ATD based Roster - Sheitan Project - Next Gen

[Sheitan]

Hey 2K Community ! First of all, I would like to introduce myself. I am French, NBA fan for more than a decade now. I work in trading for years so I am pretty good with data analysis, statistics. CelticLG has done an amazing work with his roster Attention To Detail (ATD) with others contributors and I want to thank them all.

For a year now, I am playing MyLeague/MyNBA based on the ATD roster. Mainly on CPU vs CPU (12min Quarter) where I am constantly doing adjustments of players tendencies, global sliders. From my point of view and based on the multiple hours of backtesting I found out weird behaviour in the post game, the decision making of ball handlers, the ball usage rate among a CPU team.

That’s where I will bring major updates that will completely change the game, by improving the CPU decision making and the general gameplay.

In addition, I want to correct the players ratings and the model behind to define it, because of some significant discrepancies. On that matter I would like to bring a statistical point of view of players ratings, based on a “normal distribution” which is a fundamental principle in our world. To make it simple, average players (50% of the pool) should be between 77 and 79, 25% below 76 and 25% above 80.

Player rating concept :

Spoiler

Quote:

In that way, with the changes in tendencies and ratings if you plan to multiple seasons you will have scenarios like Jeramy Grant who was a role player at the Nuggets became FP at the Pistons but he should not be a 85+. Also, player potential should be lower and players evolution over time should be limited to be closer to the reality, and not having too many 80+ players.

Player rating issues - Explanations & examples :

Spoiler

Quote:

For instance, defensive players have an overall rating way too high. If you put “intangibles” to 99 for every player to check the real player intrinsic value, Ben Simmons is the best PG, Marcus Smart is the second best SG, Bledsoe the third… Hardaway Jr. is top1 in layup… Embiid/Kawhi/Jokic are clearly underrated at closed/midrange. there are outliers that we have to handle.

Project : I will be honest with you guys, I work a lot and therefore I don’t have that much free time to do all the changes myself, but I gathered enough information to offer you an amazing and realistic basketball game. If some people can help to build the roster that would be really appreciated. We can create it step by step.

Road Map :
1. I will do the major tendencies update this week based on the scenario of jparks85 (ATD last update) and also the sliders.
2. Player ratings & CelticLG formula improvement
3. Tendencies update to make players unique based on statistical data (hook, fadeaway, ...)



Tendencies :

Spoiler

Quote:

- Post Game (everyone): - Post shot types such as “Hook”, “Fadeaway”, “Shimmy”, “Hop shot”, “Up and Under” : max 20 - Post movements such as “Post drive”, “Post spin”, “Dropstep” : max 25 - Backdown/Aggressive Backdown : max 50 Explanations & examples : Spoiler Quote: With the current values, players receiving the ball at the post will instantly shoot or triggers a post move. For instance seeing Zion throwing a fadeaway as soon as he receives the ball it’s just horrible. With the new figures these players will take their time to “work” the defender in order to take a better shot. - Post Game for players with a good handle (Zion) : - Post face-up : 100 Explanations & examples : Spoiler Quote: Post face-up is linked with backdown/aggressive backdown but the last 2 have the priority. Meaning that with 50 for both backdown and a 100 in face-up, the face-up situation will occurs 33% of the time and the backdown 66%. - Post Game for players with a good mid-range (Embiid, Jokic, Wood, Vucevic, AD, Boucher…) : - Post face-up : 100 - Spot-up mid-range : between 50 & 75 - Contested mid-range : between 25 & 50 - Triple Threat Shoot Tendency : between 50 & 95 Explanations & examples : Spoiler Quote: As mentioned above, the face-up situation won’t occur that often so do not worry with the 100 value. Also players like Embiid, Jokic, Anthony Davis use a lot the face-up with triple threat, resulting sometimes with a 45° jump shot. This is a situation that will appear with the above tendencies. They can also do a pump fake and drive to the rim. Please find below an example of what these tendencies will bring : versatility at the post video example - Everyone : - Shot tendency : max 90 (you can basically decrease the entire NBA player by 9 to do so, it won’t affect simulated games) - Touches : max 95 - Play discipline : - Super star : 80 (Harden, Giannis, …) - All star/lieutenants/6th man : 85 (Middelton, Clarkson, LeVert, …) - Role player who touch the ball or PG who call plays and does not shoot that often : 90 (G. Trent Jr, Rubio, Rondo, …) - Role player waiting in the corner : 95 (PJ Tucker, Covington, …) Explanations & examples : Spoiler Quote: With the current figures, each ball handler is keeping too much the ball and call play for himself and as a result the ball handler will shoot too often. That’s why “Touches” must be above “Shot tendency” for playmaker and stars so that the ball can move. And avoid the 99 figures as much as possible. Example : IRL Kawhi and PG did ask their GM for a PG to call plays and handle the attack, they got Rondo, but in 2K even with the current tendencies he won’t handle the attack that often. Here is my solution, and it brings a very good gameplay, where Rondo will call many plays for PG and Kawhi, and sometimes the 2 superstars will also handle the attack. Kawhi, “Shot tendency”=90, “Touches”=95, “Play discipline”=80 PG, “Shot tendency”=90, “Touches”=95, “Play discipline”=80 Rondo, “Shot tendency”=50, “Touches”=90, “Play discipline”=90 - Ball Handlers : - Off-scren 3 : maximum 50 - Step-back 3 : maximum 75 - Spot up 3 : maximum 80 - Spot up drive : maximum 80 Explanations & examples : Spoiler Quote: Currently there is a big issue where a lot of ball handlers have 99 in off-screen 3, spot-up 3, step back 3 like James Harden. This player will shoot 3 way too often with the current tendencies. In real life he does not shoot 3 that often compare to his ball usage rate, he is mainly driving after a screen and that’s why he has more than 10 assists per game. Also the stepback3 happens only when the player is driving, he stops and does a step back. With a value of 99, players do this stepback too often. James Harden does a step back 3 mainly in isolation after a “triple threat sizeup”, unfortunately a slider “triple threat sizeup stepback 3” does not exist. But even with a “stepback3”=75 he will do it a lot in a game (between 5 and 10 time at least). There is a situation if you have “spot up 3”=99 and “off-screen 3” low, where the ball handler calls a screen, he doesn’t shoot and doesn’t drive, but he sees him at “open at 3” so he is doing an horrible “moving 3”. With a “spot up 3”=80 max, those “moving 3” won’t happen again, if it does, lower again the “spot up 3”. “Spot up drive”=99 is horrible, players like Fox and John Wall will try to drive all the time, as a first choice like a bot. In 2K driving seems to have the priority on every other parameter, so we have to lower it.

Sliders :

Spoiler

Quote:

- Offense : - high values for the different shots Spoiler Quote: otherwise the success rate is too low. Higher value for close/mid as the probability to be contested is higher, so the success rate decrease. - layup success=45. Explanations : Spoiler Quote: the goal is to find a good balance between open layup, contested layup and contact layup where every layup can be missed and can go in at the appropriate percentage, and not 100% or 0% as it's the case currently. - Dunk in Traffic Frequency = 80 Spoiler Quote: (especially for Centers who are really close to the RIM, IRL they just dunk on you. If you put it lower they won’t dunk, above every slasher will dunk too much during a drive) - Defense : VERY IMPORTANT - Driving contact/Inside contact/layup strength defense (takeoff/release) stay at 50. Why ? Spoiler Quote: Otherwise it will trigger too many contact animation each time a player drive and they will miss way too much. With slider above 50 sometime you can have a defender which is really far behind the attacker and still he does a fake huge jump and he comes contesting the layup… When IRL there not that much contact in the paint, but shot are contested, and there is a big difference between contact & contested shot. I want more fluidity during drives, more contested shot than contact shot, also it won’t change the number of faults. With this setting and layup success=45 you will see very realistic behavior where players can score despite the contested shot, or can be really affected by the presence of a defender and miss, and they can also miss even if the contest is not that strong. From my point of view it brings a great balance. If you still have attackers avoiding contact or contested shot, you have to kill the badge “Slightery finisher” because it is the main responsible of weird behaviour in the paint. - Jump shot defense (gather) = 38. Why ? Spoiler Quote: Otherwise the presence of a defender near by is way too strong and CPU without the badge “Deadeye” cannot score. 38 is a good balance - Jump shot defense (release) = 15. Why ? Spoiler Quote: This happen most of the time when a player is open and shoot, then a defender tries to contest but he is far away so the contest impact is really low. That’s why I want to lower it but still keeping a small effect. At 50 the impact is way too important. - Attributes : - Defensive Awareness = 75. Spoiler Quote: The defence awareness is terrible especially during pick&roll so we try to boost it without having it too overpowered - Lateral quickness = 99. Spoiler Quote: It will mainly improve the pick&roll defence, and avoid the blow by. - Tendency : - Post shots : 20 Spoiler Quote: to avoid as much as possible having players instantly shooting when they receive the ball at the post - Attack the basket : 20 Spoiler Quote: this slider has the priority on everything, so lower it, otherwise players will drive no matter what - Look for post players : 90 Spoiler Quote: it helps to get the ball to the post when a player is posting - Throw alley-oops: 99 Spoiler Quote: otherwise there are no alley-oops, even with 99 it won’t occur that often. - Attempt Dunk : 80 Spoiler Quote: for center mainly, if slashers try dunk to often try to lower it step by step with a -5 until you find a good balance - go for on-ball steals : 5 Spoiler Quote: otherwise the CPU is trying way too often. Also the CPU will be more concentrated in staying in front of the attacker - Contest shots : 35. IMPORTANT Spoiler Quote: The definition of this slider in 2K is : Jump and try a block. With the ATD figures every good defender has like 99 in contesting shots, so even with this slider at 1 they will keep contesting almost every shot. The goal here is to define a good balance when players that are supposed to contest they do it (mainly centers) and players who are not supposed to contest that often won’t contest every shot, but no worries they will defend and glue the attacker during drive, it's just they don't try a block (basically Harden or Seth Curry). What I am doing is during the regular season I put this slider at 20 or 25, and at 35 in playoff. It brings very good behavior. As an example we see it a lot IRL where an attacker use a pick&roll to open a path to the rim, he drives, then there is a defender blocking the access to the rim but leaving the attacker open at mid-range. The defender’s goal is to defend a potential alley-oop and a layup so he prefers an open shoot at mid range. If you put contest shot at 50, the CPU will basically contest the mid-range and so in case the attacker miss the mid-range, his teammate center will grab the offensive rebound 99% of the time. Example (look at the piston’s center, blocking alleyoop, layup and potential offensive rebound) : video example - Fouls : 100 (for the CPU) - Movement : Spoiler Quote: IRL the difference between fast & slow players is not that big, in 2K it is. If you have a too low min rating, slow players will take ages to move. Especially when you call a pick&roll. This setting still allow fast player to beat their opponent with their speed, but it will change the impact of big men. If you don’t like it you can increase the max but do not decrease the min. - Max : 62 - Min : 52 Comments : If you are playing CPU vs CPU : “steal success” and the attribute “stealing” should be at 99, otherwise they are no steals. If you are playing User vs CPU : “steal success” and the attribute “stealing” should be at 50. Otherwise the CPU will steal to you too often. Maybe change the ball security & body-up sensitivity to 50.

Players Rating : On-going

Spoiler

Quote:

- If you put “intangibles”=99 you will see the intrinsic value of every player, and defensive perimeters players are completely broken like Simmons best PG (90) or Bledsoe (87) third best SG and Marcus Smart (90) the second best SG… When Jamal Murray is a 80… The balance between good attackers and good defenders is bad. - I don’t have a perfect view for now but I want to give more strength to good offensive players and lower the rating of good perimeters defenders. With the idea to keep 50% of the players around 77-79, 25% above 80 and below 76 - Increase the mid-range and close shoot for superstars like Kawhi, Embiid, Jokic. - In the formula of CelticLG’s spreadsheet, we should put more weight on the current year figure, with a ratio of 75% current year and 25% last year rating. Julius Randle did improve a lot at 3 for instance.

MyNBA sliders & special processes :

Spoiler

Quote:

- Contract : keep the original 2K settings, because we see a lot of players around 80 or old players (Dragic) getting a crazy contract (around 20m$ a year). - Off-season free agency (how to have good players signing for the minimum contract) : there are 12 days in the FA, when you arrive at the day 9, I change the “contract sliders” so that every player will accept the minimum contract. It has a pretty good result because at that time most of the players already got a contract but there are still few players around 78-79 overall ratings. So you will find few ring chasers around 78-79 going to the Lakers or Brooklyn. Then after the FA you put back the “contract sliders” to its original values.

Current issues & To be improved :

Spoiler

Quote:

- Pick&Roll defence : currently it’s horrible. When you start the game go to each CPU team and change the “help defence slider” to 99, it’s the only way to improve it a bit. - "Speed with ball" attribute : decrease everyone by 15, with a max value at 80-82 for Fox & Westbrook & Wall. Otherwise they are completely broken. - Off-season “Player progression” badges acquisition is totally random, we have to correct the acquisition manually for each player (Harden & Doncic earning defensive badges like “Off ball pest”) - Off-season “Player progression” tendencies evolution : tendencies changes drastically in post-season for every player (shot under, shot tendency). For instance the “Shot under” tendency that was 99 for Giannis and Zion will become something like 15, a really low number, and it will affect all the simulated games for the next season… So you have to change it manually. SOLUTION STILL IN TEST. For the moment what I do is simple : 50 for everyone, 75 for PF like AD and 99 for center or players without a good shoot like Giannis or Zion.

Thanks a lot for reading me, if you have questions or remarks feel free to share. Also if you want to be part of this project you can !

[pjt8405]

Quote: Originally Posted by Sheitan

Hey 2K Community ! First of all, I would like to introduce myself. I am French, NBA fan for more than a decade now. I work in trading for years so I am pretty good with data analysis, statistics. CelticLG has done an amazing work with his roster Attention To Detail (ATD) with others contributors and I want to thank them all. For a year now, I am playing MyLeague/MyNBA based on the ATD roster. Mainly on CPU vs CPU (12min Quarter) where I am constantly doing adjustments of players tendencies, global sliders. From my point of view and based on the multiple hours of backtesting I found out weird behaviour in the post game, the decision making of ball handlers, the ball usage rate among a CPU team. That’s where I will bring major updates that will completely change the game, by improving the CPU decision making and the general gameplay. In addition, I want to correct the players ratings and the model behind to define it, because of some significant discrepancies. On that matter I would like to bring a statistical point of view of players ratings, based on a “normal distribution” which is a fundamental principle in our world. To make it simple, average players (50% of the pool) should be between 77 and 79, 25% below 76 and 25% above 80. Player rating concept : In that way, with the changes in tendencies and ratings if you plan to multiple seasons you will have scenarios like Jeramy Grant who was a role player at the Nuggets became FP at the Pistons but he should not be a 85+. Also, player potential should be lower and players evolution over time should be limited to be closer to the reality, and not having too many 80+ players. Player rating issues - Explanations & examples : Project : I will be honest with you guys, I work a lot and therefore I don’t have that much free time to do all the changes myself, but I gathered enough information to offer you an amazing and realistic basketball game. If some people can help to build the roster that would be really appreciated. We can create it step by step. Road Map : 1. I will do the major tendencies update this week based on the scenario of jparks85 (ATD last update) and also the sliders. 2. Player ratings & CelticLG formula improvement 3. Tendencies update to make players unique based on statistical data (hook, fadeaway, ...) Tendencies : Sliders : Players Rating : On-going MyNBA sliders & special processes : Current issues & To be improved : Thanks a lot for reading me, if you have questions or remarks feel free to share. Also if you want to be part of this project you can !

Phenomenal info!

Quick question,
Are you aware of the default setting for both teams where each team's freelance is set to "Pound the Ball Inside"?
You have to change both teams to "Auto" at beginning of the game and wait for one possession before the game recognizes the change. Also, at the start of each half, the non-user-controlled team must be changed again as they revert to "Pound the Ball Inside" each quarter. You do not need to change the user-controlled team.

Three other areas which make a HUGE difference in gameplay are

ISO Tendencies ( many teams have WAY TOO MANY players with ISO tendencies despite not being ISO guys

Transition Spot-Up vs Cut - Many players are way off by default in this area, as Shooters should be MAX 95 for example, and Rim Runners who don't shoot or Bigs should be 0

Roll vs Pop - These need to be cleaned up across the board as many are already correct but this years data should be included



Also, the Playbooks and Play-Types are VITAL to proper gameplay. Many teams are missing "Cutter" plays in their playbook despite guys like Clint Capela cutting for Alley-Oops etc. Many player types include the 3PT play type (for example = Brandon Ingram) despite not being a higher volume 3PT shooter. That changes the offense gameplay tremendously if edited correctly.

Finally, Coach POE play a big part in simulation as well as each Head Coach and Assistant Coach grade and System Proficiency (In the "Edit Coach" menu)

[Sheitan]

Quote: Originally Posted by pjt8405

Phenomenal info! Quick question, Are you aware of the default setting for both teams where each team's freelance is set to "Pound the Ball Inside"? You have to change both teams to "Auto" at beginning of the game and wait for one possession before the game recognizes the change. Also, at the start of each half, the non-user-controlled team must be changed again as they revert to "Pound the Ball Inside" each quarter. You do not need to change the user-controlled team.

Great news I was not aware but for sure I will do it !

Quote: Originally Posted by pjt8405

Three other areas which make a HUGE difference in gameplay are ISO Tendencies ( many teams have WAY TOO MANY players with ISO tendencies despite not being ISO guys

Indeed I forgot to mention that "player tendencies" related to Isolation are also way too high like Harden having 99 everywhere, it will trigger so many isolation when he does not do it that often per possession, especially now that he plays at Brooklyn.

If you are talking about the playtype, in general the ATD Roster does a correction where not that many players have isloation in their playtype.

Quote: Originally Posted by pjt8405

Transition Spot-Up vs Cut - Many players are way off by default in this area, as Shooters should be MAX 95 for example, and Rim Runners who don't shoot or Bigs should be 0 Roll vs Pop - These need to be cleaned up across the board as many are already correct but this years data should be included Also, the Playbooks and Play-Types are VITAL to proper gameplay. Many teams are missing "Cutter" plays in their playbook despite guys like Clint Capela cutting for Alley-Oops etc. Many player types include the 3PT play type (for example = Brandon Ingram) despite not being a higher volume 3PT shooter. That changes the offense gameplay tremendously if edited correctly.

Are you talking about the original 2K Roster or the ATD ? Because from my point of view that's pretty good in the ATD on those 3 topics. For instance each playbook contains the maximum number of system which is 50 and the balance seems quite ok.

[pjt8405]

Quote: Originally Posted by Sheitan

Great news I was not aware but for sure I will do it ! Indeed I forgot to mention that "player tendencies" related to Isolation are also way too high like Harden having 99 everywhere, it will trigger so many isolation when he does not do it that often per possession, especially now that he plays at Brooklyn. If you are talking about the playtype, in general the ATD Roster does a correction where not that many players have isloation in their playtype. Are you talking about the original 2K Roster or the ATD ? Because from my point of view that's pretty good in the ATD on those 3 topics. For instance each playbook contains the maximum number of system which is 50 and the balance seems quite ok.

This definitely changes the gameplay and in my opinion changes the strength of both the Tendencies and Slider changes. The AI will look to move the ball around more with the setting changed to "Auto"


I was referencing the 2K Default Roster. Just was making a general list of important things to look out for.

I think what is important to note is that no one community-wide really knows how to effectively change the Tendencies completely for the better. We do not know the under the hood coding in which the 2K scales are built off of. We can all try and get them close to realistic, but the true test is trying to build proper tendencies off default sliders and work from there in my opinion. As someone who has deep knowledge of the player tendencies, I'll say that it is a long process to edit, but even the slightest wrong tweaking of one area can completely mess up gameplay.

Looking forward to this completed project as I'll definitely be checking it out

[paqofskillz]

I wish u had a ps4, the Atd roster i released has all 522 players rated for this season and included your post tendency changes. I do have a spreadsheet with everything player stats and shot location tendency I didn't release yet but your research is on point bro for sure some of the changes are very noticeable and personally you helped solved some of the issues I couldn't figure out yet with the post play. Thanks for your research brother

[Reddcheese]

Any update on this mon friere?

[Love13]

Hey Man, I love your work. Very detailed and constructive with examples/reasons and very well written. You touched on a few things that I have questioned through some testing such as shot type tendencies being 99 possibly trumping other tendencies, mainly the 3pt spot up/off screen/pull up etc.

I also like the lower contact/defense strength sliders, I've been doing tests with lowering those as well as lowering shot success sliders to go with them in gameplay and have had some success. With that, I do have some questions:

- How have fouls/free throw attempts been effected with your much lower values in contact and defense strength?

- With steal tendencies lowered, are you still getting realistic turnover balance? (I find this hard to get this high enough on average as is so may be kind of a mute point lol)

- As far as ratings go, I think the bell curve idea is what Celticlg was going for in the ATD roster, but in your findings you suggested the intangibles at 99. Would you recommend keeping intangibles at 99 across the board and then build the overalls around that to help create that balance with everyone on a "level playing field" so to speak?

- With the ratings system, have you found anything in regards to player progression/regression over time? I know that has been one of the major balances trying to be accomplished within the community.

- And last is more niche area but, have you by chance found anything in sliders as it relates to player lock? There's a small group of us that prefer that if we can find a good balance, but 2K seems to try and break teammate ai so I'm curious if you have found anything that effects teammate AI (Shooting mainly, I have yet to find a ton of rhyme or reason to user ai fg%) on player lock either negative or positive.

Thanks for the insights, im going to look into some things based on your explanations above.

[Love13]

Also, you should connect with Celticlg directly and share some of your findings/collaborate. I think these results could really take the ATD roster to the next level!