Is football's logic just too hard to put into a video game?

[RyanMoody21]

Quote: Originally Posted by Valdarez

I'm a little confused. You say you haven't played APF2K8, but you still find things that irk you?

Yea, Im with you this. Seems like Killroy is passing alot of judgment on how the game plays based off 2k5. I was skeptical about your experience with 2k8 when you talked about "players being stuck in animation".

While I have only played about half of the amount of games as Valdarez, very few times did I feel like I was stuck in an animation I shouldnt have been in to start with.

I hate to say this Killroy, but your very stereotypical of what I see in regards to 2k8. Guys that claim to have played the game, then come to find out they never did but have this predetermined mindset of all its flaws and errors. While it has its issues, most the user can "deal with" meaning you can take control of a defensive player and prevent / reduce the success of certain "cheese" plays.

[Valdarez]

To be fair, he's going by video footage, and he probably sees a lot of what he saw in 2K5 and draws the conclusion that it must be the same, but APF2K8 though a port to the next gen console of 2K5 had quite a lot improvements to the core engine coupled with a lot of bug fixes.

2K does use longer running animations though, this is something the Madden team latched onto and even used in one of their blogs to showcase their new blocking capabilities. There's only 1 animation I know of (reach tackle) that gets triggered in situations it shoulnd't, and it's a bug. Other than that, the animations always felt right.

One thing I would like to see with regards to the AI / Logic of the games is the ability to 'disguise' plays, which is something that players figure out and happens to some extent based on the capabilities of the engine (hot manning, routing, and such). Yet no game has provided the ability to disguise plays. APF2K8 does this to some extent with coverage shells, which are very nice, but I would like to see this concept taken to the nth degree so defense becomes more difficult to read before and after the snap.

[CreatineKasey]

Playcalling totally matters in M10 and APF.

You can't run certain routes against man and certain against zone. Double moves aren't worthwhile against zone in many cases. Slants and arrows are risky against zone. All of those routes rock against man. Fade is pretty risky against any defense that has a superior DB on the WR OR if the DB is backed up.

In M10, it's a little different. Basically, it's super hard to go vertical on zone defenses.. period. Underneath routes are there vs man and zone, but sometimes they take more time than is available to get open. Tighter throws are required to be made in M10 for the most part.

Regardless.. playcalling DOES matter and anyone discrediting that is gravely mistaken.

[Phobia]

Yea Killroy, how can you give a assessment on a game you have not played???

APF is similar to 2k5, but it is very different in terms of gameplay. The gameplay is very refined and plays a hell of lot more realistic on the field than any football game released EVER!

[johnkorean]

Let me preface by saying I know NOTHING about programming.

My gut is that football logic is possible, but that it's more difficult than any other sport. Take baseball for example. I get the feeling that this is comparatively "easy." First of all, the pitching/batting "phase" can be made separate from the fielding/baserunning "phase." Seems like it wouldn't be too hard to have the computer as batter run something like:

If (random integer 1, 100) > 60 (or something like this) OR if (the count and location of the pitch makes swinging sensible), then swing. Then, if (random integer 1, 100) > 50 (ish), then HIT!

I apologize to anyone that knows how to code for the above abomination, but I feel that the above attempt at representing logic is at the very least workable. Football just seems so much different, since on every play there are 22 individual players that have to make independent decisions. Not only that, they have to be making constant independent decisions based on what the other 21 players are doing. This is different from baseball, since your options as a fielder, for example, are "make a play on the ball" or "go stand on the base you're responsible for." Lather. Rinse. Repeat. Furthermore, you've got to account for the fact that a. some players are going to make wrong decisions, and b. the fact of a. means you need to be concerned with exploits. So if you code in that defenders are going to guard the flats, maybe the human controller will call plays to exploit this and hit crossing patterns all day long.

Like I said, I think incorporating football logic into a game is possible; I just think it's such a difficult problem to solve that we're not going to be there for a while.

[KBLover]

Quote: Originally Posted by johnkorean

If (random integer 1, 100) > 60 (or something like this) OR if (the count and location of the pitch makes swinging sensible), then swing. Then, if (random integer 1, 100) > 50 (ish), then HIT!

50% chance of getting a hit on a ball in play? Who's batting? Ty Cobb?

Quote: Originally Posted by johnkorean

I apologize to anyone that knows how to code for the above abomination, but I feel that the above attempt at representing logic is at the very least workable. Football just seems so much different, since on every play there are 22 individual players that have to make independent decisions. Not only that, they have to be making constant independent decisions based on what the other 21 players are doing. This is different from baseball, since your options as a fielder, for example, are "make a play on the ball" or "go stand on the base you're responsible for." Lather. Rinse. Repeat.

Baseball is different in that the decisions can be separated as far as the game goes because things don't happen as quickly or don't have to be "checked" as often. For example, fielders also have to decide which base to throw to, whether or not to cut off a throw, whether or not to try a diving catch. But those things can be checked when needed. If the fielder with the ball just throws it back in, no need to see if the relay throw is good. You can check if the runner decided for a double or triple only when the situation needs it (like a ball off the wall vs a grounder up the middle)

In football, everything is potentially needed because the man with the ball (the QB) can make a multitude of possible choices. The DBs always are looking at the WR or the QB (man vs zone) Blocking is always taking place.

Quote: Originally Posted by johnkorean

Furthermore, you've got to account for the fact that a. some players are going to make wrong decisions, and b. the fact of a. means you need to be concerned with exploits. So if you code in that defenders are going to guard the flats, maybe the human controller will call plays to exploit this and hit crossing patterns all day long.

Players make wrong decisions in both sports, but in baseball, the decisions don't often come all at once. Mainly when the ball is in play is when the most simultaneous decisions are made and even then, it's usually just a couple players at the given moment with key decisions. Sometimes, there's no decision at all, basically.

Football, decisions are coming constantly as there's always movement and act/read/react going on. So there's more places too check for right decisions, right reads, and proper actions, and typically, it's all 22 guys needing this. That's where it can get complicated.

The play call part might actually be the easiest part to program. Much of it is situation. If it's 3rd and 20, the defense isn't going to be looking run. If it's 4th and 1, you probably won't be seeing a cover 3, especially if the other team has Brandon Jacobs in the backfield.

The rest is tendency and personnel. Again, this should be easy. Personnel can be simply a check of ratings. If I have 3 90+ OVR WR on the field, they should be calling pass defenses frequently except in obvious run situations, and even then, if my 3 WR are on the field, they can't choose a sell-out run unless they are gambling. Likewise, if I have JaMarcus Russell and mediocre WR - they shouldn't fear the pass unless I prove I can beat them with it.

That's where tendency and results come in. Instead of/in additional to MySkill, the game should track player playcall's over their profile history. This could help the AI "scout" the player. So they'd know that if I throw to the flats a lot, call defenses against that until I show in-game I'm beating them with something else (the results part). If I can't handle the blitz, keep blitzing, but if I start beating them over the middle, back off it (or bring it from the edge instead of from the LBs, or bring 1 blitzer instead of 2 or 3, etc).

Of course, to do this, Madden and gamers need to get away from accepting having the AI just know your play call (because it's "challenging") and determine it by situation, results, and personnel like the human player has to. We need to stop accepting cheating AI and want true intelligence in the AI.

[RogueHominid]

Something I'd be curious to know that would definitely inform debates on this subject is how different levels of NFL players actually grade out over the course of the season.

We all know that players get a full report on their snaps and assignments after each game--that they're graded out. So how often does a Jake Long do his assignment correctly compared to a Jason Peters or a Eugene Monroe.

We die-hards tend to think that if you have a good rating you should do what the play says to do almost every time (and we cry foul if it doesn't happen). But my hunch is that top-tier NFL players are grading out in the 80th percentile.

My kid brother is a trainer for DEN, so I could probably find out through him how different players on that team grade out.

Knowing that info would probably put our expectations in proper perspective when it comes to performing assignments and missing them.

If, for example, you have a DE lined up wide and an OLB just inside him, it might be possible for the LT to get legitimately confused or misread the drop of one, misread a fake blitz, etc, such that he just whiffs and lets a guy in. When this happens in the game many folks lost it and cry about the game being broken. But if we have more data on just how often guys get it right in the league, it might change our perspective some on these types of plays.

[KBLover]

Quote: Originally Posted by KillRoy

Of course, one could argue that this is realistic, real NFL QBs are capable of getting a read on the defense before the snap. However, the problem could be corrected by simply taking out the ability to call hot routes at will. In all honesty, the ability to change the route at will kind of makes it too easy to play offense.

I would be one to contend the hot routes are realistic and I want realism over challenge.

Like you said, QBs can read stuff and signal/change the play at will, including giving an indication to a WR to change his route to something else, say to beat a blitz or if he sees the DB back off or move tighter pre-snap.

In fact, it could go both ways. Defenses could do this to try to "bait" a hot route that would end up going right into the teeth of the called coverage. It would give a tactical purpose to "show blitz" instead of hoping to make the LBs play shorter hook zones.

Also, defenders with high AWR should have a chance to recognize when a signal has been made and have a better chance of reacting to the hot route successfully.

Play flipping wouldn't be necessary if we could play call formations better. For example, there's nothing unrealistic about me running trips left, then "flipping it" to call trips right. The thing is, the play wouldn't just be a mirrored version - the routes could be different. The "hot" receiver could be different. But in Madden, it's just a mirror of the same play. That's what I find limiting and unrealistic.

Quote: Originally Posted by KillRoy

Why not implement a certain tactical advantage to running up the middle against a 4-3 but make it harder to run off-tackle against 4-3. And make it harder to run up the middle against 3-4 but easier to run off-tackle against 3-4. that is actually somewhat realistic because on a DIVE play the center usually blocks the middle linebacker when the defense is in 4-3 and this creates a bubble in the middle. But in 3-4 the nose guard cuts off any runs up the middle. However, its easier to run off-tackle against 3-4 because in most situations, the tight end has an advantage over the outside linebacker and the OT can get a good push on the DT thus making it easier to run off-tackle against a 3-4.

I think that advantage would depend on the TE and if he can block, as well as who that OLB is. I wouldn't make it a "most of the time" advantage, but base it on skills. The TE has an easier time locking in on the OLB, but the result of that engagement would depend on the OLB and the TE's blocking/shedding blocks. Not to mention what Mike or Jack are doing. If the defense is thinking outside run, they might have spread the LBs and aware ones could go outside on their run coverage if they read outside run.

I think 3-4 needs the most help with this tactical aspect since it has the most options for what the LBs can do and what the OG can/need to do. Probably why it doesn't work like it zone both for it's strengths and weaknesses.

Quote: Originally Posted by KillRoy

Implementing this into the game would completely change the game and make it more of a chess match than what we have now because it forces the guy playing defense to make on decision on who to control before the snap and it can also confuse the offensive player and make it harder to guess what type of play his opponent called. Like, if he takes control of the the strong safety and stacks the line to cover the run, is he going to blitz or drop back in coverage?

Definitely.

Quote: Originally Posted by KillRoy

Again... these subtle changes would create a whole new dynamic in the game.. I also think they should take one idea from MLB: THE SHOW. Not sure if you've played it, but in that game they have whats called "guess pitch" and if you guess right its usually a hit. In a football game, I think you should have a "guess play" for the defense and if you guess the correct play the offense called it results in either a sack, interception, fumble or just a failed play.

I disagree. I don't like these "automatic success" features in sports games because there's no such thing in reality. I never used Guess Pitch because it's too strong. A batter can look for a pitch (guessing) and get that pitch, but it doesn't mean an automatic hit much less a hard hit. Looking for a pitch makes it a bit easier for a hitter to hit that pitch, but even balls down the middle aren't automatic hits.

What successful "guessing" should do is make it easier for defenders to get in position, based on what was guessed. So if you guess deep pass and it was a deep pass, that should make the safeties go into position for accurately. It should NOT be an auto failure for the offense. Being in better position makes it easier to play defense, so that's where the advantage is, but there's no such thing as automatic success or failure on the field, imo.

Quote: Originally Posted by KillRoy

Did anyone watch the Steelers and Broncos game tonight? Remember that one play when Rothlesburger scrambled around in the pocket and bought himself an extra second or two and nailed a wide open reciever in the end zone? Doing stuff like this just doesn't work when you're stuck in animation. I could give you so many examples of stuff we see in REAL football that you NEVER see in a videogame. And its because the games have no sliding effect and the animations prevent you from moving real fluid like.

Stuff like that doesn't work in Madden for a lot of reasons. One is like you mentioned you can't move too much in the pocket unless you hit sprint to move faster, but then you can't make those subtle slides and moves. Plus, lineman aren't typically that aware in Madden. Rothlisberger slid around, but the OL had to maintain the blocks they did have. In Madden, they'd all have likely gotten beat by then and he'd have 3-5 guys on top of him.

Not to mention if that's against man coverage - the whole defenders stick to your every move thing can be there also and in zone, you'll likely get the super jump to tip/pick the pass as well, especially if the QB is shuffling his feet a little - the bad accuracy because he's "on the run" system kicks in too easily. It's not just about the feet, but the whole position of the body, like on that pass in the dirt to Royal when Orton was getting pressured by a free runner on the blitz. Not just his feet, but his whole body was falling away, making the pass bad.