CPU’s knowledge on matching up…

[KenXCarson]

Re: CPU’s knowledge on matching up…

It would be nice if 2K would have prompts. Like imagine if I could set a prompt to have Trae Young guard the worst offensive perimeter player or guard lol. Or have OG Anonouby guard the best wing player on the opposing team. 2K needs to hire me, I'd be an idea factory for these guys lol

[ggsimmonds]

Re: CPU’s knowledge on matching up…

It would be nice if 2K would have prompts. Like imagine if I could set a prompt to have Trae Young guard the worst offensive perimeter player or guard lol. Or have OG Anonouby guard the best wing player on the opposing team. 2K needs to hire me, I'd be an idea factory for these guys lol

Can you program AI?

Its not that they haven't thought of these things, the problem is coding AI is hard

[m29a]

Re: CPU’s knowledge on matching up…

I suspect it's likely about priorities. Putting resources into making defense smarter is not flashy and the bulk of people probably don't notice or care about it in general.

[KenXCarson]

Re: CPU’s knowledge on matching up…

Can you program AI?

Its not that they haven't thought of these things, the problem is coding AI is hard

You're probably right, it does sound a lot more difficult than one would just safely assume. I work with software, and I feel like it would just be a simple conditional formula tho, no?

If X defender has better ratings than Y defender, than he would lock onto the best offensive rated player.

It's lowkey already built in with the defensive matchups setting as is, but they just need to enhance it

[ggsimmonds]

Re: CPU’s knowledge on matching up…

You're probably right, it does sound a lot more difficult than one would just safely assume. I work with software, and I feel like it would just be a simple conditional formula tho, no?

If X defender has better ratings than Y defender, than he would lock onto the best offensive rated player.

It's lowkey already built in with the defensive matchups setting as is, but they just need to enhance it

Its far more difficult than that.

First how do you quantify "better" and second how do you address the downstream effects? E.g. if the result of your conditional formula is that we now have the 3 guarding the 1, who does the 1 guard? What logic are you going to code to handle that assignment? I.e. are you just going to allow him to default to guarding the now uncovered man or do you provide some further conditional logic?


That first part is a lot harder than it seems because the logic has to be precise. If you write it wrong you could end up with some undesirable scenarios like Marcus Smart trying to defend Durant.

[FixEverything2k]

Re: CPU’s knowledge on matching up…

Its far more difficult than that.

First how do you quantify "better" and second how do you address the downstream effects? E.g. if the result of your conditional formula is that we now have the 3 guarding the 1, who does the 1 guard? What logic are you going to code to handle that assignment? I.e. are you just going to allow him to default to guarding the now uncovered man or do you provide some further conditional logic?


That first part is a lot harder than it seems because the logic has to be precise. If you write it wrong you could end up with some undesirable scenarios like Marcus Smart trying to defend Durant.

I feel like we’re overcomplicating it. Not saying it’s easy but:

Players have 2 O grades, Perimeter O Interior O,
2 Defence grades, Perimeter D Interior D.

A team for the most part will only have so much A to A+ rated players in these categories.

If Perimeter/Interior O is much greater than the Defence rating of the player guarding them, match your best Perimeter/Interior D to them.

If all other players are not superstars, match for closest position/height and call it a day.


As it is now, if you manually put your SF on your PG, out of 10 times how many times does it completely throw off the rest of your lineup? For me it doesn’t mess it up more often than not unless a ridiculous lineup is being ran.


If we want it to start getting more complicated add a 3rd defence rating to this equation with Help D. If there’s a good help defender and a poor offensive player, put the good help defender there to help the 2nd best defender guard the best offensive player.

They have the tools in the game already they don’t need to restructure nothing, just put it to use.

[ggsimmonds]

Re: CPU’s knowledge on matching up…

I feel like we’re overcomplicating it. Not saying it’s easy but:

Players have 2 O grades, Perimeter O Interior O,
2 Defence grades, Perimeter D Interior D.

A team for the most part will only have so much A to A+ rated players in these categories.

If Perimeter/Interior O is much greater than the Defence rating of the player guarding them, match your best Perimeter/Interior D to them.

If all other players are not superstars, match for closest position/height and call it a day.


As it is now, if you manually put your SF on your PG, out of 10 times how many times does it completely throw off the rest of your lineup? For me it doesn’t mess it up more often than not unless a ridiculous lineup is being ran.


If we want it to start getting more complicated add a 3rd defence rating to this equation with Help D. If there’s a good help defender and a poor offensive player, put the good help defender there to help the 2nd best defender guard the best offensive player.

They have the tools in the game already they don’t need to restructure nothing, just put it to use.

I'm a software developer in real life. You guys are underestimating the code and logic that has to go into something like this. Programming doesn't work like "put the best defender on the best offensive threat." You have to operationalize it and you have to do so in a way to cover almost all scenarios.

For your hypothetical question that I bolded, if you were a programmer and said "well my code works more often than it doesn't" you wouldn't be a programmer for long lol.

Not saying that asking for this is asking for too much. It is something that is reasonable to hope for. But its also not as simple as you make it out to be. My main point is in regards to the comment about "hire me for ideas" the other poster made. It has nothing to do with ideas; this isn't a new idea. It has everything to do with writing an AI that can incorporate this level of complexity. Two years ago in a live stream Czar said he desperately wants to get this in the game but 2k's AI is not at a place where they could add it. I think with this year's ACE substitutions we are probably getting closer to seeing dynamic matchups in the game, but I don't know when we can expect to see it.

I don't know that this is something they can just drop on top of the current Defensive AI or if it would require something of a rewrite. If its the latter, then its what m29a said and we are down to priorities.

[FixEverything2k]

Re: CPU’s knowledge on matching up…

I'm a software developer in real life. You guys are underestimating the code and logic that has to go into something like this. Programming doesn't work like "put the best defender on the best offensive threat." You have to operationalize it and you have to do so in a way to cover almost all scenarios.

For your hypothetical question that I bolded, if you were a programmer and said "well my code works more often than it doesn't" you wouldn't be a programmer for long lol.

Not saying that asking for this is asking for too much. It is something that is reasonable to hope for. But its also not as simple as you make it out to be. My main point is in regards to the comment about "hire me for ideas" the other poster made. It has nothing to do with ideas; this isn't a new idea. It has everything to do with writing an AI that can incorporate this level of complexity. Two years ago in a live stream Czar said he desperately wants to get this in the game but 2k's AI is not at a place where they could add it. I think with this year's ACE substitutions we are probably getting closer to seeing dynamic matchups in the game, but I don't know when we can expect to see it.

I don't know that this is something they can just drop on top of the current Defensive AI or if it would require something of a rewrite. If its the latter, then its what m29a said and we are down to priorities.

My point with the more often than not was to say with the current system that is already in the game, when you make a switch in defensive settings and put your small forward on the point guard, it doesn’t completely break the game.

Theres several things in this game that are still a work in progress but are available, let this be one. If it doesn’t work during one of the matches, turn it off.

I could have sworn this existed in the game at one point when ACE was first coming in anyways.

Put it in as a beta or something.

[ggsimmonds]

Re: CPU’s knowledge on matching up…

My point with the more often than not was to say with the current system that is already in the game, when you make a switch in defensive settings and put your small forward on the point guard, it doesn’t completely break the game.

Theres several things in this game that are still a work in progress but are available, let this be one. If it doesn’t work during one of the matches, turn it off.

I could have sworn this existed in the game at one point when ACE was first coming in anyways.

Put it in as a beta or something.

It was in previously. Its just speculation but I am inclined to agree with ChaseB and think it was hardcoded rather than dynamic AI

[KenXCarson]

Re: CPU’s knowledge on matching up…

I feel like we’re overcomplicating it. Not saying it’s easy but:

Players have 2 O grades, Perimeter O Interior O,
2 Defence grades, Perimeter D Interior D.

A team for the most part will only have so much A to A+ rated players in these categories.

If Perimeter/Interior O is much greater than the Defence rating of the player guarding them, match your best Perimeter/Interior D to them.

If all other players are not superstars, match for closest position/height and call it a day.


As it is now, if you manually put your SF on your PG, out of 10 times how many times does it completely throw off the rest of your lineup? For me it doesn’t mess it up more often than not unless a ridiculous lineup is being ran.


If we want it to start getting more complicated add a 3rd defence rating to this equation with Help D. If there’s a good help defender and a poor offensive player, put the good help defender there to help the 2nd best defender guard the best offensive player.

They have the tools in the game already they don’t need to restructure nothing, just put it to use.

I was going to respond with similar points so thank you for taking the time to do so lol. They literally have an ACE coaching feature that ensures there is always a good to great scorer on the floor at all times lmao. Why can't they duplicate this for defense, but add an additional layer that allows for matchup logic? I think it's def attainable, because as you said.. the ratings are already available.

[ggsimmonds]

Re: CPU’s knowledge on matching up…

I was going to respond with similar points so thank you for taking the time to do so lol. They literally have an ACE coaching feature that ensures there is always a good to great scorer on the floor at all times lmao. Why can't they duplicate this for defense, but add an additional layer that allows for matchup logic? I think it's def attainable, because as you said.. the ratings are already available.

Lets look at 2 case studies using his proposed perimeter/inside offense and defense.

The highest rated perimeter offensive player on the Clippers is...Norman Powell at an A+. So the AI is going to put their best perimeter defender on him. PG, Kawhi, and Harden all come in at A-.

Did yall see what Ja Morant did last night? In game Ja's perimeter offense is a B I think, but his inside offense is A+. Using his model, the AI is going to put Rudy Gobert on Ja.

To operationalize something is to take an abstract concept and make it something measurable. Its easy to agree on matchups at the abstract level, but to program it you have to take the abstract and convert it to an equation that consistently yields your desired result and that is not always easy.

I think the best approach is to use the player archetypes in conjunction with the touches tendency. Build a map of sorts that says defensive archetype X counters offensive archetype Y. If its not already present, there needs to be a scale (i.e. two players can both fit the same archetype but one is better than the other). Then have some code that says

if offensive player on court with {specific scoring archetype} AND {archetype aptitude} 80< AND {touches tendency} 85<

Then
Find is defensive player on roster that has {corresponding defensive archetype} AND {archetype aptitude} 70< AND {miscellaneous checks like player not in foul trouble, player is not fatigued, height difference does not exceed 5 inches, etc}

Then
Matchup that defensive player with the offensive player.

Where it could get complicated is if/when the defensive player that fits the requirements for matching up is a bench player. We could scale it down to have the search only use players currently on the court and just keep it at dynamic matchups, then in the future expand it to encompass dynamic rotations.

[FixEverything2k]

Re: CPU’s knowledge on matching up…

Lets look at 2 case studies using his proposed perimeter/inside offense and defense.

The highest rated perimeter offensive player on the Clippers is...Norman Powell at an A+. So the AI is going to put their best perimeter defender on him. PG, Kawhi, and Harden all come in at A-.

Did yall see what Ja Morant did last night? In game Ja's perimeter offense is a B I think, but his inside offense is A+. Using his model, the AI is going to put Rudy Gobert on Ja.

To operationalize something is to take an abstract concept and make it something measurable. Its easy to agree on matchups at the abstract level, but to program it you have to take the abstract and convert it to an equation that consistently yields your desired result and that is not always easy.

I think the best approach is to use the player archetypes in conjunction with the touches tendency. Build a map of sorts that says defensive archetype X counters offensive archetype Y. If its not already present, there needs to be a scale (i.e. two players can both fit the same archetype but one is better than the other). Then have some code that says

if offensive player on court with {specific scoring archetype} AND {archetype aptitude} 80< AND {touches tendency} 85<

Then
Find is defensive player on roster that has {corresponding defensive archetype} AND {archetype aptitude} 70< AND {miscellaneous checks like player not in foul trouble, player is not fatigued, height difference does not exceed 5 inches, etc}

Then
Matchup that defensive player with the offensive player.

Where it could get complicated is if/when the defensive player that fits the requirements for matching up is a bench player. We could scale it down to have the search only use players currently on the court and just keep it at dynamic matchups, then in the future expand it to encompass dynamic rotations.

There ya go put that in the game and let’s try it, if it doesn’t work let’s have an option to turn it off. If it’s in the game then maybe it can be tweaked with updates based on feedback.

Put in a watered down version, whatever, just say ‘hey guys this is a test feature’.

I feel like half the community won’t notice it’s in the game anyways so it won’t bother them.

[Smirkin Dirk]

Re: CPU’s knowledge on matching up…

I recall OG showing off how the CPU would match up Bogut on yo Tony Allen and just stand in the paint.

That was hardcoded; no procedural behind that.


Sent from my iPhone using Operation Sports