Home

Why the narrow-sighted "AI" is a problem on Madden and NCAA

This is a discussion on Why the narrow-sighted "AI" is a problem on Madden and NCAA within the Madden NFL Old Gen forums.

Go Back   Operation Sports Forums > Football > Madden NFL Football > Madden NFL Old Gen
MLB The Show 24 Review: Another Solid Hit for the Series
New Star GP Review: Old-School Arcade Fun
Where Are Our College Basketball Video Game Rumors?
Reply
 
Thread Tools
Old 07-30-2008, 10:00 PM   #17
Banned
 
OVR: 4
Join Date: May 2008
Re: Why the narrow-sighted "AI" is a problem on Madden and NCAA

Quote:
Originally Posted by TheFuture15
All good points.

Out of sheer curiosity (I'm not trying to deflect here), are there any other sports games out there that you all feel have best-of-breed AI? Before I joined EA I was always kind of in awe that they were able to correctly model football so well (since it's by far the most complicated sport), but as we all know as the visual fidelity increases in games, so do the expectations of every other part of the game (i.e. animation, AI, lighting, etc)

I'd like to see what everyone thinks so I am not missing anything and I can research them further.
Are you arguing that since other sports games don't have state of the art AI that Ncaa should'nt? aren't ncaa and madden two highly lucrative franchises? surely they are to be held to a higher standard, or maybe thats just me that does that. No hostility here, just how I'm seeing things. You guys should be pushing boundaries not looking back at whats behind you, don't worry about other games, make yours better. Your point about football being by far the most complicated sport is noted though, this is true, that doesn't mean it isn't feasable to be able to accurately replicate the intricacies of the game if you so desired.
vsoul86 is offline  
Reply With Quote
Old 07-30-2008, 10:05 PM   #18
EA Game Changer
 
OVR: 14
Join Date: Sep 2005
Location: Calgary, Alberta
Blog Entries: 3
Re: Why the narrow-sighted "AI" is a problem on Madden and NCAA

Quote:
Originally Posted by vsoul86
Are you arguing that since other sports games don't have state of the art AI that Ncaa should'nt? aren't ncaa and madden two highly lucrative franchises? surely they are to be held to a higher standard, or maybe thats just me that does that. No hostility here, just how I'm seeing things. You guys should be pushing boundaries not looking back at whats behind you, don't worry about other games, make yours better. Your point about football being by far the most complicated sport is noted though, this is true, that doesn't mean it isn't feasable to be able to accurately replicate the intricacies of the game if you so desired.
I think your mis-reading him. He's just thinking of a game that he can use as reference. Nothing wrong with that. I understand your point though that those guys needs to be the innovators and be the game others look towards.

However, there's no game I can think off top of my head. EA NHL kinda trys this adaptive AI idea, but that's when a player keeps doing the same thing over and over again (overall defense is as bad as CFL defense). However, it does carry over games so if your in a playoff series, the opposing team will adapt to your play style over the series. Madden should at least have that.

Last edited by Joborule; 07-30-2008 at 10:10 PM.
Joborule is offline  
Reply With Quote
Old 07-30-2008, 10:16 PM   #19
MVP
 
OVR: 52
Join Date: May 2008
Blog Entries: 7
Re: Why the narrow-sighted "AI" is a problem on Madden and NCAA

Quote:
Originally Posted by vsoul86
Are you arguing that since other sports games don't have state of the art AI that Ncaa should'nt? aren't ncaa and madden two highly lucrative franchises? surely they are to be held to a higher standard, or maybe thats just me that does that. No hostility here, just how I'm seeing things. You guys should be pushing boundaries not looking back at whats behind you, don't worry about other games, make yours better. Your point about football being by far the most complicated sport is noted though, this is true, that doesn't mean it isn't feasable to be able to accurately replicate the intricacies of the game if you so desired.
I'd say you are a half-empty kind of guy?

I even posted it in there...not trying to deflect...I just want to keep the discussion going. I mean come on...where in my post am I "arguing" that we don't have to have better AI? In asking for examples, it doesn't at all mean I want to steal their ideas or follow their lead, I just want to hear other examples of games that people already consider to have non "narrow-sighted" AI.
Ian_Cummings is offline  
Reply With Quote
Old 07-30-2008, 10:18 PM   #20
Pro
 
Sigma4Life's Arena
 
OVR: 9
Join Date: Oct 2003
Location: Texas
Re: Why the narrow-sighted "AI" is a problem on Madden and NCAA

Quote:
Originally Posted by TheFuture15
All good points.

Out of sheer curiosity (I'm not trying to deflect here), are there any other sports games out there that you all feel have best-of-breed AI? Before I joined EA I was always kind of in awe that they were able to correctly model football so well (since it's by far the most complicated sport), but as we all know as the visual fidelity increases in games, so do the expectations of every other part of the game (i.e. animation, AI, lighting, etc)

I'd like to see what everyone thinks so I am not missing anything and I can research them further.
Konami's Winning Eleven / Pro Evolution soccer has excellent AI with very few holes. For example, the players have knowledge of whether they are winning or losing and adjust accordingly. I'd be surprised if there is a single line of code in Madden that addresses this. I've never played a full season in Madden or NCAA due to the numerous AI issues but I absolutely love playing against the CPU in Winning Eleven because it replicates the real deal so accurately.

In response to the other poster, I don't think the AI problem in Madden has anything to do with processing power. The problem is just as the OP said, the AI logic is too narrowly focused on following assignments instead of understanding the bigger picture.

Last edited by Sigma4Life; 07-30-2008 at 10:21 PM.
Sigma4Life is offline  
Reply With Quote
Old 07-30-2008, 10:19 PM   #21
Pro
 
rhombic21's Arena
 
OVR: 9
Join Date: Jul 2002
Re: Why the narrow-sighted "AI" is a problem on Madden and NCAA

Yeah, I don't know that any other games have it, and I understand that it's probably pretty difficult to do, but it's pretty vital.

I have an idea though, that might get us closer even if it doesn't actually do the whole thing. On offense, the CPU has priorities assigned to routes which allow the CPU QB to determine what to do if it's first option is covered, correct?

What about hard-programming in options for every player on defense to know what to do, beyond their initial command? For instance, let's look at the plays that I showed in that video. The first is a basic Cover 3 zone defense out of Dime. So the flat zone defender, for instance, could have the assignment to go to the flat if there are any potential threats to that area. Once he drops to his area, he then runs some sort of radius/route check in which he surveys the field to look for what he should be guarding. Or in the second video, the player is guarding the hook area, which is threatened initially by the receiver running the drag, but is then vacated when the receiver continues past the hash marks. In that case, his first priority is to guard the hook, but then when he sees that there's nobody near his hook zone area, he abandons that assignment and looks for the nearest receiver to guard/defend. And rather than just having generic options for every assignment, maybe even go a step further and have the playbook guys program them in separately for each specific play in the game -- since they know what the scheme of the defense is. In essence, you're doing what a real coach would do. You're effectively telling the players, "____ is your first priority. Do _____ if they do A, but do ____ if they do B instead."

I'm assuming the game has some capacity to do this as you'll see that zone players will match up to receivers in their zones -- the issue is that they just have a very small window in which they'll do so. So then you could program in alternate assignments that he should do if there is no threat to his first priority. So in the case of the flat defender there, he would go to the flat area, and then when nobody is out there and there are no threats, he then goes to his second option, and looks for a crossing/drag route and works his way back towards the hash/hook area.

Then maybe you implement some sort of design where players can potentially "mess up" these checks if they have low zone coverage ratings, while players with high zone coverage ratings can make decisions on the fly really quickly. So now you have the ability for low rated players to get confused and to have truely random busted assignments that aren't cheap and don't happen universally/predictably. So maybe if I have a really low rated player gaurding the flats, he doesn't go to his second option until very late, or on the other side, maybe he gets too aggressive going to the second option and misses a HB coming out of the backfield.

Just a thought.

Last edited by rhombic21; 07-30-2008 at 10:22 PM.
rhombic21 is offline  
Reply With Quote
Advertisements - Register to remove
Old 07-30-2008, 10:20 PM   #22
Pro
 
OVR: 8
Join Date: Sep 2003
Blog Entries: 1
Re: Why the narrow-sighted "AI" is a problem on Madden and NCAA

Ian, I agree with you that EA has done a great job with the AI for computer game. From previous post I have learned how important animation plays in making the game play look real and the A.I. to choose the “right animation” makes a big difference. You have stated that the animations for line play are going to increase (I hope dramatically) which will help the decisions on A.I. game play. I think in addition you can add assigned blocking schemes a utilities zones and match ups. On trap blocks plays the key element of this play is the use of a pulling lineman or linemen. The play creates a deception that a defensive player will be unblocked as he penetrates the line of scrimmage. The play unfolds when the pulling lineman blocks, or "traps" the player to open the hole for the running back to run through. The Pulling lineman can include a guard (on either side of the play direction), the guard and tackle, or even both guards. (For simplicity, utilize 1 pulling guard.)

BLOCKING SCHEMES
Typical blocking for a trap play includes a double team at the hole where the ball will go, the pulling lineman or linemen, and the block on the play side linebacker. The double team helps to create a large hole for the running back(s). This double team, coupled with the trap block, can create quite a large hole to run through, provides that the blocks are successful. The block on the play side linebacker will determine where the back should run. The following rules apply for the pulling guard:
• For inside traps, the guard blocks the first man past the center.
Example: If the defensive player lines up in the A Gap, then the pulling guard will trap that player. Similarly, if a linebacker stunts between the Nose and Tackle, then the pulling guard will trap the linebacker.
• For outside traps, the guard will block the first man past the tackle.
Example: If a defensive tackle lines up in the B Gap, and the Outside linebacker is in the C-Gap, then the pulling guard will trap the linebacker. (qouted from Football.com)

By assigning zones and match up A.I. to the game I think you could achieve a lot more realism.

I think it would be impossible to do this concept for all positions in one year. Please start with the offensive line schemes/defensive line schemes, then work on linebackers & Db’s the following year.

Easy for me to say.

Last edited by Agame; 07-30-2008 at 10:35 PM.
Agame is offline  
Reply With Quote
Old 07-30-2008, 10:31 PM   #23
MVP
 
OVR: 13
Join Date: Dec 2003
Re: Why the narrow-sighted "AI" is a problem on Madden and NCAA

Quote:
Originally Posted by Sigma4Life
Konami's Winning Eleven / Pro Evolution soccer has excellent AI with very few holes. For example, the players have knowledge of whether they are winning or losing and adjust accordingly. I'd be surprised if there is a single line of code in Madden that addresses this. I've never played a full season in Madden or NCAA due to the numerous AI issues but I absolutely love playing against the CPU in Winning Eleven because it replicates the real deal so accurately.

In response to the other poster, I don't think the AI problem in Madden has anything to do with processing power. The problem is just as the OP said, the AI logic is too narrowly focused on following assignments instead of understanding the bigger picture.
Winning Eleven absolutely. If I am destroying a defender (beating him almost every time down) the CPU will usually sub him out. They will adjust formations based on game score and my method of attack. If I am beating them down the sidelines, they will look to defend the sidelines. If I have holes in the middle of my defense, they will go up the middle. As a result, scoring goals and staving off attacks is unbelievably satisfying.
rsnomar05 is offline  
Reply With Quote
Old 07-30-2008, 10:33 PM   #24
49ers
 
adembroski's Arena
 
OVR: 43
Join Date: Jul 2002
Location: Prescott, AZ
Posts: 5,793
Blog Entries: 20
Re: Why the narrow-sighted "AI" is a problem on Madden and NCAA

I think most of this stuff, beyond simply shoving a guy out of bounds, needs to be incorporated into playbooks.

Assignment logic needs to be more complex, incorporating various potential events during a play.

Take, for example, man coverage vs. twin WRs. In many cases, and this is usually tied to the playcall, the corners will switch on a cross, even through it's not zone. In some cases, one will take whoever releases down field.

Then you have the running game, where pretty much ever player on the field is programmed to 'get to the ball carrier'.

What should be done is to have defensive linemen fill gaps and keep blockers occupied. Linebackers check gaps and slide down the line, while cornerbacks play outside containment and funnel anything outside back in toward the backers. This is all variable, of course... we've all heard of Tampa's "play the run on the way to the quarterback" mentality, meaning defensive linemen wont bother with tying up blockers, but rather get aggressively upfield, but that requires it's own specific logic.

It's this type of programming that will finally have screens and draws work correctly. This type of programming that will allow counters and misdirection plays become the cornerstones of some player's games. This type of programming that will finally allow trick plays to be the kind of boom-or-bust plays they should be.

A perfectly defended run play has the back see he has nowhere to go on the interior, because the D-Linemen are closing interior gaps, so he starts outside only to find himself unable to get outside the corner... who's holding his ground in the backfield, cutting off the sideline (NOT charging in for the tackle) and the back being forced to cut upfield into a wall of unblocked linebackers and safeties.
__________________
There are two types of people on OS: Those who disagree with me, and those who agree.

The first kind is wrong. The second is superfluous.

The only difference between reality and fiction is that fiction needs to be credible.
-Mark Twain.

Last edited by adembroski; 07-30-2008 at 10:37 PM.
adembroski is offline  
Reply With Quote
Reply


« Previous Thread | Next Thread »

« Operation Sports Forums > Football > Madden NFL Football > Madden NFL Old Gen »


Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is On
Trackbacks are Off
Pingbacks are Off
Refbacks are Off



All times are GMT -4. The time now is 09:48 AM.
Top -