User Pass
Home Sign Up Contact Log In
Forum > Game Changes Discussion > Announced Changes > Revision of Defensive Play Creator - Limiting Player Movement
Page:
 
Rage Kinard
offline
Link
 
Originally posted by neoliminal
The positions of the blitzers before the snap are obviously numeric, no? There's an X and a Y location for each blitzer.

Pseudo Code:

1. Count Blitzers
2. Count Blocker.
3. If Blitzers > Blockers and vision check was passed... audible TE/HB/FB in to block (as assigned in play)
4.a QB attempt Vision check to assign blockers.
b.b QB and Center assign Blocking (This is what you see before a play when the center and QB are pointing at defenders.)
5. Formulate paths for blitzing players.
6. Formulate blocking paths for blocking players.
7. Failed Vision checks have them blocking someone else.


You don't need 24 slots for this. In fact, that just seems to make it harder to work out....


How many x/y locations are there when you can move players fractions of yards vertically or horizontally?

You do understand that without specific slots defenders have to blitz from, that #4 creates exponentially more situations that have to be coded for, if defenders still have freedom of movement in the DPC.

For instance, what if DT is head up over LG and NT is head up over C. The MLB is blitzing through the A gap between C and LG. Do you want the same blocking code if MLB is on the line of scrimmage, .5 yards off the LOS, 1 yd off the LOS, 1.5 yds off the LOS, 2 yards off the LOS, etc. Where do you change the blocking assignments for the LG to pick up the MLB instead of the DT? And in reality, DCs can move that blitzer fractions of those distances. Then repeat that scenario for every combination of blitzers that DCs can throw at the SIM. Now what happens when you move the DT just slightly outside so he is only 1/2 way covering the LG. When does DT block MLB and when does he block the DT? What happens when you shift the DE so he is in the B gap between the LG and RT.

Without a set # of spots that you can blitz from, coding becomes too difficult. However, you need to give DCs enough choices so that creativity isn't killed.

The reason I went with 24 locations.

LOS is 9. It covers head up with C, inside shoulder and outside shoulder of Gs and Ts. Really you could push it to 11 or even 13. The reason for more in the back is that you need similar positioning between the tackles, and the ability to start further outside for OLB and for S, CB blitzes.

Really, I would prefer the number of locations be about 15 on the LOS and 21 off the LOS for a total of 36, but thought that would be too much. 36 is still minute compared to the countless #s of positions you can come up with on the x/y axis.
Edited by Rage Kinard on Oct 24, 2010 17:28:17
 
neoliminal
offline
Link
 
Think of this.

What if you just did holes and players?

http://upload.wikimedia.org/wikipedia/commons/1/16/American_football_Gaps_and_holes.svg

(Ignore the defensive naming for a second.)

There are 7 holes in a defensive line with a TE.

There, I've reduced it down to its lowest level...

Now path each blitzers and assign them to a hole. Let's say you have a 4-3 and the LOLB and ROLB are both blitzing.... one inside and one outside. You might end up with something like (and, using the numbers on that chart):

6-ROLB
6-RDE
2-RDT
1-LDT
3-LOLB
7-LDE

Note I've overloaded one of the holes (6).

QB looks over the defense and let's assume everything goes right and he figures out all the blitzers. He now assigns blockers, inside-out, for each blitzer.

LOT - 6-ROLB
LG - 6-RDE
C - 2-RDT
RG 1-LDT
ROT 3-LOLB
FB 7-LDE

Now you just have to figure out, for each of the assignments, where the point of contact should be. This is a fairly simple math problem regardless of what XY they are.

Where is the problem here?


Edited by neoliminal on Oct 24, 2010 17:36:09
Edited by neoliminal on Oct 24, 2010 17:35:30
 
Rage Kinard
offline
Link
 
Originally posted by neoliminal
Think of this.

What if you just did holes and players?

http://upload.wikimedia.org/wikipedia/commons/1/16/American_football_Gaps_and_holes.svg

(Ignore the defensive naming for a second.)

There are 7 holes in a defensive line with a TE.

There, I've reduced it down to its lowest level...

Now path each blitzers and assign them to a hole. Let's say you have a 4-3 and the LOLB and ROLB are both blitzing.... one inside and one outside. You might end up with something like (and, using the numbers on that chart):

6-ROLB
6-RDE
2-RDT
1-LDT
3-LOLB
7-LDE

Note I've overloaded one of the holes (6).

QB looks over the defense and let's assume everything goes right and he figures out all the blitzers. He now assigns blockers, inside-out, for each blitzer.

LOT - 6-ROLB
LG - 6-RDE
C - 2-RDT
RG 1-LDT
ROT 3-LOLB
FB 7-LDE

Now you just have to figure out, for each of the assignments, where the point of contact should be. This is a fairly simple math problem regardless of what XY they are.

Where is the problem here?




There are 7 holes, but there are also 6 heads up spots you can position players on. Remember you can put a player head up with the C, the LG, and the LT and blitz 2 defenders through the 2 inside gaps on that side of the field.

Are you going to assign the same blocking assignment if LBs are on the LOS as you are if the LBs are 2 to 3 yards off the LOS and coming through the same hole?

How do you account for the spots that are where the TE should be or outside the TE? What about S and CBs that are blitzing from wider points?

What if you have a NT head up over the C, a DE head up over the T and 2 LBs blitzing the 2 gaps on either side of the C, plus the FS blitzing from outside the T? Do you block the exact same way if those 2 LBs are on the LOS as you would if one or 2 of them moves a couple yards off the LOS? If not when do you change the assignment? When one of them is 1 yard off the LOS? When one of them is 4 yards off the LOS? What if 1 is 1 yard off the LOS and the other is 2 yards off the LOS?

Also, remember if the TE or HB blocks, that means another defender is going to be blitzing in addition to the ones you have already accounted for.
Edited by Rage Kinard on Oct 24, 2010 18:35:11
Edited by Rage Kinard on Oct 24, 2010 18:34:21
 
neoliminal
offline
Link
 
Give me a 4-3 blitzing situation from the I formation that you think is a problem to program and I'll walk you step by step through the steps that would lead to a person on person blocking scheme.
 
tragula
title
offline
Link
 
Originally posted by = Rage Kinard

There are 7 holes, but there are also 6 heads up spots you can position players on. Remember you can put a player head up with the C, the LG, and the LT and blitz 2 defenders through the 2 inside gaps on that side of the field.


More correctly, there are 3 basic alignments per O player on the LOS (head up, inside shoulder outside shoulder). With a TE that is 6*3 = 18. Now you populate that with 4 DL you have 3060 possible front 4. Then you need to add the possible LBs and DEs that are split wide.

No human player needs to learn how to handle this multiplicity by memorizing all possibilities. It is not too complicated to reduce the problem to ~4-6 basic alignments for each front (3-4, 4-3, 5-2, ...). The way GLB does it right now is flawed. I think it can be corrected, if they are willing to have a rethink and redo instead of just patching on the 2.5 years old code.


Overload blitzes should not be picked up.
- The O should have the tools to punish them.
- The O should have a way to choose blocking scheme that will nullify an overload blitz (opening a weakness somewhere else on the line).
- But when the D dials the right overload, and the O is surprised it needs to be a sack. It is bad news knowing that this will not be part of the future of the game.



 
simski
offline
Link
 
Originally posted by neoliminal
Give me a 4-3 blitzing situation from the I formation that you think is a problem to program and I'll walk you step by step through the steps that would lead to a person on person blocking scheme.


As mentioned earlier there is an infinite (Actually bounded by the number floats in an interval thats stille a huge number) number of alignments of players in the current DPC. This means that when this is being coded you have to consider a lot of different intervals. As Rage Kinard pointed out this means that moving one player very little can send you into a different case (Even though everybody else is not moved), in each different case you have a lot of replicated code which is more then likely present other places (The example of moving the DT with A-gap thread). When you fix the number of spots blitzers can be placed in this problem becomes easier since you can basically take each case at a time and handle it. However even with just 24 blitz spots and a limited amount of blitzers there are still a lot of possible combination (Some Binomial coefficients are needed to calculate that).

I think we have to keep in mind that accounting for blitzing is not an easy task since thats what good QBs do. Rage Kinard suggestion is really good since it is possible to implement and still gives the DC something to work with. 24 potential positions could be a bit much, but since some are limited to d line personal some to LBs and so on it could be possible. I think the blocking assignment problem is similar to the quadratic assignment problems (Which is NP-hard) in the sense that you have to assign blockers to blitzers while minimizing the distance a blocker has to move. The blocking assignment problem has an additional twist, since there are probabilities involved in the decision process (Who do you expect blitz).
Edited by simski on Oct 24, 2010 23:53:04
 
tragula
title
offline
Link
 
Originally posted by simski
...However even with just 24 blitz spots and a limited amount of blitzers there are still a lot of possible combination (Some Binomial coefficients are needed to calculate that).


24 slots and 7 players (order is not important) = 346,104 different combinations.
 
Rage Kinard
offline
Link
 
Originally posted by neoliminal
Give me a 4-3 blitzing situation from the I formation that you think is a problem to program and I'll walk you step by step through the steps that would lead to a person on person blocking scheme.


It's not that any 1 situation is hard to code, it's that there are tens (maybe hundreds) of millions of different possibilities that have to be coded for when you don't have set spots to blitz from.

A human can recognize slight differences in positioning, assess whether or not they are relevant, and account for it on the fly. They can make adjustments in game after seeing a new blitz one or two times.

A computer game like this has to account for every possible scenario ahead of time. A very slight adjustment, still needs to be pre-written into the code so players block according to the specific alignment of the defense and you don't risk having a bug in the blocking.

It would take years of just plugging in different blocking codes, and by the time it was done it would take days to run an individual sim because the script would be huge with the number of variables that had to be accounted for with each blocking assignment.
 
foofighter24
jumpin da snark
offline
Link
 
I keep asking this to no response: Where are these "exploit" blitzes that have no counter to them? Am I missing the team that is destroying equal competition using these?
 
foofighter24
jumpin da snark
offline
Link
 
I know the blitz where the MLB occupies the center and the RDE is split out wide, with a ROLB blitz either on the edge or in the LOT/LG gap, but that has counters.
 
The Avenger
Hulk Smash
offline
Link
 
Originally posted by foofighter24
I know the blitz where the MLB occupies the center and the RDE is split out wide, with a ROLB blitz either on the edge or in the LOT/LG gap, but that has counters.


Not against rushing plays! And alot of LOL TE blocking on the left side does nothing nor does holding a HB or FB back because the sack is in 1.5 tics too!
 
foofighter24
jumpin da snark
offline
Link
 
If we are using WL as the standard, you have 25% of the league passing for 6,000 yards. It also appears many team are averaging well over 5 YPC. Granted, the INTs are abit high, but you are talking teams with like 700-800 attempts.

How much offense do you want? 120-100 games?
 
Rage Kinard
offline
Link
 
Originally posted by tragula
24 slots and 7 players (order is not important) = 346,104 different combinations.


yes, but some of those combinations would not have to be accounted for since they would be impossible due to restriction of player movement.

For instance, you have to use at least 2 numbers from the 1-9 when using a 4 man DL and 1 when using a 3 man. I also believe that with current restrictions, you can only have a max of 6 blitzers on one side of the center. So options like 7,8,9, 21,22,23, 24.

You can also add further limitations that cuts down on the number of possibilities that doesn't cut down on reasonable creativity. For instance, you could say that at least 1 pass rusher has to be on each side of the C, and no more than 5 players can be pass rushing on either side of the C. So you could still get 6 pass rushers from C over, but it cuts down the amount of combinations that have to be coded.
 
jdros13
offline
Link
 
if on the offensive packages they changed the drop down boxes (and the coding obviously) for the HB and FB positions to allow you to determine "pass block left" and "pass block right" and made a very difficult vision check to override that assignment it would make it much easier for offenses to pick up blitzes by keeping a back in to block (although the backs would probably get run over and miss blocks a fair amount as well of course).

It would also open up the strong side for blitzing as well, which would be nice.
 
AlBarsch
online
Link
 
Originally posted by foofighter24
I keep asking this to no response: Where are these "exploit" blitzes that have no counter to them? Am I missing the team that is destroying equal competition using these?


There are none... Only the lower level OCs with limited experience are having problems trying to figure out which plays to counter an overload blitz.
 
Page:
 


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