumption is tough, because the relationship between touches and efficiency surely is something that differs from player to player and is one of the more difficult relationships to estimate without bias.


This is Chapter 19 of Basketball on Paper - the discussion of skill curves. I hate to speak for DeanO, but I don't think there is a way to "statistically prove" that efficiency falls as possession usage rises. But without such a relationship, the strategy for teams would be to allow the most efficient player to use all of his team's possessions. That doesn't happen because players are forced into less efficient possessions (e.g. worse shots) as their possession usage rises.


i read DeanO's chapter 19, and i simply don't concur with his findings. if i understand his comments correctly (and i may not be) he is saying allen iverson and jerry stackhouse are less efficient players with more touches. so let's look at this...

in 00-01 iverson scored 31 pts/g. that season i have him getting 1.5 touches/min. you will notice his Scoring FG% that season was .508, the 2nd highest of his 7 year career (not including this season), with turnovers on just 5% of his touches (about average for him). that .508 was very close to the league average Scoring FG% that season of .510....

so here you have a player who shoots or gets fouled on 54% of his touches, is shooting right at the league average Scoring FG%, but with very low turnovers on just 5% of his touches. i don't find that to be inefficient at all - on the contrary i call that being very efficient...

now look at this 04-05 season. iverson is getting close to 2.0 touches/min, is shooting a Scoring FG% of .513, right near the league average of .519, is shooting or getting fouled on 40% of his touches (he's passing more per touch this season versus in 00-01), and is getting you 30 pts/g with turnovers on - again - just 5% of his touches....

so here you see iverson getting his most and least touches/min yet being what i believe to be efficient in both cases. he was quite "inefficient" in 03-04 because he shot so poorly, his Scoring FG% that season was just .468, four percentage points below the league average. if you are shooting that poorly yet taking over 30 scoring opportunities a game (one scoring opportunity being a FGA or 2 FTAs), you're not helping your team...

as for jerry stackhouse, in 00-01 he scored 29.8 pts/g on 1.6-1.7 touches/min, both career highes for him. he did that with a Scoring FG% of .510, right at the league average that season, and right near his career Scoring FG% of .513. plus that season his turnovers per touch were at just
6%, low for him as he's had seasons with turnovers per touch of 8% and 9%. again - shooting the league average Scoring FG% with turnovers on just 6% of your touches, while shooting or getting fouled on 50% of your touches, to me is pretty efficient....

his FG% that season was just 40%, but his Scoring FG% was right at the league average because he took 810 FTAs and hit 82% of them. that's like taking an additional 405 shots but hitting them at an 82% clip...

DeanO also talks about derek fisher being more efficient with less possessions with kobe and shaq around when he was on the lakers. but right now fisher is getting his most touches/min (on golden state) in 6 seasons, and his Scoring FG% is .514, better than his career Scoring FG% of .498, and his turnovers per touch are at just 4%, about average for him, and he's scoring a career best 12.6 pts/g playing just 31 min/g (he had 2 seasons with the lakers where he played more minutes)...

If, as Bob suggests, you use career-high figures, you can get to 100%, but the vast majority of this increase is due to Camby and I'm loathe to think that a 22-year-old rookie Camby tells us much about what he could do now. In practice, I think Daniels could probably add a fair number of possessions without a huge efficiency loss, but I'm dubious on the ability of the other four guys to do so.


the vast majority of players increase their touches/min after their rookie season, camby is an exception. but in any case his touches/min are now just 8% less than his rookie season. the simulation example i ran with camby has his touches/min increasing by 9%. i don't see the difference here - the simulation shows him getting touches/min that he did get at one time during his career, so that is in fact realistic...

Off of a spreadsheet, a frontline of Foster and Camby is probably going to get pushed around a fair bit by stouter players, I would think.
....

i would think not - currently marcus camby is the 3rd best defensive rebounder in the league behind kevin garnett and reggie evans. jeff foster is currently the 3rd best offensive rebounder in the league behind danny fortson and dan gadzuric. they appear to be doing their own fair share of pushing...

yes the simulation does assume players are not getting injured, i.e. players like camby and foster playing all 82 games, but remember the simulation is a model and looks at both best case and worst case (a poor peforming player playing major minutes) scenarios...

The lack of a "pure" point guard might also affect some of these players' effective scoring. Damon Jones might revert to that role....

the past 2 season antonio daniels has passed the ball with 2/3 or more of his touches, getting about 1.5 touches/min over the 2 seasons. that's about as pure of a PG as you'll find, with higher passes thrown per touch than the mike bibby's, allen iverson's, gilbert arenas's, and steve francis's of the 04-05 world...

[/b]

[gabefarkas]

i just went back and re-read Chapter 19. you know what's been bugging me about it this entire time? the axes seem to be reversed from what they should be. the x-axis is efficiency and the y-axis is touches.

shouldn't it be the other way around? touches is the independent variable, and efficiency is the dependent variable. right?

[gabefarkas]

as for the 12 chosen players that started this discussion, it's not surprising that they will win a lot of games. all 12 of them are starter or prime backup quality. there's very little dropoff between the 4th-5th and 11th-12th players. where's the Bruno Sundovs, David Harrisons, Mateen Cleaves, Obinna Ekezies, and Tierre Browns in that group?

[Mike G]

Gabe,

To your 1st post: You might imagine that guys get touches based on their efficiency; i.e., How many touches should this guy be getting?

2nd post: There are only 10 players on this hypothetical team. One of them is only getting 4 minutes. (That's why I was leery about assumed player availablility, AKA the injury factor).

[mathom]

Yeah, the axes should be switched. Somewhere on APBR analysis awhile back I remember Dean posting that they were supposed to be switched.

[Dan Rosenbaum]

Bob, the basic question is this. If there is no relationship between possession usage and efficiency, then why would the Wolves let anyone other than Hoiberg shoot if he is their most efficient player? Why would the Sixers let anyone other than Korver shoot if he is their most efficient player? Why would Seattle let anyone other than Daniels shoot if he is their most efficient player? If you were giving advice to these teams, why wouldn't your advice simply be to have their most efficient shooters shoot every time down the court?

Maybe I am dumb, but the only thing I can think of that would make that poor advice is that forcing those players into more shots would result in their efficiency falling.

This is theory, but it is also basic economics. At low possession usages, players only try to score on the chances where they have the highest probability of success - breakaways, tip-ins under the basket, wide open jumpers, etc. But as their possession usage rises, they start to add in contested shots - floaters in the lane, jump hooks over an outstretched arm, fadeaway jumpers, etc. Push their possession usage way up and they then are forced to create lots of their own shots and some of the shots become very low percentage.

Thinking about this process we can imagine that for some players their efficiency drops off precipitously as their usage increases, while for others the drop is more gradual.

Now it is tough to estimate this relationship. For example, if we use game-to-game differences in usage to estimate this relationship, it is likely the case the higher usage games were games where the player was in a favorable match-up or was healthier. Comparisons between seasons would be better, but it still would be difficult to disentangle this relationship away from players simply getting better and worse or playing for coaches that do a better job optimizing a player's talents.

Hoiberg, I think, is a decent example of what DeanO calls a skill curve. There have been three seasons where he played at least 20 minutes per game, and I will focus on those since his minutes may have been mostly garbage minutes in those other seasons. Hoiberg's efficiency seems to fall quite a bit as his usage increases. At 20 percent or higher usage, I would imagine that his efficiency would be below average.

Season Usage Efficiency
1999-00 16.5 106
2000-01 13.2 119
2003-04 11.5 121

http://www.basketball-reference.com/players/h/hoibefr01.html

Brevin Knight, I think, has been more efficient with higher usage. His assists per 40 minutes are the highest of his career (although not that much higher than in some previous seasons) and his shooting percentages are pretty typical for him. Charlotte has organized their offense around their point guards with even Steve Smith generating a lot of assists when he played point guard.

There have been two players, Keith Bogans and Kareem Rush, thrust into a scorer's role at the SG position. Both have seen their shooting percentages fall quite precipitously as their usage has increased from previous seasons.

But I think it is very difficult to estimate the relationhip underlying these skill curves. There are too many other things going on that are difficult to control for. In other words, it is hard to do the ceteris paribus comparisons that we would like. But at the end of the day, the nagging question is still the same. If there is not an inverse relationship between possession usage and efficiency, why don't teams have Korver, Daniels, Battier, Hoiberg, etc. shooting 30 shots per game? It could be that coaches and GMs are just dumb, but I think folks rightly realize that these players' efficiency would drop quite a bit if they were used much more intensively.

[S.K.]