NBA 2K22 - Tendencies - A better Realism - For Roster creators

[Sheitan]

Hey guys, I have been working a lot on tendencies for the past 2 years, especially on the ATD roster and I would like to share with you my point of view. As I think it would be very useful for all of you, roster creators.

My job is to find out the frequency distribution for each in-game tendency, in order to improve the CPU decision making. I want to find the p25, p50 and p75.

p25 = percentile 25, 25% chance of occurence
p50 = percentile 50, the median, 50% chance of occurence
p75 = percentile 75, 75% chance of occurence



Why do we need this ? Because in statistics this is called a normal distribution. It represents very well real-life behavior, and here the CPU decision making.
By distributing tendencies values around the p25 and the p75 we are going to have smart players, with strenghts and weaknesses, as we see in NBA.
I want to avoid as much as possible extremes values (basically 0 or 99) because it leads to robotic decision.
Indeed with extremes values, the CPU won't have the possibility of choice, it will do the tendency as soon as he is in the situation to do so.

For instance, a player with a 99 stepback3 tendency will do a stepback3 as soon as the situation can trigger a stepback3.
This is dramatic in term of decision making where we want players to think, to choose, if the move is appropriate or not.

Therefore I will provide you with a big table of every in-game tendency, and for each we want :
-minimum value
-p25
-p50
-p75
-max value
-description
-concrete example

Please share if you have any suggestions or remarks, and if you want to contribute to this project :

https://docs.google.com/spreadsheets...it?usp=sharing

Everyone is able to modify the figures if you have better statistics or others remarks.

PS : there is also a thread for MyStaff for those interested to go deeper in the realism (https://forums.operationsports.com/f...-creators.html)

[tarantism]

Following. Is this list final, or are you still testing?

Thanks for posting.

[1CK1S]

Do you have anything for players shot tendencies, shot close, shot under basket, shot mid range, three point shot, and the tendencies that were left blank in the spreadsheet?

[Sheitan]

Hey guys, sorry for the late answer but I have a pretty good news for you, after a month of testing, tendencies are done and the Roster is on its last step before being live !

I am going to create a new thread specifically for it.

Basically the roster will be based on the ATD Roster for badges, playbooks, playtype and attributes.
But as I want to do multiple seasons in MyNBA and also CPU vs CPU the tendencies done in the ATD are not a good fit. But for a User vs CPU you should definitely use the ATD Roster with Vann tendencies !

So tendencies will be mine and will be defined by position and player type base on statistical ranges to create the perfect AI. Taking into account players specificities (Harden stepback 3, Trae Young floater, ...).

The goal of those tendencies is that after multiple seasons, after many trades and free agencies movement and also new players added by the draft you won't have to change anything manually and the sim games as well as the in-game will stay the same and it will be amazing.

By the way you can find the new elements in the spreadsheet. It will be used as well for any draft classes.

The only remaining information I am filling at the moment is the "players type" for each team, so that, the tendency allocation will be done automatically.

By "player type" I mean :
- all around
- two way
- offensive
- defensive
- playmaking
- athletic
- 3&D
- outside scoring
- inside scoring
...

If you want to fill out some information feel free, everyone can edit the spreadsheet. That would be really appreciated !

The next step after will be to create the Roster starting from the last update of the ATD Roster. It won't be that long I suppose.






I have some examples of the amazing AI to show you :

Jokic with a 45° face-up midrange : https://www.youtube.com/watch?v=KYQYrb79L8Y

[Jeffruel]

Just started looking at the spreadsheet. Just want to make sure I am correct on this thinking. Are you saying for instance a 3 and D forward should have his 3 point shot tendency set to 44, spot up set to 20, and off screen set to 10?

If that's the case, I do have a couple follow up questions (nothing to criticize, genuinely just curious and want to understand your reasoning.)

I worked on the ATD project last year and was curious to know what makes your tendencies different from that. I know you mentioned that yours would work nicely for multiple years down the road when everyone has switched teams, but why is that?

I create draft classes on PS5 and I always create them with ATD in mind so that is why I am curious. I like the though of people able to create general player types and use this, so that is why I have a few questions.

You also mention in the first post the p25, 50, and 75. Do you have these numbers listen anywhere for each tendency or player or player type??

Also, what system is this for?

Thanks, and sorry for the questions.

[Sheitan]

Quote: Originally Posted by Jeffruel

Just started looking at the spreadsheet. Just want to make sure I am correct on this thinking. Are you saying for instance a 3 and D forward should have his 3 point shot tendency set to 44, spot up set to 20, and off screen set to 10?

My bad do not consider it... I changed the player type "rebounding" by "3&D" because "rebounding" was kinda useless but I didn't change the tendencies figures. I'll correct it asap.

There are also some stuff about the sim tendencies that I should change to better represent the shot repartition of each player type, the spreadsheet is not totally finished.

Quote: Originally Posted by Jeffruel

I worked on the ATD project last year and was curious to know what makes your tendencies different from that. I know you mentioned that yours would work nicely for multiple years down the road when everyone has switched teams, but why is that?

Firstly, the current ATD roster is built for a User vs CPU using Vann tendencies, he has done an amazing work with hundred hours of testing in-game. I helped him by the way for some post plays. The roster is great and I recommend everyone willing to play User vs CPU to use the ATD Roster.

My roster is focused on CPU vs CPU and unfortunately there is quite a difference with the experience User vs CPU and especially with the 2K22 with the strong defense. It requires more a statistical approach, with small variations between player type to avoid "hardcoded moves" and to create an AI that think and takes the best decision at a certain moment.

For instance look at the "play discipline", if you have big gaps between player type it destroys the balance of tickets shots in CPU vs CPU. We found it last year with Deflo.

Quote: Originally Posted by Jeffruel

You also mention in the first post the p25, 50, and 75. Do you have these numbers listen anywhere for each tendency or player or player type??

It's available in the spreadsheet but in an hidden sheet. I need to update it as well.

Quote: Originally Posted by Jeffruel

what system is for ?

PS5 but anyone willing to play CPU vs CPU on an other platform could take up my data.

No problem I am always happy to discuss and hopefully my answers were clear ! I appreciate your points

[dahoops]

This is some pretty amazing work. I am a member of an online league that is primarily CPU vs CPU and we have struggled with the realism aspect. We have had a lot of focus on the game sliders. The flaws in the default NBA roster alone make that something of a challenge. But, scaling the tendencies as you have demonstrated makes a lot more sense. We have kicked around the idea of tendency and rating modification before. But more on an individual player level. Nerfing guys per se, which kills the realism in a whole different way.

I have been in the league on and off for a number of years and in the early days, we had sometimes used custom rosters. As of late we had focused primarily on the Default 2k roster and incorporated their updates as they are rolled out. But they seem to have very little impact on the overall play of the players.

Over the past few years, I have come to believe that the tendencies have more raw value than individual ratings as the tendencies dictate how the players use the skills that the ratings represent.

If it is ok with you I would like to share your data with my following league members. I am not clear if it will help us immediately. As we run our season concurrently with the NBA. But for future seasons it should help us immensely with determining how to set up the roster for the most realistic results.

[Evtnj]

Quick question on your spreadsheet I see your Touch column for freelance touch values. Where is your column for Freelance Shot? What’s the abbreviation? Only one I can’t find. Thanks.