|
|
Post by Ian Noble on Mar 3, 2018 19:02:35 GMT
Testing Player RatingsFor the first time ever I conducted a thorough review of five individual player ratings on the sim engine the other day. Dunk, Speed, Quick, Jump and Strength The conclusion: none of these ratings make any difference to players performances in the sim!! The results are here and here. This is quite a big discovery. It's bad because it's taken me this long to test it (my excuse is: how would I have known to test it?). It's bad because it means athletically gifted players are penalised by the sim and always have been ever since EA Sports created it! It's good because now it's tested we can do something about it. It's also not as bad as I am making out here because the negative effect on athletically gifted players matters only in so much that their athletic ratings deviate from the league average, and even with an athletic guy like Lebron James that's only 84 rating points out of a total that's over 1500. It explains why Dirk Nowitzki was a sim-god for all those years - low athleticism, high skilled players are few and far between but unicorns like Dirk are perfect sim players. It will also explain why high athleticism, low skill players (e.g. Josh Smith) suffered the most. Me and Walt are discussing how this will affect future rating changes because it's quite hard to get your head around the implications! Walt already assigns a league average 70 'Hardyness' rating points to every new change because Hardyness affects injuries and we don't have injuries, but the rating nevertheless contributes to each player's overall rating. It seems logical going forward to also revert these non-sim ratings (Dunk, Speed, Quick, Jump and Strength) to their league averages, which is usually somewhere around 65 to 70. But how this affects what we consider to be a high player rating or a low player rating going forward is the most difficult concept, because the bar might possibly have changed, or has it? It's something I'm trying to wrap my head around.
|
|
|
|
Post by Andrei Kirilenko on Mar 3, 2018 19:33:12 GMT
Looking closer, doesn’t it appear that the physical attributes do possibly have an impact on rebounding? Simmons rebounded 1.5 more average with 99 ratings, and Embiid rebounded below with 0 ratings.
|
|
|
|
Post by Walt Frazier on Mar 3, 2018 20:18:50 GMT
Kanter avg 1.5 more rpg with maxed physicals. He also avg more steals and blocks. Maybe physical has an effect more than we first thought, just not on points (which more or less makes sense).
As the rating change guy, I'd say equalizing dunk across the board makes sense, and then for the physicals, maybe I should shrink the range a little bit? And/or prioritize other areas before their physicals, which I sort of do already. Some combo of that, likely.
|
|
|
|
Post by Ian Noble on Mar 3, 2018 20:32:28 GMT
Let me do more testing first I think, just to make sure.
|
|
|
|
Post by Hanamichi Sakuragi on Mar 4, 2018 0:07:23 GMT
can you explain first, what you did?
|
|
|
|
Post by Walt Frazier on Mar 4, 2018 1:39:38 GMT
can you explain first, what you did? He simmed an entire season with players' ratings adjusted. So he put, for example, the one time - Embiid's Dunk rating at a 1 and then simmed like 16 seasons to see what the average was, then simmed 16 seasons with Embiid at a 99 dunk. He did the same type of thing with Kanter, George Hill, Ben Simmons, and then also ran similar tests with Strength, Quickness, Speed, and Jump. |
|
|
|
Post by Hanamichi Sakuragi on Mar 4, 2018 4:11:32 GMT
How about the W-L record Ian?
There is a chance that the physicals are more of a team aggregate thing in deciding which team will win.
|
|
|
|
Post by Ian Noble on Mar 4, 2018 8:46:39 GMT
How about the W-L record Ian? There is a chance that the physicals are more of a team aggregate thing in deciding which team will win. W-L records were not affected either. I should have recorded them to be more thorough I suppose. |
|
|
|
Post by Hanamichi Sakuragi on Mar 4, 2018 9:28:18 GMT
How about the W-L record Ian? There is a chance that the physicals are more of a team aggregate thing in deciding which team will win. W-L records were not affected either. I should have recorded them to be more thorough I suppose. I think, what you should try is to first simulate without adjusting anything. Then the adjusted once. Then compare... If you can record team stats, that will be better.  I just hope we do not find out huge problem.. |
|
|
|
Post by Ian Noble on Mar 6, 2018 20:54:48 GMT
W-L records were not affected either. I should have recorded them to be more thorough I suppose. I think, what you should try is to first simulate without adjusting anything. Then the adjusted once. Then compare... If you can record team stats, that will be better. I just hope we do not find out huge problem.. Hana you were absolutely correct to suggest this! I was premature in my analysis which could've been rather disastrous. How embarrassing!  Here is the spreadsheet for 11 seasons of Strength testing: link |
|
|
|
Post by Ian Noble on Mar 6, 2018 21:26:21 GMT
I'm doing the same for all physical stats now: speed, quick, jump, strength and also fatigue. I wont have the work finished until tomorrow though.
|
|
|
|
Post by Brian Scalabrine on Mar 6, 2018 22:29:29 GMT
This kind of stuff is why you're a top notch commish Ian
|
|
|
|
Post by Ian Noble on Mar 7, 2018 12:00:22 GMT
I've spent a few hours putting this together!  Here are the results of testing All Physical Ratings. I'm not going to look at them right now because I need a break, but I'll come back later and take a proper look. If any of it doesn't make sense let me know. |
|
|
|
Post by Hanamichi Sakuragi on Mar 7, 2018 12:56:19 GMT
Physicals changes the performance of the team.. Did not finish it all.. but in the quick, you will see that the bigs average more points than the guards.
But, in the 1 Quick, the bigs tapout at 18.9 or above, 4 times.. The In 99 Quick, they reach it once. The guards reached 13.0 or more, once. In 99 quick, did it thrice.
It could be about the sample, but I think that is conclusive already.
Something changes with the physical, and with the way I am looking at it, it is really an aggregate thing.
|
|
|
|
Post by Ian Noble on Mar 7, 2018 13:10:42 GMT
Speed
??
Quick
Helps stealing quite a bit
Jump
Helps rebounding a lot
Helps blocking a little
Strength
Helps scoring, blocking, rebounding
Decreases turnovers
Causes more fouls
Fatigue
Affects everything because minutes decrease
|
|
|
|
Post by Ian Noble on Mar 7, 2018 18:25:02 GMT
 there is an inverse 20 win disparity between 1 and 99 Strength. 1 Strength = 64.6 wins on average over 5 seasons 99 Strength = 44.6 wins wtf |
|
|
|
Post by Ian Noble on Mar 7, 2018 20:14:32 GMT
there is an inverse 20 win disparity between 1 and 99 Strength. 1 Strength = 64.6 wins on average over 5 seasons 99 Strength = 44.6 wins wtf I backed this up with another 10 simmed seasons on 10 teams, so we have 100 sets of results. This is batshit insane. ResultsIncreased Strength = Decreased Game WinningW.T.F. Even though Increased Strength = Increased scoring, blocking, rebounding and decreased turnovers (albeit more fouls) This is by far the biggest WTF, shitloads of stuff must rely on the Strength rating but not necessarily manifest in player stats... edit: The problem I feel like we're currently faced with - do we want to know these things?! What constructive conclusions can we draw with this information?! Gah! edit2: One variable to rule them all, one variable to find them, one variable to bring them all and in the darkness confused the hell out of them! |
|
|
|
Post by Ian Noble on Mar 7, 2018 20:54:58 GMT
Okay I've had a moment to collect my thoughts about this now and, even though there is a negative correlation,
1) in reality it's not going to be so extreme in the actual sims because Strength is distributed in a quite varied fashion (whereas in the example above an entire roster was set to either 1 or 99 throughout)
2) the best teams will tend to have stronger players? Perhaps? Maybe the negative affect on wins is counter-balanced when Strength is not uniformly rated as high as 99? Although having a big strong point guard like Ben Simmons might come back to haunt me...
|
|
|
|
Post by Andrei Kirilenko on Mar 7, 2018 21:12:08 GMT
No wonder Jeremiah keeps winning games. His team is weak.
|
|
|
|
Post by Shaquille O'Neal on Mar 7, 2018 21:35:17 GMT
there is an inverse 20 win disparity between 1 and 99 Strength. 1 Strength = 64.6 wins on average over 5 seasons 99 Strength = 44.6 wins wtf I backed this up with another 10 simmed seasons on 10 teams, so we have 100 sets of results. This is batshit insane. ResultsIncreased Strength = Decreased Game WinningW.T.F. Even though Increased Strength = Increased scoring, blocking, rebounding and decreased turnovers (albeit more fouls) This is by far the biggest WTF, shitloads of stuff must rely on the Strength rating but not necessarily manifest in player stats... edit: The problem I feel like we're currently faced with - do we want to know these things?! What constructive conclusions can we draw with this information?! Gah! edit2: One variable to rule them all, one variable to find them, one variable to bring them all and in the darkness confused the hell out of them! Does it mean more muscles = less basketball iq? 😂 |
|
|
|
Post by Ian Noble on Mar 7, 2018 22:08:43 GMT
D5 Starting-5s Ranked by Avg Strength Rating
75.4 Denver Nuggets
73 Minnesota Timberwolves
72.6 Boston Celtics
69.8 Los Angeles Clippers
69.4 Golden State Warriors
68.8 Charlotte Hornets
68.8 Indiana Pacers
68.4 Houston Rockets
68.4 New York Knicks
68 Miami Heat
67 Brooklyn Nets
67 Cleveland Cavaliers
67 Toronto Raptors
66.6 Utah Jazz
66 Orlando Magic
65.6 Chicago Bulls
64.8 Memphis Grizzlies
64.4 Philadelphia 76ers
64.4 Portland Trailblazers
64.2 Milwaukee Bucks
63.8 San Antonio Spurs
62.8 Sacramento Kings
62.4 Washington Wizards
62.2 New Orleans Pelicans
62 Detroit Pistons
61.8 Oklahoma City Thunder
61.4 Atlanta Hawks
58.2 Los Angeles Lakers
55.6 Dallas Mavericks
53.6 Phoenix Suns
D5 Full Rosters Ranked by Avg Strength Rating
65.86 San Antonio Spurs
65.64 Utah Jazz
65.54 Los Angeles Clippers
65.23 Minnesota Timberwolves
64.71 Boston Celtics
63.92 Chicago Bulls
63.38 Denver Nuggets
63.25 Philadelphia 76ers
62.92 Orlando Magic
62.58 Miami Heat
62.46 Toronto Raptors
62.14 Charlotte Hornets
61.93 Dallas Mavericks
61.54 Milwaukee Bucks
61.20 Brooklyn Nets
61.07 Golden State Warriors
60.86 Indiana Pacers
60.50 Washington Wizards
60.25 Portland Trailblazers
59.69 New York Knicks
59.67 Detroit Pistons
59.23 Houston Rockets
58.77 Cleveland Cavaliers
58.54 Memphis Grizzlies
58.00 Sacramento Kings
56.29 Atlanta Hawks
55.64 Los Angeles Lakers
55.64 Oklahoma City Thunder
55.10 New Orleans Pelicans
53.42 Phoenix Suns
|
|
|
|
Post by Alex English on Mar 8, 2018 1:35:52 GMT
Sooo higher strength ratings make your team lose more games... and my starting lineup has highest average strength... I think we've found one big reason as to why my team has historically underperformed.
The fact that strength even has a negative correlation with wins makes no sense at all. What's going on with the sim to cause this?
|
|
|
|
Post by Ian Noble on Mar 8, 2018 10:08:59 GMT
Sooo higher strength ratings make your team lose more games... and my starting lineup has highest average strength... I think we've found one big reason as to why my team has historically underperformed. The fact that strength even has a negative correlation with wins makes no sense at all. What's going on with the sim to cause this? There's a bigger disparity when a whole team has 1 Strength compared to when a whole team has 99 Strength. On my test database Phoenix's entire roster has a 52.6 average Strength rating, with a 53.6 rating for the starting 5. Denver's has 64.5 for the full roster and 70.8 for the starting 5. Denver - Ish Smith, James Harden, Tobias Harris, Lamarcus Aldridge, Brook Lopez Phoenix - Tyler Ulis, Caris LeVert, Kyle Kuzma, Trey Lyles, Henry Ellenson If I swap only the Strength ratings for both teams at both extremes of the D5 spectrum we get this: Wins per SeasonPhoenix (with Denver's Strength): 7 - 14 - 10 - 11 - 13 = 11 wins average Denver (with Phoenix's Strength): 64 - 55 - 55 - 61 - 67 = 60.4 wins average Phoenix (with Phoenix's Strength): 12 - 16 - 20 - 13 - 18 - 19 = 16.3 wins average Denver (with Denver's Strength): 61 - 55 - 54 - 53 - 55 - 57 = 55.8 wins average At maximum extent it's a 5 win disparity at either extreme end of D5, which is a big concern. How do we minimise this effect when making rating changes? Scale-down the Strength rating on strong players, scale-up the Strength rating on weak players? |
|
|
|
Post by Ian Noble on Mar 8, 2018 10:24:09 GMT
The fact that strength even has a negative correlation with wins makes no sense at all. What's going on with the sim to cause this? This is the biggest mystery to me. Higher Strength ratings lead to higher statistical output from players, in almost all categories, but less wins from teams... it doesn't make sense at all, but we need to come up with some method of mitigating the effects that isn't going to ruin Walt Frazier's life with complexity whenever he comes to make rating changes! |
|
|
|
Post by Hanamichi Sakuragi on Mar 8, 2018 11:19:01 GMT
Changing the way how to rate, will only benefit the players who will be adjusted after the change..
|
|
|
|
Post by Ian Noble on Mar 8, 2018 11:34:28 GMT
Changing the way how to rate, will only benefit the players who will be adjusted after the change.. That's not helpful  And technically it's not even true Take a player like Lebron James. 2nd highest Strength (90) in D5. Would you trade him away for someone weaker? Do Golden State suck? In Golden State's lineup both Pau (70) and Zaza (79) have above league average Strength and their roster is one of the Strongest in D5. Talent matters more. The Phoenix vs Denver approach I showed above yielded a 5 win disparity. But the skill of the players is a much more important factor in success, as the results show. There may be a 5 win disparity but there's a gulf of 40 wins between the two teams because of the other ratings that matter more. Future rating changes will have to take this into account. How that happens needs to be calculated somehow. |
|
|
|
Post by Andrei Kirilenko on Mar 8, 2018 12:16:33 GMT
I hate to be that guy... but this sample size is honestly too small to draw any sort of conclusions. You have Phoenix winning 7 games one season and 14 games the next. Denver wins 55 one season and 67 the next. with these huge variations, a sample of 5 doesn't actually give us enough data. How do we know which of these are outliers and which of these are normal? I don't want to create more work to figure this out, but I do think we shouldn't jump to conclusions.
|
|
|
|
Post by Ian Noble on Mar 8, 2018 12:48:31 GMT
I hate to be that guy... but this sample size is honestly too small to draw any sort of conclusions. You have Phoenix winning 7 games one season and 14 games the next. Denver wins 55 one season and 67 the next. with these huge variations, a sample of 5 doesn't actually give us enough data. How do we know which of these are outliers and which of these are normal? I don't want to create more work to figure this out, but I do think we shouldn't jump to conclusions. Good call I'll do more testing to start with! |
|
|
|
Post by Hanamichi Sakuragi on Mar 8, 2018 12:56:39 GMT
Changing the way how to rate, will only benefit the players who will be adjusted after the change.. That's not helpful And technically it's not even true Take a player like Lebron James. 2nd highest Strength (90) in D5. Would you trade him away for someone weaker? Do Golden State suck? In Golden State's lineup both Pau (70) and Zaza (79) have above league average Strength and their roster is one of the Strongest in D5. Talent matters more. The Phoenix vs Denver approach I showed above yielded a 5 win disparity. But the skill of the players is a much more important factor in success, as the results show. There may be a 5 win disparity but there's a gulf of 40 wins between the two teams because of the other ratings that matter more. Future rating changes will have to take this into account. How that happens needs to be calculated somehow. I don't know if you understood what I mean.. But, in essence, it will benefit the players that will have rating change after a change in rating players, in terms of REALISM.. But, further testing should be done. |
|
|
|
Post by Alex English on Mar 9, 2018 1:18:29 GMT
I hate to be that guy... but this sample size is honestly too small to draw any sort of conclusions. You have Phoenix winning 7 games one season and 14 games the next. Denver wins 55 one season and 67 the next. with these huge variations, a sample of 5 doesn't actually give us enough data. How do we know which of these are outliers and which of these are normal? I don't want to create more work to figure this out, but I do think we shouldn't jump to conclusions. True, though in every single case, the maximum win total from the high strength tests is lower than the minimum win total from the low strength tests. The sample size is definitely small, but it's looking pretty good like something is going on here, even if we can't say for sure how large the effect is. Ian Noble how much work is it for you to post team stats along with this? That might give us some clues. Like, maybe strength is correlated with fouls, so 99 strength teams send their opponent to the line 50 times per game while 1 strength teams never send their opponent to the line. It would be interesting to see if we could find any sort of relationships like that. |
|
Boston Celtics
Brooklyn Nets
Minnesota Timberwolves
Cleveland Cavaliers
New Orleans Pelicans
Miami Heat
Denver Nuggets
And technically it's not even true 