Association Stats are a ROSTER issue

[elknavo]

Re: Association Stats are a ROSTER issue

Youalreadyknow:

Just to clarify I'm talking only about stats from simulated games in season or association mode, NOT actual gameplay.

I looked up that spreadsheet and applied the shot and touch tendencies to the entire Nets team. Deron Williams usually averages around 6.5 APG. Those changes seemed to get Deron to around 7 to 7.5 APG. I'm not sure if the touch tendency changes actually affected simmed stats at all, but the shot tendencies did. It's an improvement, but not nearly enough.

Another big reason for too few assists is far too many FT's being shot. This is mainly because role players get far too many FTA. Sim a season with the stock roster and you'll find that the lowest FTA team shot around as many FTA as the top FTA team in the NBA- the average team shoots about 500-1000 FTA more than they should, and correspondingly fewer FGA, so fewer assists. This happens not because the foul tendency is too high (fouls are around the correct range) but because Draw Foul tendency is too high for most role players.

So, by correcting Draw Foul tendencies and shot tendencies on the Nets, I managed to get Williams to around 8.4 APG. This at least starting to get somewhat reasonable, as it's sort of conceivable that Williams' assists might drop off some in Avery Johnson's slow paced offense. Of course the greatest effect on assists seem to be teammates-- any Mavericks PG will get decent assist numbers. I can't seem to figure out what it is about the mavs teammates that make Kidd get 12 APG while Nash can barely manage 7. I also haven't found any tendency other than shot tendency that actually seems to have a strong effect on assists.

And it's not just superstar PG's that lack assists, it's most players and teams. The top assist teams get around the right amount of assists, but the lower assist teams get way too few- around 15/game, while the lowest NBA team last season got 19. Most of the PG's with unrealistically low APG numbers seem to come from these low assist teams. As far as I can tell coach profiles, plabooks and offense ratings have little or no effect on this- player personnel seems to be the main factor.

Youalreadyknow, you seem to know a lot about this game's mechanics. If you know what it is that makes high APG possible please tell me- the numbers on that spreadsheet help the issue but double figure APG is still far away for Deron even though he's managed it for 4 consecutive seasons. And I can't seem to get Nash to even hit 8 APG whatever I do, after he led the league.

[Drewski]

Re: Association Stats are a ROSTER issue

Have to say I just simulated a team built around Dwight/CP3 (Magic), with a solid roster all around (Made it to the ECF many, many times. No ring unfortunately), and I also noticed as far as simming that PG stats really suffer. CP3 generally averaged 14-15 points and 6-7 assists per season. And I ran from 2011-2012 all the way through to 2018. Never a year when he was close to even 8 apg.

Is what it is, but no doubt PG's assists are much lower than they should be when simulated.

[ffaacc03]

Re: Association Stats are a ROSTER issue

Youalreadyknow:

Just to clarify I'm talking only about stats from simulated games in season or association mode, NOT actual gameplay.

I looked up that spreadsheet and applied the shot and touch tendencies to the entire Nets team. Deron Williams usually averages around 6.5 APG. Those changes seemed to get Deron to around 7 to 7.5 APG. I'm not sure if the touch tendency changes actually affected simmed stats at all, but the shot tendencies did. It's an improvement, but not nearly enough.

Another big reason for too few assists is far too many FT's being shot. This is mainly because role players get far too many FTA. Sim a season with the stock roster and you'll find that the lowest FTA team shot around as many FTA as the top FTA team in the NBA- the average team shoots about 500-1000 FTA more than they should, and correspondingly fewer FGA, so fewer assists. This happens not because the foul tendency is too high (fouls are around the correct range) but because Draw Foul tendency is too high for most role players.

So, by correcting Draw Foul tendencies and shot tendencies on the Nets, I managed to get Williams to around 8.4 APG. This at least starting to get somewhat reasonable, as it's sort of conceivable that Williams' assists might drop off some in Avery Johnson's slow paced offense. Of course the greatest effect on assists seem to be teammates-- any Mavericks PG will get decent assist numbers. I can't seem to figure out what it is about the mavs teammates that make Kidd get 12 APG while Nash can barely manage 7. I also haven't found any tendency other than shot tendency that actually seems to have a strong effect on assists.

And it's not just superstar PG's that lack assists, it's most players and teams. The top assist teams get around the right amount of assists, but the lower assist teams get way too few- around 15/game, while the lowest NBA team last season got 19. Most of the PG's with unrealistically low APG numbers seem to come from these low assist teams. As far as I can tell coach profiles, plabooks and offense ratings have little or no effect on this- player personnel seems to be the main factor.

Youalreadyknow, you seem to know a lot about this game's mechanics. If you know what it is that makes high APG possible please tell me- the numbers on that spreadsheet help the issue but double figure APG is still far away for Deron even though he's managed it for 4 consecutive seasons. And I can't seem to get Nash to even hit 8 APG whatever I do, after he led the league.

Elk, did you edited the whole roster ? all teams and players ? ... cause only doing so for 1 or 2 or 5 or 10 teams do not provide the intended result ... if FGs are too few, then all other aspects become too few, like steals, rebounds, assists, blocks ... also, if teams FG%s are low (like they seem to be even after edits) they would also cause fewer assists.

You are right about some of the things mentioned not related to the sim engine, but the things that do, have to be edited for all players and teams in order to see a noticeable effect ... so having a worthy "sim stats roster" (as well as a sim gameplay roster), requires for an extensive (and I mean extensive) editing and testing. I believe that this year, there are flaws within the sim stats engine that wont be entirely overcome by roster edits, thus at this point is just speculation.

Having said that, wait for youalreadyknow work to be done, to see up to what extend has he been able to address such issues, he seems to have studied the game variables and is taking his time (which is a great sign) so we should let him finish so he can then share his findings with us.

[elknavo]

Re: Association Stats are a ROSTER issue

Touches translated to simmed stats in 2K11. The reason I know is because my Touches (based on usage) always tranferred over to the Metric Stats category of Usage. So I am not sure if this has changed in 2K12, but I know it worked in 2K11.

And as far as blocks, that is incorrect. I have seen Dwight average 2.8 blocks in a season during my 2K11 Seasons. Not sure where that information came from.

Just simmed a season with the Clippers, most recent official 2k roster. Blake Griffin (96 Touches tendency) has the following stat line

33.1 MPG
18.7 PPG
1.2 APG
47% FG
62% FT

Same as before, and Blake Griffin's touch tendency is lowered to 0
32.6 MPG
19.1 PPG
1.2 APG
45% FG
67% FT

Note the ridiculously low assists despite decent pass rating and 3.8 APG last year. As with Chris Paul, 2k's sim engine thinks that if you are on the Clippers you are not allowed to get any assists. So yeah, unless Blake Griffin is so good he can put up 19 PPG without touching the ball, this has to prove that the touch tendency means nothing for sim stats.

As for the blocks thing, if you saw Dwight do 2.8 BPG then that is probably the highest simmed season BPG ever recorded since 2k11. Overall though blocks are too low. It's not nearly as horrible as the assists issue where the wrong guy leads the league every year and half the best players in the category do 50% worse than they should, but it's an issue. Guys like Javale McGee block more than 3 shots per 36 minutes, and that is simply flat out impossible in 2k sim stats. And for a laugh try simming a season with the '94 Nuggets and watching Mutombo (4.1 BPG that year, 99 BLK) usually come in under 2 BPG. It's broken. Not as bad as assists, but it's broken.

[elknavo]

Re: Association Stats are a ROSTER issue

Elk, did you edited the whole roster ? all teams and players ? ... cause only doing so for 1 or 2 or 5 or 10 teams do not provide the intended result ... if FGs are too few, then all other aspects become too few, like steals, rebounds, assists, blocks ... also, if teams FG%s are low (like they seem to be even after edits) they would also cause fewer assists.

You are right about some of the things mentioned not related to the sim engine, but the things that do, have to be edited for all players and teams in order to see a noticeable effect ... so having a worthy "sim stats roster" (as well as a sim gameplay roster), requires for an extensive (and I mean extensive) editing and testing. I believe that this year, there are flaws within the sim stats engine that wont be entirely overcome by roster edits, thus at this point is just speculation.

Having said that, wait for youalreadyknow work to be done, to see up to what extend has he been able to address such issues, he seems to have studied the game variables and is taking his time (which is a great sign) so we should let him finish so he can then share his findings with us.

ffac, I did not edit the whole roster. As I said I edited the Nets and was successful at getting them to approximately the correct amount of field goals made. Even if you do that, they just seem to have ridiculously low numbers of assists, regardless of playbook or coach profile. Editing the rest of the league should not make much of a difference as to the Nets' assists. It's clear that it is possible for teams/players to have high numbers of assists without league-wide edits-- after all, Jason Kidd pulls his career high in APG every single year in 2k's sim engine with default rosters. Michael Conley always does 2 APG more than he ought to. So we know for sure that it is possible- and until I figure out what mechanism actually does control the APG, what point is there to going to the effort of editing the whole league?

What unsolvable issues are there with the sim stats engine? The only one I can identify that is clearly unsolvable is blocked shots, and that one isn't really as bad as the assist thing. Blocked shots have declined in recent years in the real NBA, and nobody's done 4 BPG in a long time. Everything other than assists is, as far as I can tell, pretty accurate or easily editable to be accurate. And I know there's a way to fix the assist issue, which is why I keep racking my brain trying to figure out how.

Youalreadyknow seems to have the secret, but then again he said the key was touches tendency, and I think I just proved that touches tendency has no effect on sim stats at all, so perhaps he doesn't know as much as it seems.

[elknavo]

Re: Association Stats are a ROSTER issue

Have to say I just simulated a team built around Dwight/CP3 (Magic), with a solid roster all around (Made it to the ECF many, many times. No ring unfortunately), and I also noticed as far as simming that PG stats really suffer. CP3 generally averaged 14-15 points and 6-7 assists per season. And I ran from 2011-2012 all the way through to 2018. Never a year when he was close to even 8 apg.

Is what it is, but no doubt PG's assists are much lower than they should be when simulated.

Drewski, this is true but take note that this is only true for some PG's. CP3, Deron Williams, Steve Nash, and Westbrook will always have career low APG. But it's not the case that all PGs will suffer from it. Check out Mike Conley or Jason Kidd- Conley does 2 APG more than his career high, and Kidd leads the league every year with his career high APG while he's nearly 40 years old! ROndo also does just fine with assists.

Trade any of those underperforming PGs to the mavericks or grizzlies and they'll have decent APG. I've ruled out playbooks and coach ratings and coach profiles as the reason- swap the whole roster of the mavs and clippers and you'll still see CP3 with 6.5 APG and Kidd with 11. So there must be something about the ratings of the SG/SF/PF/C of the Mavericks and Grizzlies that enable PGs to get high assist numbers, and something about the ratings of the Clippers/Nets/Thunder/Suns that makes their PGs forget how to pass the ball. If we can figure out what causes this we can fix it.

From my testing the following things help significantly:
1. Draw Foul tendency:
The sim stats engine gives players, especially bench players, way too many FTA. There are a few too many fouls but a huge excess of FTA, so we fix this with the draw foul tendency.

2. Shot Tendency
This works in 2 ways: First, your PG's shot tendency is inversely related to APG. Cut CP3 down to 0 shot tendency, and he goes from 6.5 APG or so to around 10, which would be perfect if it didnt also make him average 2 PPG. Part of why Jason Kidd's assists are too high is that his shot tendency is lower than it should be so he takes about half as many shots as he should. Most star PGs have a shot tendency higher than they should, so this helps some.
Secondly, your temmates' shot tendencies are often ridiculously wrong in the default roster, especially role players. For example DeAndre Jordan will take about 1/3 as many FGA as he did lat season. Correcting these will give the PG more teammates to give assists to.

Using these methods I've managed to get CP3/Deron Williams around 8.4 APG. But there must be other factors I haven't figure out yet- Michael Conley's teammates have decently high shot tendencies and his isn't too high, and the Grizzlies do shoot way too many FTA in the default roster- so why does he overperform his pass rating and stats by so much and get 8.5-9 APG? If we can figure out what it is about the grizzlies, that might be the key.

[qpc123]

Re: Association Stats are a ROSTER issue

Correct me if I'm wrong, but I don't think pass rating has a big(or any) effect on assists per game. The 'Pass' rating effects things like accuracy and the ability to throw certain types passes not how often a player passes, but how well they do it(granted better passes will lead to more assists). Shot tendency and attack/pass out will have a greater effect in generating APG IMO.

[elknavo]

Re: Association Stats are a ROSTER issue

simulated season, Mavericks, most recent official 2k roster

Kidd (93 Pass rating) averages 11.3 APG in 33 mpg

Kidd (0 Pass rating) averages 2.1 APG in 33 mpg

You are wrong, so you have been corrected. The Pass rating is one of the primary factors- I just need to figure out the other ones. It can't be attack/pass out because Nash has very high pass out tendency yet has like 40% lower APG in this game than real life.

[youALREADYknow]

Re: Association Stats are a ROSTER issue

Correct me if I'm wrong, but I don't think pass rating has a big(or any) effect on assists per game. The 'Pass' rating effects things like accuracy and the ability to throw certain types passes not how often a player passes, but how well they do it(granted better passes will lead to more assists). Shot tendency and attack/pass out will have a greater effect in generating APG IMO.

The Pass rating (along with Offensive Awareness) sets the "ceiling" on a player's ability to get assists. The combination of Touches and how those touches are used (Shot Tendency for player and Touches across the rest of the team) determines how close to that ceiling the player gets.

[youALREADYknow]

Re: Association Stats are a ROSTER issue

From my testing the following things help significantly:
1. Draw Foul tendency:
The sim stats engine gives players, especially bench players, way too many FTA. There are a few too many fouls but a huge excess of FTA, so we fix this with the draw foul tendency.

2. Shot Tendency
This works in 2 ways: First, your PG's shot tendency is inversely related to APG. Cut CP3 down to 0 shot tendency, and he goes from 6.5 APG or so to around 10, which would be perfect if it didnt also make him average 2 PPG. Part of why Jason Kidd's assists are too high is that his shot tendency is lower than it should be so he takes about half as many shots as he should. Most star PGs have a shot tendency higher than they should, so this helps some.
Secondly, your temmates' shot tendencies are often ridiculously wrong in the default roster, especially role players. For example DeAndre Jordan will take about 1/3 as many FGA as he did lat season. Correcting these will give the PG more teammates to give assists to.

Using these methods I've managed to get CP3/Deron Williams around 8.4 APG. But there must be other factors I haven't figure out yet- Michael Conley's teammates have decently high shot tendencies and his isn't too high, and the Grizzlies do shoot way too many FTA in the default roster- so why does he overperform his pass rating and stats by so much and get 8.5-9 APG? If we can figure out what it is about the grizzlies, that might be the key.

My entire mindset and philosophy is based on the assumption that every single aspect of the roster has been edited to reflect the real world. That includes tendencies such as Draw Foul, Overall Shot Tendency, and individual Shot Location tendencies. Everything works together and defaults are simply not acceptable IMO. Your points, especially about Shot Tendencies for role players, are spot on.

Second, there are indeed certain teams that produce higher assist totals for the starting PG on their roster. I have not conclusively found out what causes the huge effect on the stats but I suspect it to be related to the Coach Profiles and the quality of the primary scorers on the team. When the PG is the best player (OVR) on the team, then that player typically ends up with fewer assists than their ratings should allow. It takes a lot more effort to get them to a more realistic assist rate. Hopefully you or someone here can find further correlations so we can all learn from it and make our sim stats even better. I'm more focused on gameplay at the moment than sim stats but will share anything that could help.

Also it makes zero sense to target per game numbers without taking into account the per minute totals. There are only a few players that have taken a lot of manipulation to get close to their rate adjusted averages after using real world data to come up with league wide tendencies and attributes.

Has anyone done research on whether play types have an impact on sim stats? That's one area that I haven't explored yet.

Another thing I've noticed is that Season Mode produces slightly more realistic sim stats than Association Mode with all features turned on. I wonder what features within Association Mode are throwing stats in a different direction.

[bedwardsroy19]

Re: Association Stats are a ROSTER issue

Another thing I've noticed is that Season Mode produces slightly more realistic sim stats than Association Mode with all features turned on. I wonder what features within Association Mode are throwing stats in a different direction.

I noticed this especially. I was trying to figure this out myself.

[youALREADYknow]

Re: Association Stats are a ROSTER issue

ffac, I did not edit the whole roster. As I said I edited the Nets and was successful at getting them to approximately the correct amount of field goals made. Even if you do that, they just seem to have ridiculously low numbers of assists, regardless of playbook or coach profile. Editing the rest of the league should not make much of a difference as to the Nets' assists. It's clear that it is possible for teams/players to have high numbers of assists without league-wide edits-- after all, Jason Kidd pulls his career high in APG every single year in 2k's sim engine with default rosters. Michael Conley always does 2 APG more than he ought to. So we know for sure that it is possible- and until I figure out what mechanism actually does control the APG, what point is there to going to the effort of editing the whole league?

What unsolvable issues are there with the sim stats engine? The only one I can identify that is clearly unsolvable is blocked shots, and that one isn't really as bad as the assist thing. Blocked shots have declined in recent years in the real NBA, and nobody's done 4 BPG in a long time. Everything other than assists is, as far as I can tell, pretty accurate or easily editable to be accurate. And I know there's a way to fix the assist issue, which is why I keep racking my brain trying to figure out how.

Youalreadyknow seems to have the secret, but then again he said the key was touches tendency, and I think I just proved that touches tendency has no effect on sim stats at all, so perhaps he doesn't know as much as it seems.

It's possible that the Shot Tendency changes are responsible for the huge changes I've seen from the default rosters since I edit Touches and Shot Tendency at the same time. If so, then I apologize and stand corrected.

It is clear though that assists are a statistic that we have control over. If only Blocks were the same way. Another less glaring issue is that most passing PF/C end up with fewer assists than would be expected and it's largely due to the horrible Pass ratings that are given to them by default.

I wonder if these low pass ratings by non-PG have any indirect impact on team stats but haven't had time to do a proper test.

[DukeC]

Re: Association Stats are a ROSTER issue

It's possible that the Shot Tendency changes are responsible for the huge changes I've seen from the default rosters since I edit Touches and Shot Tendency at the same time. If so, then I apologize and stand corrected.

It is clear though that assists are a statistic that we have control over. If only Blocks were the same way. Another less glaring issue is that most passing PF/C end up with fewer assists than would be expected and it's largely due to the horrible Pass ratings that are given to them by default.

I wonder if these low pass ratings by non-PG have any indirect impact on team stats but haven't had time to do a proper test.

Where do you define the "minimum" for the Pass rating? Around 60-65?

[elknavo]

Re: Association Stats are a ROSTER issue

The Pass rating (along with Offensive Awareness) sets the "ceiling" on a player's ability to get assists. The combination of Touches and how those touches are used (Shot Tendency for player and Touches across the rest of the team) determines how close to that ceiling the player gets.

Did you see my post recently in this thread where I show that Blake Griffin gets pretty much the same offensive production with a Touches tendency of 0 as with his normal Touches tendency of 96? You've mentioned it a number of times and I'm sure it's vitally important for gameplay, but for simulated stats, unless you can show me some compelling evidence to the contrary, I must conclude that it has no effect at all. It certainly hasn't seemed to affect the APG issue either way, whatever Touches changes I've made to PGs or their supporting casts.

I think your observation that Pass rating sets the ceiling is mostly correct, but how do you explain Mike Conley? In a simmed season I've seen his assists will be around 8.5-8.9, which is usually good for top 5 in the NBA and certainly seems to overachieve compared to his 82 rating. Any idea what's with that?

[youALREADYknow]

Re: Association Stats are a ROSTER issue

Did you see my post recently in this thread where I show that Blake Griffin gets pretty much the same offensive production with a Touches tendency of 0 as with his normal Touches tendency of 96? You've mentioned it a number of times and I'm sure it's vitally important for gameplay, but for simulated stats, unless you can show me some compelling evidence to the contrary, I must conclude that it has no effect at all. It certainly hasn't seemed to affect the APG issue either way, whatever Touches changes I've made to PGs or their supporting casts.

I think your observation that Pass rating sets the ceiling is mostly correct, but how do you explain Mike Conley? In a simmed season I've seen his assists will be around 8.5-8.9, which is usually good for top 5 in the NBA and certainly seems to overachieve compared to his 82 rating. Any idea what's with that?

82 Pass rating is still elite compared to most of the league in this game though. I've never seen Conley above 9.5 APG in all of my sims so far while guys like Rondo can get up around 11 APG. I can try to lower his Pass rating and run a test to see if he still outperforms by such a wide margin.

I'm leaning towards the ratings of other players on the team being the most important driver for reaching the assist potential for the PG or primary ball-handler. Another great example is LeBron. He clearly has all of the tools and supporting cast to reach very high assist totals but since he's the highest rated player on the team his assist totals suffer. Paul for NO is the highest rated player on the team. Williams for NJ is the highest rated player on the team. Nash for PHX is the highest rated player on the team.

Rondo, Conley, Kidd, etc are not the top rated players on their teams in most rosters and that's why I think they are getting the assists.

Another great test would be to lower these player's defensive attributes in order to lower the OVR to make them the 3rd or 4th best player on the team and see whether their assist totals improve.

If you're willing to do some of these tests too then I'm sure we can find this out in just a few hours. I can also test Touches on PG's to see if the sim stats are different.