Team Building and Trade Logic: Your Ideas

Hello lads and gals! I'm posting this thread both here and at OS so I hope you don't mind if I copy-paste the post between the two threads. In a nutshell, what I want is what the thread title says: ideas on team building and trade logic.

Just some pointers, if you're going to post your feedback:
- Saying "I hate this" or "this makes no sense" without any explanation doesn't help. I've felt that posts that had a mere mention of a trade and then "'nough said" are the least helpful of all, as that gives me no idea as to what the state the teams were in and whether the trade actually made sense. If it doesn't make sense in the current NBA, doesn't mean it doesn't make sense in your save's world. As pieces move around and your world diverges from the NBA, things that might fail in the real world might apply perfectly to your save. I'm not saying this as a cop-out. I'm saying that before you post, look at the movement in depth. Is a team trying dump salaries because it's looking to rebuild? Did the trade cause a logjam in a position or cause a position to become highly understaffed? The more factors you bring up in your post (even if it's 10 things that add up to the trade not making sense), the more helpful it will be.
- "X team would never do this". Think twice before saying that. We've seen trades and movements by GMs that had us pulling our hair out, we're all fans and we've been befuddled at times. If we made the AI amazing and the Pistons made moves that actually benefited them wouldn't that make the game extremely unrealistic? (Just kidding, but boy I wish I knew what Dumars is thinking.) Don't think in a box. When building team logic, you can't build 30 different logics and update them ever year. You have to build 1 logic that's dynamic. That can support the weirdest of custom teams. That can support any team situation. So try to think outside the box. Forget about what team it is, and think of whether that move benefited the team (or not) in the world currently existing in your save.
- This is not a thread to discuss user control and commissioner features. I want us to discuss how to improve the logic. If you feel that we wouldn't need to have this discussion if only you controlled the league, fine, you don't have to participate. But there are people that want a living league that acts on its own around them, and I want NBA 2K to move towards that goal tangibly (which is why I was hired to work on the modes and AI).

Now that all the disclaimers are out of the way, feel free to reply. I sincerely hope this thread becomes a hub of constructive discussion, and that hateful posts are kept to a minimum.

And a final note: Around this time last year I was planning on creating a modding tool that would expand NBA 2K's franchise feature. There was an improptu podacst posted here a while back where JaoSming and I were discussing all the features and expanded logic we would add. Knowing what I know now, it's a humbling experience to come in and see how much goes into the logic over how teams trade and build their rosters. I'm not saying its perfect, far from it. But there is logic there, and it's an attempt to have something that works for any team, any league, any point in time and any situation, which is why I insisted that you do not base the discussion on what one of the 30 NBA teams would do today. Small things might seem simple enough to solve. "This makes no sense, I should make sure that it never happens." It's not that simple though, more often than not. Based on all the parameters that go into the AI making a move, it might've made perfect sense.

And with all that, being able to be here, at the VC studio and work directly on the logic that goes into everything that happens, makes me very excited. Which once again, brings me to what I'm here for. I want to hear your ideas from basic groundwork on how teams should look to build themselves up, to very specific details on what should make a trade work or not work.

I can only speak to the last gen version (shoddy and lazy PC port) but the one thing that stands out to me is the impatience of the AI's logic. It seems like (since I obviously can't know) the AI is taking deals on the table because they're on the table rather than letting them go and see what comes along later. This may relate to what standards it's setting for its players and what it can get in return. But it results in the AI being seemingly quite active, with no long-term planning.

The AI for example often offers me deals that are "better" for me in that I'll get a 75 and 72 for a 71 and 68. But for me I'm considering that I'm losing say a good three point shooter and a reserve ball handler and getting back two bigs when I already am fine with that rotation (even if these bigs are "better") for a variety of factors. The AI seems to not take these specifics into account as much and is quite willing to take on plenty of SF's who can't shoot or whatever as long as I'm letting it "win" the trade even if it already has six of them. At other times it seems the AI just desperately wants one type of player and I can have everything I want as long as they're getting it.

This impatience seems to be reflected in one of the big complaints about teams resigning/extending veterans during the season. Noticeable one that happened in my association is the Bobcats extending Ben Gordon to a four year contract worth $9+ million a year. During the season. Now, this does "make sense" in terms of him being their "best" perimeter scorer, he's averaging 20ppg, etc. But he's also 30, has been getting worse over the last few years, etc. Maybe this is a problem with his ratings or whatever, but real life teams would seem to be more patient and long term thinking in not only extending Gordon but also considering whether his position is "filled" or not. Especially depending on if the team is terrible or contending.

To add to that a pithy complaint, teams simply should not be extending veterans during the season. You pretty much never see this in real life.

I think that overall there being more evidence of there being a long term plan or personality to each team would be a goal to pursue. In that you can keep that core logic, but "assign" traits to the GM's and such. So even if they aren't the "real" ones you could have a Sam Presti type who is patient and stockpiles picks, or Morey who is patient and stockpiles players, or you could have a Larry Brown who is impatient and stockpiles randomly, or a Billy King who is impatient and stockpiles bad contracts for below average players, etc. Even if this was "random" and changed every few years, it would still grant more of a personality to each individual team. So maybe one team won't give me a young player almost seemingly no matter what because they're high on potential, while another team is quite willing to jettison young guys or good contracts if I give them a half decent veteran because their Russian owner is demanding contention.

The reason that's what stands out to me is that when you look at say the Thunder, Heat or Spurs in the real NBA, those are cores and even rotations that have been on the team for years now with some exceptions. I sim four seasons into any association and basically every team is completely restructured as if they were all constantly jockeying to make the best move.

Maybe this has nothing to do with how the logic actually works, but that's the perception I get from what I've messed with the game anyway.

Also, you suck and your game is gay.

To add to that a pithy complaint, teams simply should not be extending veterans during the season. You pretty much never see this in real life.

Kobe?

Last edited by deihatein on Wed Jan 29, 2014 5:03 pm, edited 1 time in total.

Some basic opinions:

I don't know how much logic is already build in as far as a rebuilding vs. win-now mode for franchises. Old veterans may be too overvalued in some cases, especially for non-contenders (but I like prospects and rebuilding anyway, so I may be biased). I see Boston trade young guys like Sullinger or Olynyk for mediocre veterans on the wrong side of 30. Orlando trading Nicholson for Jermaine O'Neal. Chicago trading Gibson and a 1st rounder for Elton Brand and Antic. Trades like that.

Also, contract value and years seem to not matter in trades with the AI. Real life example of the Raps giving away Rudy Gay to my Kings and getting little substantial talent back from us. By far the best player, but Toronto might have got pennies on the dollar talentwise in the trade. In the game, Toronto wouldn't let me take Gay for the expirings unless I gave them a 1st and somebody like Thomas or McLemore.

Back to any possibility of assigning a 'mode' and strategy to a team's moves, a rebuilding team will accept a mediocre player on a lengthy contract and give up expirings. I (as Sacramento) can trade Jason Thompson (usually around a 70 or less overall) and his 4yr/$25M for a $3M expiring of Kwame Brown (only 5 overall less than JT) and a 2nd rounder (still pretty good from a bad team like them). There's no way Philly would do that IRL, taking on long-term money and taking PT away from their own young players like Moultrie and Noel. So maybe find a way to lower the value of players whose contracts outweigh their production or perceived talent, as was the case with Rudy.

Probably more of a minor thing, but quality teams seem to trade their rotation players (the ones after their top 2 or 3 overalls) rather frequently. Memphis trading Tony Allen (for Ronnie Brewer), Cleveland trading Varejao (for Calderon), Spurs trade Danny Green (for Will Bynum). Not only is their return often underwhelming or need not taken in account (Cleveland needs Calderon with Kyrie and Jarret Jack?) , but players who seem to be permanent fixtures of their teams IRL are traded without any thought. Teams sometimes seem to trade too much in general, and I rarely see a consistent core of players remain together on a team after 2 or 3 seasons, even if that team is a contender.

I don't know if certain trades are a result of the roster being used (maybe some roster makers rate old veterans like Brand or O'Neal high, which is why they have so much trade value), but I think the contract value in trades and the high trading frequency wouldn't be affected by that. My personal opinion is that mediocre, mid-level players are swapped far too often, while there are no superstar trades (especially in last contract years). Of course there is a fine line to walk in triggering trades. It used to be utterly impossible to complete any trades in some previous games I've played (NBA Live 2005 was my first and only until 2k on my PC now), where now as the player you have the freedom to actually make a trade now and then. The thing I'd like to see the most is a way for teams to adapt their trading strategy in a noticeable way in rebuild or win-now modes.

Hey, Leftos.

I've been satisfied with NBA 2K's trade logic. I've faced some interesting teams in 2K14 as a result of it. I think one thing that might help the trade logic a little is if the AI attempts to trade and acquire players with a little more bias towards the Coach Profiles (what the team's current coach employs).

If I'm thinking correctly, the programmers wouldn't have to code this team-by-team. The game would see which team is looking to trade or acquire a player and factor in the Coach Profile of the team's current coach. In other words, a universal trade logic system that looks for common trends in the coach profile to help determine what kind of players would help that coach's team.

It will benefit a team if the players are capable of doing what the AI asks of it. If a team has a Crash Glass setting of 85, it seems counter intuitive to trade a center with very good offensive and defensive rebounding abilities for another center that is shorter and a mediocre rebounder. The latter will struggle to crash the glass consistently like his coach wants.

Scenario 1: A team with a "Fast Break" setting of 90. They would be more likely to consider acquiring a player if he has high Layup, Speed, Quickness and Hands ratings. Essentially someone who is fast, can catch passes on the go and can still score on contact for a possible And-1. Someone that fits their team's offensive strategy of opting for fastbreaks.

Scenario 2: A team with a "Take Inside Shots" setting of 35, "Take Mid-Range Shots" setting of 65, "Look For Post Players" setting of 40 and a "Crash Glass" setting of 40 that is looking for a backup big man. They wouldn't place as much importance on offensive rebounding or inside scoring attributes (Post Hook, Post Fadeaway, Standing Dunk). They would be more interested in bigs that are capable jump shooters with range.

Scenario 3: A 3rd seeded "win now" team with a coach that is a bad teacher (D or F). Perhaps the team may prefer older role players with more NBA experience, decent overall ratings and less potential as opposed to raw prospects. If I understand correctly, the Teacher rating influences how much skill a player gains during his weekly practices. Why take a chance on someone unproven if they won't gain much from their coach anyway?

If a coach is fired, then the new coach's profile settings would help determine what kind of trades are made moving forward.

Those above me have made great points; however, I think one major issue has been left out is that it is tough to mimic with AI (unless we're talking about Watson). The problem is that all players in the game are viewed as numbers by the computer, not as "people". So what may seem like a good trade to the computer, would be completely illogical to the player since we view the player as a "player" and not a number. Its hard to explain with sentences, so I'll try to explain myself with an example:

From the computer's standpoint, lets pretend the Spurs get offered a player with a stat line of 17pts, 10rebs, 3asts per game. In return, the team offering the trade wants a player with a 15pts, 10rebs, 3asts. Oh, and the player being offered to the Spurs 5-years younger. If that is all you knew about the trade, I'd think the majority of people would accept...I mean, why not? If a team offers me a player with a very similar stat line, but younger...I'd take it. This is how a computer views the trade, which makes sense to me why they accept seemingly "weird" trades. I saw that because here is what the player sees...

Memphis Grizzlies have offered Zach Randolph to the Spurs in return for Tim Duncan. Would most players approve this? Probably not. Would the Spurs approve this in real life? Probably not.

My point is, computers don't "know" how to think like us, process emotion like us, and understand the intangible things that humans do. The tougher question is how to get to this level within a game. At this point in time, research and development hasn't given mainstream software the capability to do such things. IBM's Watson is getting close, but that project's sole purpose is AI...NBA 2k has a lot more things to handle.

That's my two cents...

Those above me have made great points; however, I think one major issue has been left out is that it is tough to mimic with AI (unless we're talking about Watson). The problem is that all players in the game are viewed as numbers by the computer, not as "people". So what may seem like a good trade to the computer, would be completely illogical to the player since we view the player as a "player" and not a number. Its hard to explain with sentences, so I'll try to explain myself with an example:

From the computer's standpoint, lets pretend the Spurs get offered a player with a stat line of 17pts, 10rebs, 3asts per game. In return, the team offering the trade wants a player with a 15pts, 10rebs, 3asts. Oh, and the player being offered to the Spurs 5-years younger. If that is all you knew about the trade, I'd think the majority of people would accept...I mean, why not? If a team offers me a player with a very similar stat line, but younger...I'd take it. This is how a computer views the trade, which makes sense to me why they accept seemingly "weird" trades. I saw that because here is what the player sees...

Memphis Grizzlies have offered Zach Randolph to the Spurs in return for Tim Duncan. Would most players approve this? Probably not. Would the Spurs approve this in real life? Probably not.

My point is, computers don't "know" how to think like us, process emotion like us, and understand the intangible things that humans do. The tougher question is how to get to this level within a game. At this point in time, research and development hasn't given mainstream software the capability to do such things. IBM's Watson is getting close, but that project's sole purpose is AI...NBA 2k has a lot more things to handle.

That's my two cents...

This is pretty easy to fix.
Players should be assigned status: Franchise player, Star, Starter, etc. Logic should take into account:
- Player status
- GM philosophy (tanking? rebuilding? contending?)
- Positional need
- Player age
- Player's contract to talent ratio

To be extra tricky, if players have a listed hometown, the formula could include a desire to play in a certain market.

This is pretty easy to fix.
Players should be assigned status: Franchise player, Star, Starter, etc. Logic should take into account:
- Player status
- GM philosophy (tanking? rebuilding? contending?)
- Positional need
- Player age
- Player's contract to talent ratio

To be extra tricky, if players have a listed hometown, the formula could include a desire to play in a certain market.

I agree with your points; however, the problem still remains. Everything you mentioned requires a variable, that's the only way the computer will "see" it. How would you do it realistically? For example, if go to hard one way, teams will turn down ridiculously good deals. For example, lets say the Mavs are in the playoff hunt and Miami is playing awful. Lebron is in a contract year, so the AI wants to dump him before the deadline to get something in return. Turns out they want to send LBJ in return for Dirk. Would Dallas turn down a LBJ-for-Dirk deal? I'm pretty sure Dallas would say "Yes!".

As for contracts, I feel like most sports games already handle this pretty well. It can be tough to get out from bad contracts while getting anything of value in return.

AI is a very tough nut to crack, especially considering the level players want it at. Until natural processing becomes more mainstream (and easy), video game AI / logic is going to remain stagnant with little advances.

An important issue is that the CPU teams are at a disadvantage. They're driven by static logic. Meanwhile, the human players keep hammering at the logic until they find the holes.

1. Instead of, or in addition to, logic that tries to make CPU teams mimic real life GM's, make CPU teams mimic human gamers. Give the CPU teams an option of acting more like human gamers.

2. Let the CPU learn trade and roster building logic by watching and mimicking the way human players trade players and build rosters.

3. There are sliders for game play. Add sliders, team by team and league wide, for roster building and trade logic.

Value overall vs value a specific attribute.

Trade superstars 0 to 100

Value cap room

Value Youth

Value Experience

Value Chemistry

Consider Record or Team Attitude (rebuild vs win now vs add a specific piece like a 3PT shooter or Rebounder or Post Defender or Bench Scorer)

4. Let the CPU cheat by simming a number of seasons (or games) with the new roster vs the old roster (pre and post trade.) Let that result influence the trade decision.

5. Make it harder to trade by limiting the number of times a day and/or total number of times someone can shop a player. Make this an option, so those who want more of a challenge can have it.

6. What are people's house rules? Incorporate some house rules (fixes) people have used into trade logic. Give people the ability to add custom house rules such as limiting the number of trades they can make in a season, and other house rules people use.

To add to that a pithy complaint, teams simply should not be extending veterans during the season. You pretty much never see this in real life.

Kobe?

Kobe is a star, It' different. For example, in my associations, Utah always give new contracts to Jefferson and Biedrins, even if that make them going over the salary cap. THIS should not happen!