User Pass
Home Sign Up Contact Log In
Forum > Discuss GLB Issues With Catch22 > NOT A BUG - CLOSED > wierd pass coverage bugs where the defender starts running at the guy while he does a streak
Page:
 
Archeon
offline
Link
 
no doubt this is the best example:

http://goallineblitz.com/game/replay.pl?game_id=311088&pbp_id=6892492
 
Link
 
yeah it happens a lot and bort needs to explain why this just started happening this season
 
Black Peter
offline
Link
 
Pass defense is starting to become a joke. Maybe his attempted fixes to zone have just fricked up everything else. Not much a DC can do to adjust to this BS either. QBs and receivers have become supermen, while DBs and LBs have become blind beggars holding out tin cups. Just becoming absurd.
 
ksmilkman
offline
Link
 
Originally posted by Archeon
no doubt this is the best example:

http://goallineblitz.com/game/replay.pl?game_id=311088&pbp_id=6892492


Well you've got the over-run there, but you've also got the WR adjusting to the underthrown ball, then the athletic YACs to the end zone. Could we please require WRs and HBs to have some reasonable amt of vision to pull of this crap? I mean I have a speed/agility WR w/ vision in the mid-30s and he's only dropped one ball all season. IMO, effective receivers that can catch balls in traffic, come back to snag underthrown balls, and beat DBs after the catch should need vision comparable to quality DBs (i.e. 60+).
 
bacardi
offline
Link
 
Originally posted by Darkstrand
It's really getting frustrated watching defenders time in and time again run at the player they are covering while the player does a streak. This doesn't just happen to linebackers

At first I thought it was just LBs.. but you can see the CBs doing the same things. It's as if bort programmed in a bump-and-run tactic, and then removed the physical contact from the bump.

http://goallineblitz.com/game/replay.pl?game_id=304883&pbp_id=4355081

Then I thought maybe it was more QB pump fakes... but it's become so prevalent that unless every QB in the game put all their boost points into it, while every CB and LB decided to forgo vision this season, it makes no sense.

Originally posted by Archeon
no doubt this is the best example:

http://goallineblitz.com/game/replay.pl?game_id=311088&pbp_id=6892492

I think that's a great example... watching all 5 CBs do the same thing. I've seen it happen a few times against my team as well, and it gets real annoying watching this dance against the 5WR set.

Originally posted by Darkstrand

heres another example of borts brilliant coverage programming that's probably unrelated but included for your own amusement:
http://goallineblitz.com/game/replay.pl?game_id=294132&pbp_id=6927576
lol that example is just plain funny :]
 
bgharst
offline
Link
 
Yeah DS just pointed this out to me. And I agree this is pretty absurd. Game planning goes out the window when you have bugs like these. But what I find funny on that 5WR replay is the QB decided to pass it to the one WR that beat his guy the least and got doubled up by the FS.
 
Bort
Admin
offline
Link
 
It does have to do with trying to close the gap when on "close" type coverage, and when he does, he doesn't have the agility to turn around a full 180 right away (nobody really does). I'm not sure what the appropriate fix is, though, as far as the logic goes. If I make him not move forward and instead always wait for the man to reach him, he will get burned all day by hooks or curls. This isn't exactly a new thing - I've been wrestling with it for 7 seasons. Anyone have any ideas?
Last edited Jan 17, 2009 15:51:43
 
InRomoWeTrust
Lead Mod
offline
Link
 
Originally posted by Bort
It does have to do with trying to close the gap when on "close" type coverage, and when he does, he doesn't have the agility to turn around a full 180 right away (nobody really does). I'm not sure what the appropriate fix is, though, as far as the logic goes. If I make him not move forward and instead always wait for the man to reach him, he will get burned all day by hooks or curls. This isn't exactly a new thing - I've been wrestling with it for 7 seasons. Anyone have any ideas?


Perhaps we get a new tactic? Somehow to let us choose between jumping the hook/curl or something that allows to favor more of the streak types.
Last edited Jan 17, 2009 16:03:00
 
Darkstrand
offline
Link
 
Originally posted by Bort
It does have to do with trying to close the gap when on "close" type coverage, and when he does, he doesn't have the agility to turn around a full 180 right away (nobody really does). I'm not sure what the appropriate fix is, though, as far as the logic goes. If I make him not move forward and instead always wait for the man to reach him, he will get burned all day by hooks or curls. This isn't exactly a new thing - I've been wrestling with it for 7 seasons. Anyone have any ideas?


People say it occurs on all coverage types but yes I figured it was basically the WR getting outside the coverage radius and the defender attempting to close.

I think there are a lot of ways. The first way I thought of was the most straight forward but I"m not sure how much it will increase simulation time.

You must have some functions governing movement. So, you can easily create/use a function for instantaneous velocity. You can take a linear approximation of the velocity for both players and approximate if the receiver will be inside the defenders FUTURE coverage radius in the next few ticks or whatever unit you use.

If not you don't have to apply the same level of negative acceleration you are currently applying where the defender starts running the other direction but just a slight one so that the WR catches up. If it's a hook route the linear approximation of the velocity is going to change rapidly so the defenders acceleration should decrease by enough to get him running in the other direction.

basically by looking at velocity instead of just the position it will help with hook routes. JUST LIKE A REAL PLAYER ZOMG REALISM ZOMG

Last edited Jan 17, 2009 16:08:45
 
bacardi
offline
Link
 
Originally posted by Bort
It does have to do with trying to close the gap when on "close" type coverage, and when he does, he doesn't have the agility to turn around a full 180 right away (nobody really does). I'm not sure what the appropriate fix is, though, as far as the logic goes. If I make him not move forward and instead always wait for the man to reach him, he will get burned all day by hooks or curls. This isn't exactly a new thing - I've been wrestling with it for 7 seasons. Anyone have any ideas?

i can understand wanting to close the gap, but a CB would never move "down-field" while a WR is moving up-field unless the ball is in the air. the only time a CB should be heading down-field is if he sees a WR cutting sideways or back down (your hook/curl examples).

if there isnt a simple way to fix it with "do not move down-field if WR is heading up-field unless the ball is in the air," one suggestion might be to use those "shadow steps" you've implemented with the LBs... in parallel to how LBs are taking baby steps to see if its a run, CBs could take baby steps up-field (away from the WR) until they've committed on whether the WR will cut or not.
 
bacardi
offline
Link
 
Originally posted by Darkstrand
basically by looking at velocity instead of just the position it will help with hook routes. JUST LIKE A REAL PLAYER ZOMG REALISM ZOMG

yah that makes a lot of sense, both in terms of the problem and solution. if the root of the issue is that the CBs are only looking at WR *location* to determine whether or not to close, the obvious (but maybe difficult?) solution would be to take into account the WR speed and direction. similar to what i was saying, even if a CB is supposed to maintain *close* coverage, he would never charge at a WR who is also charging full speed at him.
 
Bort
Admin
offline
Link
 
Hm, I just got an idea to just compare the y-axes of the two players' velocity vectors and make the CB unable to move forward if they are both going the same way. Though it seems like I tried it before and it worked ok, but not in all cases. Off to test I guess.
 
Bort
Admin
offline
Link
 
Originally posted by bacardi
Originally posted by Darkstrand

basically by looking at velocity instead of just the position it will help with hook routes. JUST LIKE A REAL PLAYER ZOMG REALISM ZOMG

yah that makes a lot of sense, both in terms of the problem and solution. if the root of the issue is that the CBs are only looking at WR *location* to determine whether or not to close, the obvious (but maybe difficult?) solution would be to take into account the WR speed and direction. similar to what i was saying, even if a CB is supposed to maintain *close* coverage, he would never charge at a WR who is also charging full speed at him.


They look at the WR's location + their velocity vector * a calculated closing distance, using some vector algebra. It just doesn't always come up with the right location when the vectors are heading directly at each other.
 
dsl20
offline
Link
 
Originally posted by Bort
It does have to do with trying to close the gap when on "close" type coverage, and when he does, he doesn't have the agility to turn around a full 180 right away (nobody really does). I'm not sure what the appropriate fix is, though, as far as the logic goes. If I make him not move forward and instead always wait for the man to reach him, he will get burned all day by hooks or curls. This isn't exactly a new thing - I've been wrestling with it for 7 seasons. Anyone have any ideas?


Well, when CB's in real life are playing "Close" defense on a WR, they normally don't start the play 4 yards away from the LOS. They are helmet to helmet with the WR at the LOS, and cover from there. Maybe if the CB's could start the play closer to the LOS, then they wouldn't have to run up to the spot they should have started the play in.
 
tautology
offline
Link
 
It seems as though the current logic simply checks for relative position after a set period, then attempts to calculate a "shortest path to intercept" a calculated position which will represent the proper coverage position for the CB with respect to the WR.

A good CB does not attempt to match a position, but rather match the vector and position of the player he is covering.

Therefore, the formula should attempt to resolve achieving both a matched vector and a matched position with his coverage assignment. Not sure off the top of my head how to do that mathematically, but logically it seems feasible.


1) Calculate WRs current position and vector, calculate predicted position and vector after 1 tick, 2 ticks, and 3 ticks.

2) Change CB current vector to position the CB at a point with a vector that most closely matches the predicted WR position and vector after 1,2 or 3 ticks, and uses the three separate calculations as "limiters" to avoid the diving behavior.


In other words, the solution to most closely match position and vector after 1 tick will be used UNLESS that solution creates an untoward position and vector relationship at 3 ticks. The best path decision should be the decision that allows a highly acceptable position and vector after 3 ticks, while closing as quickly as possible within that solution after 1 tick.

Setting parameters for acceptable matches in vector and position would perhaps be delicate. Resolving the equations for potential different behaviors 3 ticks out might be daunting. You would also certainly need to account for acceleration differences...if the WR is slowing the CB needs to adjust towards the predicted future vector, as well as position (in case that wasn't obvious...it probably was )

Generally speaking, I would think that you would want a minimum 80% match in vector magnitude and 95% in direction. I am sure a radius could be be used for acceptable coverage distance, outside of which coverage is unacceptable.

Something like that....didn't really think that through clearly, but I think the ballpark is in there somewhere...
 
Page:
 


You are not logged in. Please log in if you want to post a reply.