Sonicmage's SportsCrunch Ratings

[SonicMage]

Fellow O/S Roster makers! If you're interested in finding attributes and tendencies for your roster, current or historic, you're in luck.

What's SportsCrunch?

Its my fancy name for software I've been writing in my spare time the past few years, dedicated to generating ratings for sports games based on real sports statistics. It's a database that scrapes and compiles data from top stats websites like NBA.com, Basketball-reference.com, 82games.com, and others to have the most comprehensive compilation of statistics anywhere. It then uses this data to give you a ratings model of an NBA player given several different input variables. Want ratings for a player based on stats that span his whole career or just with a single team? Want Kobe's ratings based on all his playoff performances? All are possible.

Are you making your own roster?

Yes, for 2K18 I plan on doing a few rosters. Depending on time, I really hope to make a current 2017-2018 current rosters with weekly updates. Every week the ratings would be updated to reflect the stats from the previous week. There are also a couple of historic rosters I wish to tackle as well. My hope is still that some of the other brilliant roster makers get wind of this and make their own roster designs even better using these realistic ratings.

What are your formulas?

I'm totally open to discuss strategies I took coming up with these formulas, but straight up giving the formulas would not be completely useful. My system uses lookup tables to figure out the ratings from the generated values of the formula, so you'd need both in order to get the end result you'd want. Since I'm directly giving you almost all the ratings you'd possibly need to generate anyways, the work has been done for you.

Plus, modern stats websites are still lacking in reliability. Basketball-reference is still incomplete in some ways like shot types, and unfortunately NBA.com's stats for anything more than a couple years old are full of gaping holes, missing entire quarters worth of play-by-play data in some cases. So I painstakingly took on the task the past couple years to fill in those gaps using data from other sites to make sure everything was complete. For older stats, I trust my database more than any website out there.

Are you going to show us the ratings already?

Yup. You'll find the google doc spreadsheet at:

Sonicmage NBA 2K18 Ratings 2017-18 season
Sonicmage NBA 2K18 Ratings 1997-2017
Sonicmage NBA 2K18 Ratings 1973-1997
Sonicmage NBA 2K18 Ratings All-Time

Feedback is welcome. Let me know in the comments below if you plan on incorporating these into your rosters. If you're on PS4, I'd be glad to download them and give them a go! Enjoy!

Release Notes

10/27/2015: First release. Shot tendencies for 2000-01 season through 2014-15.
10/30/2015: Added Blocking and Shot Block Tendency. Added players dating back to the 1985-86 season.
11/01/2015: Corrected tendencies for Shot Under Basket, Shot Close, Shot Midrange, Shot Three, and Blocked Shot. Changed from linear to exponential scaling.
11/04/2015: Added Touches Tendency.
11/15/2015: Added Midrange and 3pt Shot location tendencies. Temporarily removed Block Attribute while scale is being fixed.
11/18/2015: Added improved Block attribute.
11/22/2015: Added Steal attribute. Added Pass Interception and On Ball Steal tendencies.
12/04/2015: Added Alley Oop Pass tendency.
12/05/2015: Bug fixes for shot location tendencies, and tweaks to shot distance rating tables.
12/21/2015: Added Standing Layup, Driving Layup, Standing Dunk, Driving Dunk, Alleyoop, and Putback tendencies.
12/31/2015: Added Close Shot location tendencies. Fixed Shot distance and Shot area definitions to match 2K's.
01/11/2016: Added Stepback Jumper, Use Glass, Floater, Post Hook Left, Post Hook Right, Post Fade Left, and Post Fade Right tendencies.
01/28/2016: Added Dish To Open Man tendency.
01/31/2016: Added tendencies going back to the 1973-74 season.
02/03/2016: Added tendencies for ABA players.
02/15/2016: Added 3pt, Shot, Dish To Open Man, and Block Shot tendencies for the 2015-16 season, along with full rosters for each team.
02/21/2016: Added Free Throw attribute. Weekly update of the 2015-16 season.
02/22/2016: Added Block, Steal, and Touches for the 2015-16 season.
02/29/2016: Weekly update of 2015-16 ratings.
03/07/2016: Weekly update of 2015-16 ratings.
03/16/2016: Weekly update of 2015-16 ratings. Added Offensive Rebound attribute.
03/28/2016: Weekly update of 2015-16 ratings. Added Defensive Rebound attribute. Added Eraser, Offensive Crasher, and Defensive Crasher badges.
04/04/2016: Weekly update of 2015-16 ratings. Added Pass Interception and On Ball Steal for the 2015-16 season. Added Lob City Passer, Interceptor and Pick Pocket badges.
04/26/2016: Added Standing/Driving Layup, Standing/Driving Dunk, Putback, Stepback Jumper, Floater, Use Glass, Alley-oop, Alley-oop Pass, and the Lob City Passer Badge for the 2015/16 season.
06/01/2016: Added Standing Shot Close, Mid, and Three rating columns. Added all Durability attributes.
06/02/2016: Added Bank Is Open, Post Hook Specialist, Corner Specialist, Fade Ace, and Lob City Finisher badges
07/23/2016: Added Standing Layup attributes. Added attributes for the 2015-16 season.
07/24/2016: Added Post Hook and Post Fadeaway attributes. Added some tendencies for the 2015-16 season.
08/19/2016: Added Driving Layup, Moving Shot Close, Moving Shot Midrange, and Moving Shot Three attributes. Added Tear Dropper and Dimer badges.
09/19/2016: Changed spreadsheet to represent ratings for 2K17 that match those already completed for 2K16. Added Passing Vision attribute.
09/20/2016: Added Shot Close attribute.
10/07/2016: Re-adjusted shooting rating scales. Adjusted badge scales, adding Hall of Fame level. Added Putback King badge.
10/20/2016: Added extended summary support for reduction of N/A's when surrounding season stats are available.
10/25/2016: Added foul tendency.
10/28/2016: Added Shoot From Post & Roll Vs Pop tendencies.
11/07/2016: Added first player ratings for 2016-17 season, for games through 11/03.
11/09/2016: Added shooting Hot Zones. More 2016-17 ratings, for games through 11/07.
11/30/2016: Shot Close readjusted to create correct percentages in sim stats.
12/23/2016: Standing Layup and Driving Layup have been adjusted for correct percentages in sim stats.
01/10/2017: Offensive Consistency, Defensive Consistency, and Defensive Anchor badge added to the sheet.
03/16/2017: Play Types added to the sheet.
03/22/2017: Pick & Popper and Pick & Roller badges added to the sheet.
07/09/2017: Open Shot Three and Contested Shot Three added to the sheet.
07/13/2017: Off Dribble Shot Three added to the sheet.
07/18/2017: Contested/Open/Off Dribble Shot Mid-range added to the sheet.
08/03/2017: On Ball Defense IQ, Low Post Defense IQ, and Pick & Roll Defense IQ added to the sheet.
08/04/2017: Defensive Stopper badge added to the sheet.
08/19/2017: Speed added to the sheet.
08/20/2017: Speed With Ball added to the sheet.
08/24/2017: Added sheet for all-time rosters (15 teams complete)
09/08/2017: Finished sheet for all-time rosters.
09/15/2017: Added sheet for post-2000 and all-time rosters for NBA 2K18.
09/17/2017: Added sheet for pre-2000 rosters for NBA 2K18.
09/20/2017: Added Play Types sheet with new 2K18 play types.
09/25/2017: Added Spot Up shot/drive tendencies and Off Screen shot/drive tendencies.
09/26/2017: Added post and midrange shooting badges.
10/18/2017: Added Draw Foul attribute.
11/14/2017: Added first release of 2017-18 ratings.
11/22/2017: Tweaked Post Hook, Floater, Post Hook Left, and Post Hook Right
11/29/2017: Tweaked Tear Dropper badge.
11/30/2017: Tweaked Post Fadeaway, Post Fade Left, and Post Fade Right
03/06/2018: Contested Shot Three tendency added for historic players.

If you are making rosters with these edits, PM me your 2Kshare info, the platform, and a link to any posts here on O/S regarding your rosters, and I will add it to the list below.

ROSTERS THAT USE THESE EDITS
--------------------------------------
2K17
------
NBA 2k17 Historic Roster by D-Will254 (Xbox One)

2K16
------
Historic Roster 2k16 by D-Will254 (Xbox One)
Sonicimage tendencies by Thetruth +others by Chmielu9123 (PS4)
2000-01 Roster (PS4) by CD90 (PS4)
Best of the 2000s Work in progress by Jaiveer (PS4)
NBA 2K16 10-11 Roster Now done V1 by Jaiveer (PS4)
09-10 Roster Now Done and Shared by Jaiveer (PS4)

[CD90]

First of all i want to thank you for sharing this with us.
I'm working on a classic roster and have done it before so i know how much time it takes to not just figure out formulas for ratings/tendencies but also collect all the statistics.
Can't even imagine how much time you spend on this.

Because it is so time consuming most people keep their finds close to their chest. So for you taking the community first approch here: Thank you very much.


I will definitly follow these thread and most likely also use some of your ratings/tendencies for my 00-01 roster.

You are talking about 49 formulas. Could you tell us for which ratings/tendencies they will are?

And now on to the already postet shot tendencies.
For a roster i made for 2k13 i tried using something like this.
Player takes 50% of his shots from the paint ---> 50 Take shot in paint tendency.
So all tendencies combined it would be 100.
But this way for almost every player their specific tendency from one area was pretty low. The way you rate it it looks quite similar (low numbers).

I kind of liked the sim stats i got for the 2k13 roster, but gameplaywise it wasn't that great. I had to raise players' tendencies because they didn't want to shoot even if wide open. I guess 30 (3p shooting) was the number a player would consider shooting. Also had to raise the shooting sliders alot.

So i wonder how it plays out with your tendencies. Pretty sure sim stats will be great but i'm a little sceptical about the gameplay part. Relation between the tendencies seem to be good, but will players hesitate to shoot? Guess a special slider set has to be used? You got one?

Can't wait for your next update.
Thanks again!

[SonicMage]

Quote: Originally Posted by CD90

First of all i want to thank you for sharing this with us. I'm working on a classic roster and have done it before so i know how much time it takes to not just figure out formulas for ratings/tendencies but also collect all the statistics. Can't even imagine how much time you spend on this. Because it is so time consuming most people keep their finds close to their chest. So for you taking the community first approch here: Thank you very much.

Thank *you* for taking the time to look at my little project here. I used to be one of those secretive about their findings as well, mainly due to the massive time commitment you mentioned, but with this new generation, the game's attribute tables have become much too complicated for that. It can no longer be a one-man show. Between the ratings, accessories, shoes, playbooks, salaries, shot signatures, it's just too much. So it's time to be more collaborative.

Quote: Originally Posted by CD90

I will definitly follow these thread and most likely also use some of your ratings/tendencies for my 00-01 roster. You are talking about 49 formulas. Could you tell us for which ratings/tendencies they will are?

They span across several different areas: shooting, rebounding, defense, passing. Some I'm more confident in than others, so I'll definitely be tuning as I go and make sure the core formula for any individual rating is something I feel good about before I release.

Quote: Originally Posted by CD90

And now on to the already postet shot tendencies. For a roster i made for 2k13 i tried using something like this. Player takes 50% of his shots from the paint ---> 50 Take shot in paint tendency. So all tendencies combined it would be 100. But this way for almost every player their specific tendency from one area was pretty low. The way you rate it it looks quite similar (low numbers).

I've certainly seen that approach before, using the percentage of shots taken from distances and using that value flat out as the rating. I'm not a big fan, and I'll tell you why. The low numbers you mentioned, combined with a lack of variety. You'll see the way I approach these ratings is with a big emphasis on variety between players. After all, isn't that the goal of having these tendencies, to find the distinction (and similarities) between players to expose their individuality?

For me, tendencies are usually about how *often* and attributes are about how *good*. For the how often tendencies, I like to look at pace-adjusted per minute metrics. So for the overall shoot tendencies, I simply used shots per minute and adjust using the team pace factor. However, I didn't just use field goal attempts. If a player gets fouled while shooting the ball, that's also a shot attempt but it doesn't get counted in the overall FGA. So I also add shooting fouls drawn. For the years I've released here, play by play data is available so we have the exact number of times a player was fouled during the shot. For years prior to 2000, there's an estimate we can use, highlighted in Dean Oliver's book, which is to take the number of free throw attempts and multiply by 0.6. But this per minute value we get doesn't just give us a nice number between 0 and 100 like the game wants. So what I do is use rating lookup tables. I look at the generated values and come up with a good scale table that can convert our generated value to a number that fits within the game's scale. For instance, with the overall shot tendency, I saw that 1986 Jordan was the highest, so he set the bar as the "100" value. Actually, since we're dealing with a huge amount of players, I usually allow a few players to be "100" and scale up to a value just below theirs. I don't have it right in front of me but I think I used 16.0 as the "100" value. I decided to use a linear scale from 0.0 to 16.0, so for instance anybody who has a value between 0.0 and 0.16 gets a '0', between 0.16 and 0.32 gets a '1', 0.32 to 0.48 gets a '2', etc. I don't always use a linear scale like this but for shot tendencies it works out. For the shot distance tendencies, I took a very similar approach but used a different max value than 16.0.

Quote: Originally Posted by CD90

I kind of liked the sim stats i got for the 2k13 roster, but gameplaywise it wasn't that great. I had to raise players' tendencies because they didn't want to shoot even if wide open. I guess 30 (3p shooting) was the number a player would consider shooting. Also had to raise the shooting sliders alot. So i wonder how it plays out with your tendencies. Pretty sure sim stats will be great but i'm a little sceptical about the gameplay part. Relation between the tendencies seem to be good, but will players hesitate to shoot? Guess a special slider set has to be used? You got one?

I have the same concerns with my own generated tendencies, that sim stats will be great but then in gameplay the shot attempts will be too low for some players. I'm not overly concerned that my tendencies are much lower than the default because I know for instance, that the defaults gravely overestimate how much close range and midrange shots that modern NBA players take. It's clear that a custom slider set will be needed to get everything looking right but I'm not going to spend a lot of time on it here in the beginning. Based on my research, there are many factors in gameplay that determine whether a player shoots or not, some that have seemingly no relation with shooting at all, so I'd rather not spend an awful lot of time getting things perfect now, just to have it be all messed up when I release new values for some other rating that may have an effect later on. The other concern I have is that the gameplay designers now have the ability to tweak shot tendencies on the fly, which could throw a wrench in something I spent weeks to perfect. It's better to let their potential trigger-happiness settle down before really trying to get everything perfect, so for now the focus will be on getting numbers in the right ballpark.

Thanks so much for your feedback, this is excellent information and I look forward to hearing more of your input!

[Coachjosh21]

This is awesome. I never understood why people don't release their findings. it boosts the quality of rosters we can create as a community. I look forward to incorporating your numbers in my historic roster.

[SonicMage]

I did some SimCast Live CPU-CPU matchup simming tonight. I have to say, even with the CPU shot tendency sliders at all 50, things are looking pretty good! Too many points in the paint per game (72!), but I think that's more due to other factors that need to be fixed like the bullet passing for the easy score down low which is also causing too many assists, and to a lesser degree the aggressive overabundance of driving to the rim. Mid range jumpers are a tad low but I think that'll rise once the points in the paint passing is toned down, and at least the correct players are taking midrange shots now. Amount of threes taken was a bit on the high side, but it looked pretty decent considering the average 3-point shot tendency is lower than the defaults.

I also forgot to mention the players that had the largest tweaks for each category compared to the defaults, just to show you who really needed some fixing.

For shot tendency:

'08 McGrady: 50->87 (+37)
D. Schroeder: 41->72 (+31)
J. Lucas: 33->63 (+30)
Ish Smith: 44->68 (+24)
M. McGary: 35->58 (+23)
'04 Garnett: 53->75 (+22)
'04 G. Trent: 34->56 (+22)

'01 Madsen: 90->39 (-51)
'01 D. George: 98->58 (-40)
'06 Mourning: 92->53 (-39)
Papanikolaou: 83->46 (-37)
J. Ingles: 81->44 (-37)
'01 I. Rider: 98->62 (-36)
'07 Wesley: 81->46 (-35)

No surprise to see the scrubs way off here, but what's with all the guys from the new classic teams being totally misrated, McGrady a 50?! I'll tell you what happened, a lot of the new classic guys are straight up copy-paste in tendencies from other versions of themselves in the game, no kidding. The other McGrady in the game from the '00 team is also a 50 and underrated. Sad but true.

I don't have all the shot distance tendencies from the defaults marked down yet, but I'll be sure to tell you guys who got fixed the most there once they're ready. As a teaser though, lots of Warriors were way overvalued from 3-point range.

[PhiPsi1]

@Sonicmage -

First off...Thanks for sharing this with the community...I'm thinking that as this thread starts growing and growing with more stats compiled throughout the year, it could become the OS basketball version of FBG ratings (for Madden)!

I have a potentially noob question...

So since all of the shot tendencies are from the 14-15 seasons, do you think we can still plug these values in for players that have switched teams...An example off the top of my head is Josh Smith with his move from Houston to the Clips...Could we plug it in or are these players that have made trade moves better to wait it out a bit to compile more stats?

[STLRams]

Wow, this looks pretty cool. So how has the game play been looking so far. Are teams offense looking more balanced/better, are the players playing more true to life?

[SonicMage]

Quote: Originally Posted by PhiPsi1

@Sonicmage - So since all of the shot tendencies are from the 14-15 seasons, do you think we can still plug these values in for players that have switched teams...An example off the top of my head is Josh Smith with his move from Houston to the Clips...Could we plug it in or are these players that have made trade moves better to wait it out a bit to compile more stats?

A fair question. I would say go for it. Some of these ratings are going to be close to 2K's methods and some aren't. One thing I wouldn't do is try to edit some players and not others, it'll throw everything out of whack because our scales are different and wouldn't play nicely together. You would have some players with much higher tendencies than others when you shouldn't. To use your example of Josh Smith, 2K has him rated at a 90 from close shot tendency and I have him at 43. Both would make him the highest rated on the Clippers, but one overwhelmingly so. So don't cross the streams, please, it would be bad.

One thing this system doesn't do is forecasting. Otherwise, I'd be a millionaire on FanDuel, right? Eventually, I'd like to scale these based on natural progression to try to do some estimating but that's much much further down the line. That also means it doesn't account for player movement, coaching changes, etc. But the reason I say to go for it anyways is because for most ratings, players don't tend to deviate too too much from one year to the next, at most maybe a three to four point swing, unless they're some bench rider who suddenly gets a ton of minutes or vice versa.

My hope is to have new ratings based on this year's stats by the end of the year. Due to some unfortunate circumstances, my code for collecting data was suddenly made unusable and my backup was way too ancient so I'm left having to rewrite it from scratch; safe to say it'll be a while. Luckily the database stayed perfectly intact so all the old stats and ratings are still there.