NBA Elite 11 Video: Real AI

[ehh]

Quote: Originally Posted by rEAnimator

From a technical standpoint though, mocapping these sequences as animations would be impossible both from a money standpoint and a memory standpoint. It is financially prohibitive to fly out every NBA star to do massive amounts of moves in all possible situations. And from a memory standpoint it is absolutely impossible to fit that much animation on current generation consoles. By recording inputs, we can record 100 times more data for all the contexts needed, and do that for 1000 times more players than if it was animation. We can also get people to mimic real life player tendencies a lot easier by using the joystick, and in the case of Kevin Durant, get the real life players to do it.

Thanks for spending some time to come in here and explain this a little more.

This quoted part is where I'm still a little lost. Since everything you said is true about resource limitations, what is the source for all these recordings? Obviously you had KD up as the coverboy to record his stuff but what about everyone else in the league? Are you guys going through game footage, scouting reports, etc? So will you have some guys in the lab watch Melo or LeBron game tape and then do their best to mimic their game with the sticks in Elite and record that data? Based on what you're saying I think that's how it'll work.

Just curious as to how this will be implemented across the board.

[HMcCoy]

I love the concept. Thats exactly how I try to play...signature sequences of moves guys actually do, from the correct spots on the floor.

Looking forward to this...

[michaeljordanjr]

stepsix answered my concern about whether its dynamic enough to change the script.

Because I can foresee people memorizing the chained moves and anticipating them if the A.i. player uses it over and over.

I think Real AI is a great addition to basketball games and one that is long over due. Instead of predictable, generic AI behavior we get signature and dynamic behavior.

Despite the unrealistic dunk in the video, Elite is shaping up nicely.

Now for some HD 5 on 5 CPU vs CPU footage.

PLEASE - No recording a random guy playing the game, if you're recording a user player make sure its someone who KNOWS the controls and the game. Unlike the other NBA game's user footage where the guys were running around like headless chickens not knowing anything, I'd like Elite videos to look like actual basketball players playing basketball.

[magicman32]

Quote: Originally Posted by JerseySuave4

why would they ever put that sequence from 1:09-1:16 in this video? Usually in a video that you're putting together to hype up your game you won't put something so blatantly unrealistic or screwed up in there. The guy does a standing dunk from the nearly the top of the key on the right side. If he can do that then imagine what he could do in a dunk contest. Forget dunking from behind the free throw line, that guy would be doing it behind the 3 point line. I just hafta shake my head at this video because i dont understand why they would put things in their video. That were so ugly. I'd rather real AI use the animations, physics and movement that the human players have to use to play the game instead of having stuff scripted.

yeah. Live has had a problem with momentum and weight for a while in their games. Some of their animations for certain players are off. Like they are superhuman.

They need to improve on this big time.

[PVarck31]

I get it now. It makes sense. Lets just hope it works well. I still want to see where they are on physics though.

Oh and rEAanimator, or stepsix, could you give us an update on the sliding problem? As someone mentioned there was an issue where the foot planting software was fighting the physics.

[stepsix]

Quote: Originally Posted by ehh

Thanks for spending some time to come in here and explain this a little more. This quoted part is where I'm still a little lost. Since everything you said is true about resource limitations, what is the source for all these recordings? Obviously you had KD up as the coverboy to record his stuff but what about everyone else in the league? Are you guys going through game footage, scouting reports, etc? So will you have some guys in the lab watch Melo or LeBron game tape and then do their best to mimic their game with the sticks in Elite and record that data? Based on what you're saying I think that's how it'll work. Just curious as to how this will be implemented across the board.

That's exactly how we do it - we need base layers of sequences to cover all game situations, and then to provide differentiation for players that have distinctive styles, we study tape and record specific sequences from the areas on the court that they typically get the ball in. You can see how this would then work with our play system...if a play has Melo coming off of a screen for an iso on the wing, we would then record Melo specific sequences from that position.

Quote: Originally Posted by michaeljordanjr

stepsix answered my concern about whether its dynamic enough to change the script. Because I can foresee people memorizing the chained moves and anticipating them if the A.i. player uses it over and over. I think Real AI is a great addition to basketball games and one that is long over due. Instead of predictable, generic AI behavior we get signature and dynamic behavior. Despite the unrealistic dunk in the video, Elite is shaping up nicely. Now for some HD 5 on 5 CPU vs CPU footage. PLEASE - No recording a random guy playing the game, if you're recording a user player make sure its someone who KNOWS the controls and the game. Unlike the other NBA game's user footage where the guys were running around like headless chickens not knowing anything, I'd like Elite videos to look like actual basketball players playing basketball.

Not to worry - we've been practicing all year with the controls - any sequences that go in get reviewed and pulled if they don't look right.

[Jano]

Now thats great to hear stepsix this is exactly what this series needed should add a much needed facelift for the AI.

Should make playing against the AI a lot more fun and definitely make each game feel different.

Does this system apply to the off ball players too? So will guys like Ray Allen be hangin out on the perimeter? And what about the perimeter shooting big men like Troy Murphy?

[TUSS11]

Quote: Originally Posted by stepsix

That's exactly how we do it - we need base layers of sequences to cover all game situations, and then to provide differentiation for players that have distinctive styles, we study tape and record specific sequences from the areas on the court that they typically get the ball in. You can see how this would then work with our play system...if a play has Melo coming off of a screen for an iso on the wing, we would then record Melo specific sequences from that position. Not to worry - we've been practicing all year with the controls - any sequences that go in get reviewed and pulled if they don't look right.

How difficult is it to predict what Melo would do in this situation? Say he starts a jab --> pump-fake --> jab --> shoot combination and I've played the game enough to predict that this particular sequence will end with a shot. Will he diverge from the sequence and attack off the dribble instead of shooting if he sees I'm playing him too tight? Not by me jumping or getting out of position, but by me simply applying too much pressure?

I'm just afraid that by recognizing the sequence I'll be able to play up on him and time the block because I saw it coming.