APBRmetrics

[AGBCZAL]

I've been playing around with B-Ball, which I got from APBR file , simulating 95-96 NBA season and I've got a few questions about it

First, what is the usual difference between results generated by the same line-ups? As i've changed John Stockton to Terrel Brandon playing the same 36 minutes one time I've got about 56 wins and later about 58 (820 games season) I've been wondering about that.

And the second question is purely theoretical. As I understand it, in simulation assist is "just a pass which happened to become an assist". But as I tried reducing Stockton's possesion factor by 50% and 75% percent, he handled the ball less, had less assists but the team won around the same (about 1 win difference). So the question is what about players who seems to be assist dependent (which have majority of their shots assisted)? Shouldn't the loss of a few hundreds assists hurt them?

[bchaikin]

First, what is the usual difference between results generated by the same line-ups?

when repeating an identical simulation of 8200 games total, you shouldn't see more than a 4 game swing at most over an average of 82 games. in other words if a team is 41-41 in real life with a per game point differential of zero, in the simulation you shouldn't see a difference of more than about 39-43 to 43-39 over the span of 8200 games, and its usually less than 2-3 games (1 to 1.5 games either way). from just one single 82 game season to the next the difference can easily be more (as could flipping a coin 82 times)...

As i've changed John Stockton to Terrel Brandon playing the same 36 minutes one time I've got about 56 wins and later about 58 (820 games season) I've been wondering about that.

brandon had an excellent season in 95-96. he was the 2nd highest scoring PG in the league per game that season, had the 2nd most points per minute of all PGs, shot a ScFG% (combining 2pters, 3pters, and FTs) of over 56%, got slightly more steals per minute than stockton did, and to top that off had the lowest rate of turnovers per touch of all starting PGs in the league (and with 40% less ast/min than stockton)...

if you simulate penny hardaway from 95-96 on the 95-96 jazz replacing stockton the jazz will win even more games, because that season hardaway had one of the very best seasons since 77-78 by a PG in terms of wins generated through the stats he produced. but he also had 1/3 less assists than stockton did...

And the second question is purely theoretical. As I understand it, in simulation assist is "just a pass which happened to become an assist". But as I tried reducing Stockton's possession factor by 50% and 75% percent, he handled the ball less, had less assists but the team won around the same (about 1 win difference). So the question is what about players who seems to be assist dependent (which have majority of their shots assisted)? Shouldn't the loss of a few hundreds assists hurt them?

when you reduce stockton's touches/min by 75% in the software his touches/min actually only end up getting reduced by a little more than 1/2. you then have a starting PG who gets about 400-500 assists but still shoots a ScFG% of 60%+ with more steals than turnovers (about 135 steals and 115 turnovers) playing 36 min/g. that same season the atlanta hawks had a starting PG (mookie blaylock) with just such stats - 400-500 assists with more steals than turnovers, and that team won just 9 less games than the jazz did, with blaylock shooting worse than stockton, and without an all-NBA 1st teamer like karl malone...

do you believe there is anything in particular that would make the jazz players more assist dependant than the hawks players? that season the jazz had 68% of all their FGM assisted on (highest ratio in the league) but the hawks just 54% (2nd lowest in the league)...

also that season the hawks and jazz PFs played the most minutes on their respective teams. if prior to the season they had swapped PFs, grant long for karl malone, do you think the hawks would have won more games and the jazz less? would karl malone have scored just as well with a PG that got just 1/2 the assists than his previous PG, or would long have scored more?...

just because stockton piled up the assists doesn't mean his teammates could not score without him. stockton played close to two decades in the league and a number of players who played for the jazz during that time also played for other teams. check their stats of how they played on the jazz versus other teams to see how well they scored with and without stockton...

also in 0203 and 0304 steve nash was the starting PG on the mavs. the team went 112-52 those two seasons with ast/fgm of .586. in 0405 and 0506 without nash they were slightly better as they went 118-46, but with ast/fgm of just .513. so they seemed to play just as well with or without the assists...

[AGBCZAL]

Firstly, thanks for the answers.

I've been looking for players in 82games.com who had their assisted fga changeing from one season to another and it seemed to me as there was no general trend in increasing their efg%. And Dirk Nowitzki was able to keep his level of shooting despite the departure of Nash. I guess I was only interested in Korver or Hoiberg type of players who seems to have about 90% of their fga assisted. Of course I am sure that Malone or Hornacek could score without Stocton's assists.

I simulated A. Johnson replacing Stockton for those 36 min and he seemed to be about 5 wins worse than Stockton. I cheked his numbers and it seemed to me that he was very good at ballhandling with just 3% of his touches being TOs and passing with 77% of his touches (95-96 season). Of course, he was poor rebounder and his ScFg was only average. And of course his defense was not as good as Stockton's. So it seems that these Mark Jackson, Avery Johnson type of palyers are overrated by people liking "passing" pg. For me the fun part is analysing the results of the simulations and understanding why something happened. That leads to reevaluations of players as one who may look good at the first glance is in fact much worse (By this simulation I get that D. Stoudamire is in fact on of the worst starting PGs in 95-96 despite scoring 19ppg).

One thing I would like to ask is how to compute toches/min reduction? If rdeucing by 75% means reducing by half, how much is reducing by 50%?
25?

[bchaikin]

So it seems that these Mark Jackson, Avery Johnson type of players are overrated by people liking "passing" pg.

yep...

most remember stockton primarily as a passer. but during his heyday (say 86-87 to 94-95) he averaged 15 pts/g playing 35 min/g but with very efficient shooting, a very high ScFG% of 60% (and low turnovers per touch). that's the best overall shooting of all players during that 9 year stretch except three - reggie miller, barkley, and mchale....

stockton also had by far the most total steals (over 300 more than the next player), and the 2nd highest rate of st/min next to alvin robertson. he averaged 218 steals a year - that's alot of defensive stops to his credit. heck, he could have gotten 1/2 the assists he got and he still would have been an excellent player...

as great as steve nash is, he has averaged just 65 steals a season over the past 7 years. imagine him getting 150 more steals each season like stockton did...

you can get a ton of assists but if you don't shoot well, or get steals or blocks, or play defense, or rebound, or do something else positive, those assists aren't going to generate a significant number of wins...

For me the fun part is analysing the results of the simulations and understanding why something happened.

it is a nice diagnostic tool...

That leads to reevaluations of players as one who may look good at the first glance is in fact much worse (By this simulation I get that D. Stoudamire is in fact on of the worst starting PGs in 95-96 despite scoring 19ppg).

among the PGs try jalen rose from that season, he's even worse...

One thing I would like to ask is how to compute toches/min reduction? If rdeucing by 75% means reducing by half, how much is reducing by 50%?
25?

depends on (1) the touches/min of the player's teammates, and (2) the ratio of how often each of those players shot, passed, got fouled, and turned the ball over per touch...

[AGBCZAL]

When simulating J. Rose I get that he is about 14.5 wins worse than A. Hardaway that season (per 36 min). That's quite huge. I was wondering what are the gaps at other positions at various seasons.

I've got an interesting result when simulating D. Wesley - he seemed to be as good as Stockton, but than I saw that he was one of the best shooting and rebounding PGs that season and I understood why he was so good. That left me wondering about some present players. How do B. Davis, B. Knight or T. J. Ford end up in simulation than compared to guys like S. Nash or C. Billups? As i've checked the latter two were shooting over 10% better in Scfg%last season.

Also I wanted to ask how does the software calculate game pace. I've looked at some teams and they were quite close to the real life.

And how did you rate the players defensively before 82games.com data was available? Using some stats or just by watching players?

[bchaikin]

When simulating J. Rose I get that he is about 14.5 wins worse than A. Hardaway that season (per 36 min). That's quite huge.

he wasn't too good that year - very high rate of turnovers per touch for someone with such high touches/min...

I was wondering what are the gaps at other positions at various seasons.

at 40 min/g and 82 games 14-15 wins between the best and worst starters at a specific position is pretty typical. in 05-06 for example the difference between PFs elton brand and juwan howard on a 40 min/g and 82 game basis was 16-18 wins, depending on which team you did the simulation on (brand was very good that season)...

the gap between the best and worst SGs in say 87-88 is larger, specifically because jordan had his best season statistically then. this 06-07 season the gap in SFs will also be large because adam morrison is playing so poorly statistically (yes he's not a starter but he will end up playing about 2300 minutes).....

I've got an interesting result when simulating D. Wesley - he seemed to be as good as Stockton, but than I saw that he was one of the best shooting and rebounding PGs that season and I understood why he was so good.

high ScFG% (one of the very best 3pt shooting PGs), high rebounding, also good steals, good defender, what's not to like...

That left me wondering about some present players. How do B. Davis, B. Knight or T. J. Ford end up in simulation than compared to guys like S. Nash or C. Billups?

knight and ford two of the worst starting PGs, nash and billups two of the best...

Also I wanted to ask how does the software calculate game pace. I've looked at some teams and they were quite close to the real life.

attention to detail?...

And how did you rate the players defensively before 82games.com data was available? Using some stats or just by watching players?

watch everything you can, read everything you can, take notes. more subjective than now, so god bless roland beech and his efforts for making our current analytical lives easier...

[AGBCZAL]

I've got a few interesting results when simulating PGs. B. J. Armstrong, D. Harper, B. Price and D. Barros ended up being inthe top 11. That seems to show that Scfg% is one of the most important stat to look upon as all these had high Scfg. On the other hand, R. Strickland, J. Kidd ended up average. But a player like T. Hardaway ended up pretty poor. I've than simulated him from 96-97 season and he got about 4 wins more, so it seems that 95-96 season was not good for him.

Anyway, I've simulated PGs on Utah, Toronto and New York (as they all had PG that played 36 min in substitution pattern) and it seems that some players can look better on different teams. So I was wondering what are biggest difference in wins than compared to the best or the worst player in that position when playing different teams (I mean that for example M. Jackson was about 1.3 wins better in NY than in Tor if compared to the worst PG I've found there - B. Shaw).

Also I would like to ask if there is any difference which periods of time a player plays in substitution pattern or the only thing that matters is playing time in your sumulation.

By the way, are you going to add a few more seasons to your simulation wihich you put in APBR files?

[Mike G]

[AGBCZAL]

When simulating Kerr instead of Stockton I get Utah win as much as they do with Stockton. Really, an interesting case, but Kerr gets about 12ppg with TS% of 0.664 and only 1 to per game. He also gets only 3.4 assists but with 1.1 steal per game and thats more steals than turnovers. When simulating Stockton gets about 15ppg with TS% of 0.646, he gets 11.4apg and 1.7 steals but 3 TOs per game. However, K. Malone and J. Hornacek get 27.6 and 15.4 ppg with Stockton but 28.6 and 15.9 with Kerr. Kerr is rated as +2% in defense and Stockton as -3.5%.

[Mike G]

So in simulation, in spite of Stockton's greatly superior defense, the Jazz win just as much with Kerr at point. Main difference is that he takes a couple less shots per game, I'm guessing; leading to Malone and Hornacek shooting a bit more.

And no apparent reduction in scoring efficiencies despite having to create shots without assists. I wonder if this is realistic. Really, I suspect that it is not. Perhaps it's unrealistic to plug in a player like Kerr at PG, even though on the Bulls he was one of several who were classified as such, but without the traditional duties.

Kerr had more steals than turnovers overall in his 5 seasons in Chicago; but not at any other time in his career. This seems almost certainly to be a causal relationship, which did not transfer with Kerr to his subsequent stops in SA and Por. He also enjoyed his best shooting with the Bulls.
_________________
40% of all statistics are wrong.

[bchaikin]

I've got a few interesting results when simulating PGs. B. J. Armstrong, D. Harper, B. Price and D. Barros ended up being inthe top 11. That seems to show that Scfg% is one of the most important stat to look upon as all these had high Scfg.

not only overall shooting efficiency or ScFG% but also turnovers. 3 of the 4 you mentioned had not only high overall shooting but also low turnovers per touch (lower than the average PG that season)....

in harper's case he was also a very good defender...

...a player like T. Hardaway ended up pretty poor. I've than simulated him from 96-97 season and he got about 4 wins more, so it seems that 95-96 season was not good for him.

take a look at his stats 95-96 vs 96-97. if you extrapolate his turnovers to minutes played you'll see he had about 20% more turnovers per minute in 95-96 than in 96-97 (with similar touches/min)...

...what are biggest difference in wins than compared to the best or the worst player in that position when playing different teams.

that's alot of permutations to run, i don't know offhand. it all depends on what the player you are simulating brings to the team and what the team needs to generate more wins. i do remember running penny hardaway 95-96 on different teams and seeing a large difference of 6-7 wins over 82 games, but don't know if that's the most difference you'd see....

also as a team wins more games its hard to improve it much with another great player versus a poor team adding a great player...

Also I would like to ask if there is any difference which periods of time a player plays in substitution pattern or the only thing that matters is playing time in your simulation.

players touches/min are affected by the touches/min of the other players, and how often per touch they shoot, pass, etc. thus their statistical offensive production can change based on how their touches/min change based on who their teammates are....

also playing players more in the last 4 minutes of a quarter versus the first 4 minutes of a quarter will result in the former getting more FTAs as teams are in the bonus more often towards the end of a quarter...

By the way, are you going to add a few more seasons to your simulation which you put in APBR files?

not sure, maybe at some point..

And no apparent reduction in scoring efficiencies despite having to create shots without assists. I wonder if this is realistic. Really, I suspect that it is not.

let's use the example from above - mookie blaylock played the same amount of minutes in 95-96 as stockton did but with 1/2 the assists, on a team with similar pace - 90.9 vs 91.3 poss/48min. the hawks that season without blaylock got less assists per game than the jazz got without stockton (13.8 vs 14.9 ast/g). the players playing the most minutes on that hawks team other than blaylock were their SG and PF - steve smith and grant long. the players playing the most minutes on the jazz that season other than stockton were their SG and PF, jeff hornacek and karl malone...

the hawks won just 9 less games than the jazz that year. is there something in the makeup of smith and long that makes you believe they were better at creating their own shots than hornacek and malone? are you under the impression that malone and hornacek, if traded to the hawks, or if blaylock was traded to the jazz, would score less, or shoot worse, because they couldn't create their own shots with a PG that got 1/2 the assists that stockton did?...

[Mike G]

[bchaikin]

Hmmm, well Mookie was a pretty good PG.

hmmm, well steve was a pretty good shooter...

don't expect an exchange of dialogue to your questions if you refuse to respond in kind...

And no apparent reduction in scoring efficiencies despite having to create shots without assists. I wonder if this is realistic. Really, I suspect that it is not...

blaylock got 1/2 the assists stockton did in 95-96, the rest of the hawks less ast/g than the rest of the jazz. how would malone and hornacek do with blaylock as a PG?...

What I wondered was whether Steve Kerr could fill Stockton's shoes without the Jazz players suffering dropoffs in production or efficiency.

do you wonder the same about doc rivers of the 95-96 spurs?...

[metalloids]

Can you please post a link to the B-Ball simulator? I haven't heard of it, and can't find it. Thanks.

[Mike G]