Warrior General Games

There is, in my opinion, no AI problem more urgently in need of being addressed than those exposed by the screen passes, specifically:

1) blocking in the open field, which is unrealistic
2) defensive logic in open field situations, which is idiotic in several ways

These two problems feed off each other, and are intertwined, so it is impossible to address one fully without addressing the other.

Fortunately, a solution exists which is, I think, easy to implement. I will discuss other solutions which are far more complex, but my suggestion is to use the easy solution, which I think can be implemented quickly.

My suggested solution is:
1) To define an area called the "box area" which will extend from the outside point of the right side TE would line up (if there was one) to where the left side TE would line up (if there was one), and extending 5 yards upfield and downfield from the line of scrimmage. On a regular pro football field, this area would be roughly bounded by the hash marks and extend +-5 yards from the line of scrimmage.
2) That outside the "box area" all blocking uses pass blocking logic (ie players on the wings and downfield try to shield defenders from the player with the ball but do not drive block the defenders).
3) That defenders who have used any sort of backward "angle of pursuit" logic on a play will afterwards assume a much more conservative tackling approach than normal and will be much more likely to make a tackle at the cost of being much more likely to be dragged for extra yardage and of having almost no chance to force a fumble (exceptions: inside their own 3 yard line and perhaps down late in the game).

Another idea which I think goes well with the above, but which is not essential is to
4) remove the current run focus / pass focus choice for blockers and replace it with a strength focus / agility focus. A strength focus would give a bonus to a blocker's effectiveness when he has a sizable strength advantage over his opponent, but a negative when he faces a strength disadvantage, while an agility focus would give a bonus when a blocker has an agility advantage over his opponent, but a negative when he faces an agility disadvantage. No focus would be neutral.

So those of you who don't like to read a lot can stop here. For those interested in the details, and other, more complex possible solutions, here goes:

I was watching a football game a couple of weeks ago and I noticed a tackle trying to block a cornerback on a screen pass. The announcers commented on how difficult it is for such a big, heavy player to try to block such a quick, fast player in the open field. This started me thinking about the GLB screen plays, where I regularly see big heavies pancake 2 or 3 much quicker defenders at a time on screen plays.

The problem, I recognized, was that once the pass is caught, the defender runs straight for the man with the ball, without any attempt to avoid blockers. Blockers calculate an intersection route and so usually get to the defenders. At this point the defender tries to go through the blocker, and the blocker tries to go through the defender, and big, heavy tackles usually pancake fast cornerbacks, and then start eyeing the next defender to block.

This got me thinking about all sorts of elaborate logic that could be written where the defender would take a path so that, if he is much faster than the blocker, he tries to avoid the blocker, and in any case doesn't try to go through the blocker, so he's still standing and able to make a tackle once the running back who caught the screen tries to get downfield. Meanwhile, I thought, blocking logic could be added so blockers would try to seal off the defenders from the running back. And there would need to be a chance that the quick defenders could get past the attempted seal.

As I thought about the blocking additions, I started to recognize that these are already what blockers do with pass blocking logic. Then I started to think that if the blockers would only pass block to begin with once they were in the open field, that there wouldn't really be any need to write all the fancy logic for defenders I had thought up. The defenders could go ahead and run straight at the ball carrier (which isn't completely realistic but it's easy to implement since there's no code that needs to be written to change defensive paths). The pass blocking logic will effectively simulate the defenders and blockers squaring off on the open field and the defender trying to slip by while the blocker tries to seal him off. It also will allow the defenders to avoid being pancaked and be in position to at least make the tackle 5-10 yards downfield in most cases, which is what defenders would do in real life. Screen passes would still be effective against many defenses, but wouldn't go for so many huge gains.

I also thought about how this same logic should be applied to defenders playing 10 yards downfield, and that's when I came up with the idea of the "box area". Basically, any time a defender is in the open field (outside the "box area"), he gets the benefit of the extra space and blockers are only able to seal him off not try and drive block him resulting in unrealistic pancakes in wide-open field against much quicker players.

Naturally, I also considered special teams plays like kickoffs and punts. I thought, should the same sort of logic apply? My thought - heck yea! It makes no sense seeing 2/3 the kicking team getting pancaked on every kickoff or punt play, then having to get back up and make a tackle. That's completely unrealistic. I think that once the ball is kicked, the "box area" should be completely removed from kickoff and punt plays and ALL blocking from that point can only be pass blocking. (Actually this change is not mandatory for my idea so I don't want to push it too hard but just wanted to state my thoughts and if these are left as they are now that's ok too - let's focus on fixing the biggest problems which are on non-special teams plays.)

As I was thinking of any impacts to the game, I thought of the need to get rid of the current choice of run block / pass block focus since, with this change, much more blocking would be pass blocking. However, I liked the idea of changing the choice to strength focus / agility focus / no focus. Suppose as a result of scouting a tackle is expected to go up against a very quick but not very strong DE. The tackle could go into the game with a focus on trying to outmuscle his opponents and keep them neutralized that way. He would have worked on that focus all week during practice. He is prepared to have improved blocking against opponents where he has a strength advantage. So on most plays when he blocks that weak DE he gets a bonus. However, if on one play he gets matched up with a huge, super-strong defensive tackle he won't have prepared for that during his practices and on that play he'll have a disadvantage. It think this is a good way for offensive players with all-around builds to scout and help neutralize defenders with extreme, imbalanced builds, just like would be done in real life.

As I continued to think about all this, I realized that there was still one issue that I saw on so many screen plays that needs to be fixed. Time after time, I would see the halfback running down the sideline after catching a screen pass, and defenders would take a careful pursuit angle to cut him off and get in a position between the running back and the goal line, then once they got close, instead of forcing the running back out of bounds or just trying to delay him so other defenders could catch up, the defender in good position would try an aggressive tackle and many times would miss that tackle and allow a touchdown or many more yards.

The more complicated solution I thought up was to have defenders recognize if they are a "last man" or "last few men" and also recognize sideline proximity to use that to their advantage and then give a little ground to effect a force out or else cause the runner to need to either try to cut back horizontally into the field of play or run straight through the defender with little to no chance of doing a fake, etc. However, I saw that could be complicated.

So then I recognized that defenders in this situation had almost always taken a backwards angle of pursuit to get in this great position, and it hit me that that was a great trigger! If a defender goes to all the trouble to give up ground taking a negative angle of pursuit then that defender is NOT going to suddenly become aggressive again and fail to make the careful but conservative tackle. The defender is no longer in a normal situation where he is attacking the ball carrier. He is trying to cut off a touchdown or big(ger) play and will use much different tactics when tackling. He will not advance on the running back, so will be very unlikely to fall for any fakes or jukes or spins. Instead he will happily give a bit more ground and use the sideline to his advantage plus anything he can do to slow the running back down so other defenders can catch up. So I thought the easiest way to code the effect of all this is for defenders who have previously taken a negative pursuit angle to receive a significant bonus to tackling and fake avoidance at the cost of giving up a little extra yardage very easily and of having negligible chance to force a fumble.

Of course there are some situations where a defender might still want to be as aggressive as possible when tackling, including:
1) inside his own 3 yard line
2) when down on the scoreboard late in the game

Another possibility for my list is when a first down is on the line but I see many real-life defenders who still play it safe even if they have to give up a first down (although the announcers invariably criticize this choice) so I didn't add that to my list.

I think it would be enough if defenders inside their own 3 yard line or after a certain point in time late in the game if down on the scoreboard would always try to tackle aggressively like they do now, and these were the only two exceptions. Perhaps someone else can add another exception I've failed to recognize.

Ok, that's it. I've thought this through and considered sophisticated logic but I'm comfortable that the comparatively simple adjustments I've recommended here will fix the problem with minimal pain and maximum benefit. I hope you agree but welcome all comments.

Selected ideas from other posters

Overall, there seems to be some sentiment that the suggestions above could be a step in the right direction. But they would not be the final solution and, once they are implemented, it would be easier to get to the "next step" items which also need some work

Things that could be improved by the suggestions above
Better simulating the difficulty of open-field blocking and that defenders should generally win the open field battle (suggestion 2)
Simulating how in the open field blockers often just try to act as a "shield" of sorts (suggestion 2)
Somewhat countering the too-strong combos of Outside Blocker VA and Downfield Blocker VA (suggestion 2)
Making it harder for offensive blockers to take advantage of defenders' poor pathing (suggestion 2)
Simulating somewhat how defenders will take out an offensive blocker and make him unable to block another defender (suggestion 2)
Simulating how a defender will break down in order to have a tackling advantage and prevent a breakaway (suggestion 3)
Simulating how a defender who breaks down will have less chance to force fumbles (suggestion 3)

Things that will not be improved by the suggestions in this thread and will need additional AI work later to improve
Forcing RB back towards the middle of the field (tragula, staz)
Some existing SA's and VA's can give a bonus to pass blocking and don't anticipate this change, so they won't make sense (and will probably need to be modified) (tragula)
RB's doing a better job of setting up blockers by slowing down or cutting back (staz, cherrie)
RB's have more room to maneuver in the open field and should take advantage by doing a better job setting up fakes, spins, etc. (staz)

We may have to send Bort a "bulleted summary" but this is well worth the read. I'm all for things that make this game more realistic. It is obvious that the pancake totals far outweigh what could happen in reality and while that is not the problem, it is indicative of a problem. This solution appears very well thought out and I'd love to see it implemented. After that, perhaps we can do something to make the fumble situation more realistic as well.

tbh, I did not made it to the end of the suggestion.

Originally posted by Sly

2) That outside the "box area" all blocking uses pass blocking logic (ie players on the wings and downfield try to shield defenders from the player with the ball but do not drive block the defenders).

Pass blocking is "chicken fight" where the blocker is "letting" the defener lead as long as he doesn't eneters the pocket.
Run block is either drive away, or seal block
Open field block are much similar to standard run block in the goal of the block. However since it is much harder to get a good position on the defender (due to speed, open space and angles), most of the time the lbocker with just try to land a big hit on the defender, getting him away from the ball carrier just enough.

Originally posted by Sly

4) remove the current run focus / pass focus choice for blockers and replace it with a strength focus / agility focus. A strength focus would give a bonus to a blocker's effectiveness when he has a sizable strength advantage over his opponent, but a negative when he faces a strength disadvantage, while an agility focus would give a bonus when a blocker has an agility advantage over his opponent, but a negative when he faces an agility disadvantage. No focus would be neutral.

I always think about the run/pass focus as the OL stance, since it is more or less constant during the whole game. Your suggestion make less sense in RL. In GLB most OL (except the C) will have less strength and less agility from the DL they face. I am not sure this part of the suggestion make sense at all.


I agree that defenders angles (and the tackling mechanics) are really bad. In RL DB will force the HBs to cut toward the middle of the field, where the LB can take him down. You never see that in GLB.

Thanks for the comments, although it would be nice if you read the whole thing first.

Although you are focused on "real life", for this solution I needed to consider not only what happens in real life, but also what mechanics are offered by the game, and try to simulate a solution through existing mechanics offered by the game. You mention that both a drive block and a seal block would be considered forms of a run block in real life, but in the game all run blocks are drive blocks with a high chance of pancaking the defenders. I suggested pass blocking logic because among the existing AI options it best simulates the "getting him away from the ball carrier just enough" that you describe as what actually happens in most open-field blocks. So this is the AI option that best simulates the real life result that you and I agree is what effectively happens.

As for the situation where an O-lineman expects to face a defender who is both stronger and more agile than he is, the O-lineman in that case could choose "NO FOCUS" and would not receive any bonuses or negatives. In my experience, well-built O-linemen may occasionally face defenders who are both stronger and more agile, but often face more extreme builds, that might be, for instance, much more fast or agile, but less strong. I have seen such builds in DE's and DT's, and also remember that O-linemen don't only block D-linemen but also linebackers and defensive backs. If the game plan was to run outside a lot and an O-lineman expected to be pulling and lead blocking on most run plays, he would probably be facing players who were more agile but less strong.

I'm glad we agree that the tackling logic for "last men" on breakaway plays needs to be improved.

GG, Sly.

Seriously, though. I'm not sure your easy change would be quite so easy. Defining "the box" on a real football field isn't hard. Here, it would have to be calculated for every play and blocking logic would have to likely be totally rewritten to handle the two different types of blocking.

The problem with screens is the Outside Blocker VA. A 30% boost to Speed, Agility, Vision, and Blocking turns an offensive lineman with 68 speed and agility to an offensive lineman with 88 speed and agility (with similar vision and likely insane blocking). Add in Downfield Blocker and poor pathing and you have the screen problems of today.

Originally posted by AirMcMVP
GG, Sly.

Seriously, though. I'm not sure your easy change would be quite so easy. Defining "the box" on a real football field isn't hard. Here, it would have to be calculated for every play and blocking logic would have to likely be totally rewritten to handle the two different types of blocking.

The problem with screens is the Outside Blocker VA. A 30% boost to Speed, Agility, Vision, and Blocking turns an offensive lineman with 68 speed and agility to an offensive lineman with 88 speed and agility (with similar vision and likely insane blocking). Add in Downfield Blocker and poor pathing and you have the screen problems of today.

And the screen play has, I think, shed light on just how bad open field blocking and defensive pathing work.

Originally posted by Sly
Thanks for the comments, although it would be nice if you read the whole thing first.

But that would make it the longest thing I read since high school , lol.

Originally posted by Sly


Although you are focused on "real life", for this solution I needed to consider not only what happens in real life, but also what mechanics are offered by the game, and try to simulate a solution through existing mechanics offered by the game. You mention that both a drive block and a seal block would be considered forms of a run block in real life, but in the game all run blocks are drive blocks with a high chance of pancaking the defenders. I suggested pass blocking logic because among the existing AI options it best simulates the "getting him away from the ball carrier just enough" that you describe as what actually happens in most open-field blocks. So this is the AI option that best simulates the real life result that you and I agree is what effectively happens.

I think the game mechanics has all the needed feature. It is more a problem with the pathing the players takes and their assignments. Except for a "big hit" in outside blocks (more when pulling and less in screens).
The pass block will not do the trick, first since the mechanics is wrong (at least the current one is a hit and release technique). Although the SAs and VAs applied to it will be weird. Like a 3rd down bonus to blocking on screen passes ?

If anything close to the Box idea and the "reverse angle" aspect could be programmed it would be great to at least see it in test. I don't see much problem in establishing the box. It would be relative to the ball position at the beginning of the play and we always know where that is. The direction the defender is taking is also always known. It seems to me that these changes could be tried rather easily.

It would be nice if someone who has more programming knowledge that I do could opine about the difficulties (or lack of difficulties) in defining a "box area". I know that on every play, all players and the ball are tracked in each segment by a longitude/latitude position. It therefore seems to me that it would be a relatively easy task to define a longitude/latitude area in the first segment from the starting position of the ball, and check each player who attempts to block to determine if that player is in the "box area" or not.

I don't agree that "blocking logic would have to be totally rewritten" since the whole point of my suggestion was to avoid a complete rewrite by using logic which has already been written. By using something similar to the existing pass blocking logic in open field situations, I am hoping to avoid a complete rewrite and instead use some easy to copy existing logic. I do agree that there will need to be SOME programming work, but it is my hope that it would be far less than other options.

It might make it easier if on passing plays, once a pass is caught, to have the box area disappear completely and ALL blocking could be pass blocking. I think either way is ok, whichever is easier to program.

As tragula states, "it is more a problem with the pathing the players takes and their assignments." I don't think anyone disagrees. But the difficulty of programming different path and assignment logic led me to the practical solution of using something similar to the existing pass blocking logic in open field blocking situations. While tradula's point is well taken that this blocking is not EXACTLY what would happen in real life, I think that in the game it would EFFECTIVELY do the same job with the minimum amount of programming. I therefore think pass blocking WILL "do the trick" in terms of offering a practical solution that can be implemented rapidly and with far less programming effort than other options. Of course, I again would appreciate someone with more of a programming background than I have to speak to this question.

I read the entire post. This brings things to mind, and I know some might be points you have made, but I'm just going to rattle off what's in my head.


1. Defenses don't use angles to their advantage, nor do they attack blocks.
-Often times, you'll see a LB come up and take out the FB to clear the way for other defenders to come in and hit the RB. I don't see this in GLB. Yes, players will get hit by the FB, but they don't seem to take him out of the play. I'm talking full force "try to knock him on his ass or take him down" intentions. Could be going for the legs, coming in full force, etc., but I'd like to see some way of making that happen. Could even have a VA to accompany it.

-The pursuit angles are solid, but there doesn't seem to be any "break downs" from the defensive players. You know, the ones where a defender gets into the open field against a ball carrier, stops, squares his shoulders and waits for the ball carrier to make a move before he attempts to bring him down. It's usually very quick, but this allows the other defenders time to swarm to the ball. This may happen in the game, but off the top of my head I can't picture a time where I've seen this happening.

-Defenders don't "force" players inside. On a pitch play, you'll often see a CB shed his block, and show a little bit up the field in an attempt to force the RB to cut inside. This is basic defensive tactic, and is used to turn the RB back in for the rest of the defense. Of course, the CB may be blocked, or the HB might not cut back as far as the CB wanted, but they should at least try. This might be a tactic, an aggression slider. Breaking down and "showing" is more of a passive maneuver, where as going for the tackle is aggressive.*

2. Ball carriers don't set up blocks

- The best running backs in the games are ones who use their vision to set up their blockers. They may start to run outside of the WR, only to cut it back inside. This causes the CB to pursue outside, and when he goes to cut back, the WR is there to meet him. This can happen just about anywhere on the field, and if this logic was introduced, I think we'd see a lot more cut backs. Of course, your ability to set up blocks would depend highly on your vision

- Sometimes, a ball carrier will slow up a bit to allow the blocker time to get in position. In GLB, it seems that ball carriers just run full speed from the instant they get the ball. Of course they slow down when they make cuts, but that's just a side effect of the cut and not the intention of the ball carrier.

3. Defenders should generally win the 'open field' battle.

-It's tough to block in the open field. With the defender able to easily go backwards, forwards, left or right, a blocker at a slight disadvantage. On top of that, the defender can see the ball carrier, while the blocker will generally have his back to him. Makes it tough, and the blocker will constantly have to respond to the defender's movements

4. Open field tackles are tougher

- An open field tackle is generally considered tougher than a tackle in traffic. The ball carrier has more space to evade the defender and is tougher to lay a solid hit.

- It's tougher to force fumbles due to the reason above. If you cannot get a solid hit on the ball carrier, the chances of you popping the ball loose are going to be greatly reduced. Making a tackle from behind, though, changes the picture. The ball carrier is often holding the ball without too much security, and a square punch to the ball or a solid rip can pull that ball out of there. High tackling skills would be needed

-Defenders are more likely to be run over, juked, faked, etc**



* - This gave me an idea. Open field tackling tactics. "Break Down vs Go for Tackle" There would need to be a way to prevent more than one player from simply breaking down. If the ball player is "slowed" or "stopped", then SOMEBODY has to attack.

** - If the tactic is introduced, Break Down reduces the chance of being juked and faked out, but increased the chances of being run over. Go for tackle does the opposite.

Great thoughts! You've put some time into your post and it shows. You did a good job describing many elements of the AI logic and where they break down. I guess one outstanding question is, given that it might be a huge programming effort to fix all these issues, do you think that my suggestions of pass blocking logic outside the box area, etc. might go a long way to approximating what all the more complex logic you suggested does? Given limited time and ability for more programming I think my original suggestion will be a nice 85 percent solution.

I don't know if this has been said in this thread, but you are missing one very important point in downfield blocking... that is the ball carrier's ability to set up blocks for his blockers. Players are taught to use their blockers at a young age. Further, when you get to college football level, you learn how to help your blockers by setting them up. A faster, more agile player can get around a big ugly... but if the ball carrier moves to keep the blocker between the CB and himself... then he is likely going to either a) be pancaked, or b) be out of position to make a play as he tries to get around the blocker.

While this idea is nice, you would just be turning this around the opposite way... so instead of blockers blocking effectively, there would be no reason to run a screen play, or sweep for that matter, at all.

Overall, I think the realism is solid in this situation as it is. I do know that my CBs in previous seasons made a ton of tackles for short gains on screen plays... so I'm sure a player can find a way to the runner already.

Here is just one example... the game this play is from was a 2nd round playoff game against a VERY good team. They threw a lot of screens, as my defense is blitz heavy. If you go through the game, you'll see my players make a LOT of plays on the screens. Again, this leads me to believe things are fine in this area, and it is likely more of a build and defensive play call issue for you rather than game mechanics.

http://goallineblitz.com/game/replay.pl?game_id=690575&pbp_id=17769284

Originally posted by cherrie
I don't know if this has been said in this thread, but you are missing one very important point in downfield blocking... that is the ball carrier's ability to set up blocks for his blockers. Players are taught to use their blockers at a young age. Further, when you get to college football level, you learn how to help your blockers by setting them up. A faster, more agile player can get around a big ugly... but if the ball carrier moves to keep the blocker between the CB and himself... then he is likely going to either a) be pancaked, or b) be out of position to make a play as he tries to get around the blocker.

Ha, two posts above yours, #2

Originally posted by Sly
Great thoughts! You've put some time into your post and it shows. You did a good job describing many elements of the AI logic and where they break down. I guess one outstanding question is, given that it might be a huge programming effort to fix all these issues, do you think that my suggestions of pass blocking logic outside the box area, etc. might go a long way to approximating what all the more complex logic you suggested does? Given limited time and ability for more programming I think my original suggestion will be a nice 85 percent solution.

I definitely like the idea of "pass" style blocking. Granted, once a blocker gets his hands on the defender, he might try to drive him, but they generally seem to just try and get between the runner and the defender and act as a "shield" of sorts. Pass blocking is similar, so I think a slight tweak to that sort of logic would work effectively. It would also seem to be a great step in the right direction, and would most likely make other updates to open field defender/runner/blocker interactions a little easier to add.