4/24/2013 7:20 PM
Posted by slid64er on 4/24/2013 5:50:00 PM (view original):
Posted by slid64er on 4/24/2013 5:49:00 PM (view original):
1-10   OFF: Shotgun: Default    DEF: 4-3: 4-3
100 QB Rick Smith
100 OL1 Floyd Best
100 OL2 Roy McCoy
100 OL3 Charles Ramirez
100 OL4 Thomas Nelson
100 OL5 John Ellis
100 TE John Smith
100 WR1 Kevin Rosa
100 WR2 Douglas Curtis
100 WR3 James Delgado
100 WR4 Christopher Goodman
100 DL1 Mike Hamilton
100 DL2 John Sanders
100 DL3 Jose Gregory
100 DL4 Charles Warren
100 LB1 Marcus Armstrong
100 LB2 Steven Moore
100 LB3 John Thomas
100 CB1 Edward Rush
100 CB2 Jeffrey Bell
100 SS Robert Torres
100 FS Raymond Thomas
SHOTGUN - Rick Smith takes the snap. [Pass Defense - BLITZ: Moore - long cover]
The defense is starting to get pressure on Smith. Smith throws to the wide open Goodman at the L 14 (deep). Pass is off the mark. Incomplete.


Missing wide open receivers seems pretty common.  Should we ratchet the completion percentage up when they are wide open and turn it down when they are covered/well covered?


Can you please address this?
It does actually adjust the chance of completing a pass (throwing on target) based on how open the player is.  In this case, the QB just missed the target which should be acceptable for a deep pass.  Not saying it can't be tweaked, but the adjustment is in there.  You also have to look at the quality of your QB, and pressure on the QB can affect his chance to throw on target as well.
4/24/2013 8:30 PM
Posted by norbert on 4/24/2013 7:10:00 PM (view original):
When short on the WR/CB match up and LBs have no one to cover, LBs will try to cover WRs.  I may have to adjust their effectiveness but essentially it is looking at their ratings on how well they actually cover the players.  Theoretically, with the mix of ratings being different for covering further down the field versus short, the LBs should have a natural disadvantage in covering deep.  However, if they are covering a WR short, then the comparison would essentially be the same as covering a RB/TE.  Basically, we don't penalize an LB that actually has the skills to cover a WR just because he is covering a WR.  Don't forget to add rating comparison in to the check for whether a player is open or not, because even a 1-on-1 match up can result in the receiver being covered.

There really seem to be two concerns here: what players are covering different areas and how does that affect the open result AND players participating in tackle attempts in locations they probably shouldn't be able to get to from their cover locations. Luckily they can be treated pretty independently, and I'll look into this closer over next week.

I think the S moving up and participating in a tackle can be addressed in some minor adjustments, but it requires a lot of testing to make sure it doesn't screw up other parts of the play.  It's really just shifting the influence set by the cover for Safeties and then adjusting the logic that determines which players can participate in the play at each location.  I would try to limit it to move one location per step, which would mean a S in long cover could meet a receiver in the medium area if he caught the ball in the short area, but would not be able to make a play in that short area.

The second part is to review the logic that determines which players are in cover (and their influence).  I think the cover check is pretty decent, but maybe I need to present more detail on what each type of cover actually means and what that cover result means for the rest of the play.  Probably some room for improvements there as well once I get more feedback.

I think these are good and focused examples of what will help make this a better engine, and I will always believe there is room for improvement.  I think working on some of this might help further define how much game settings will matter and i think that's a good thing.

Here are the ratings for the players involved in the 4-3 v shotgun example:


Pos Starter A SPD D WE ST STR BLK TKL H GI E T TOT
WR   20 56 51 62 57 38 23 17 54 33 51 47 509
WR   48 64 67 75 48 27 26 1 44 45 61 31 537
WR   36 52 33 27 52 31 58 16 39 25 61 36 466
WR   44 64 48 77 34 44 32 30 66 41 63 41 584
                             
                             
Pos Starter A SPD D WE ST STR BLK TKL H GI E T TOT
LB   45 39 45 48 58 56 11 62 32 41 40 41 518
LB   43 37 47 30 54 60 18 61 22 29 38 46 485
LB   53 39 37 51 37 52 27 58 29 44 60 57 544

I'm fine with not penalizing LBs in coverage, but something is wrong when LBs can consistently cover WRs short or deep.  The speed difference alone should render them ineffective in trying to track a receiver.

This is also the first I've heard of different coverage ratings for different areas of the field.  Why?  Speed is speed.  Elusiveness is elusiveness.  Hands are hands.  Theoretically, the faster receivers go downfield and are covered by faster defenders (DBs) while slower receivers (TEs and some RBs) don't go downfield and are covered by slower defenders (LBs).  This may be the reason the LBs are so good in coverage against WRs, especially if the speed impact is lessened as passes get shorter.  Even on short passes, I would give a decisive advantage to a speedy WR on an average speed LB.  No way the LB keeps up with the WR on a bubble screen, slant or stick.  Especially if there is no safety help short.

I think you're on the right track with adjusting the logic about where players can participate in a play.  I really think it's wrong for someone to be able to influence the entire field from sideline to sideline and from long coverage to tackles in the short area.  I also think you need to look at the cover influence again.  The example above with a 4-3 covering a shotgun consistently well is an issue for anyone who looks at the game logically.
4/24/2013 8:34 PM
Posted by norbert on 4/24/2013 7:17:00 PM (view original):
Posted by slid64er on 4/24/2013 5:53:00 PM (view original):
What happens when a ball is "in and out of the hands" of a receiver?  This does not show up as a drop.  Was the pass bad?  If the pass was not bad, why wasn't it caught or ruled a drop?
This is a dropped pass.  Is it not showing a drop for the player?

There are basically 4 results of a throw:
  • off target - receiver has no chance to catch it
  • reach - receiver has to reach to catch the ball, thrown high, behind, etc.  Chance to catch based on receiver ratings.  If doesn't catch, is not a drop.
  • slightly off - fairly easy catch. Still chance to drop based on ratings. Will count as a drop.
  • on target - easy catch.  Chance to drop based on ratings (low hands player may drop). Will count as a drop.
There are basically 3 ways a pass can be incomplete:
  • QB throws off target (or throws at limit of receivers reach)
  • Receiver drops the ball
  • Pass is defensed
All of these should be evident in the PBP.  It might make it easier to start adding more tags in the PBP like we did for "throw to ABC 20 (short)", so maybe add a (drop) in the PBP when the pass is actually a drop.
For whatever reason, it does not show up as a drop in the statistics.  That is the cause of my confusion.  Getting this to register as a drop should be an easy enough fix. 
4/24/2013 8:39 PM
Posted by norbert on 4/24/2013 7:20:00 PM (view original):
Posted by slid64er on 4/24/2013 5:50:00 PM (view original):
Posted by slid64er on 4/24/2013 5:49:00 PM (view original):
1-10   OFF: Shotgun: Default    DEF: 4-3: 4-3
100 QB Rick Smith
100 OL1 Floyd Best
100 OL2 Roy McCoy
100 OL3 Charles Ramirez
100 OL4 Thomas Nelson
100 OL5 John Ellis
100 TE John Smith
100 WR1 Kevin Rosa
100 WR2 Douglas Curtis
100 WR3 James Delgado
100 WR4 Christopher Goodman
100 DL1 Mike Hamilton
100 DL2 John Sanders
100 DL3 Jose Gregory
100 DL4 Charles Warren
100 LB1 Marcus Armstrong
100 LB2 Steven Moore
100 LB3 John Thomas
100 CB1 Edward Rush
100 CB2 Jeffrey Bell
100 SS Robert Torres
100 FS Raymond Thomas
SHOTGUN - Rick Smith takes the snap. [Pass Defense - BLITZ: Moore - long cover]
The defense is starting to get pressure on Smith. Smith throws to the wide open Goodman at the L 14 (deep). Pass is off the mark. Incomplete.


Missing wide open receivers seems pretty common.  Should we ratchet the completion percentage up when they are wide open and turn it down when they are covered/well covered?


Can you please address this?
It does actually adjust the chance of completing a pass (throwing on target) based on how open the player is.  In this case, the QB just missed the target which should be acceptable for a deep pass.  Not saying it can't be tweaked, but the adjustment is in there.  You also have to look at the quality of your QB, and pressure on the QB can affect his chance to throw on target as well.
I think this needs to be adjusted some more.  I've seen several plays where there is no pressure and there is a wide open receiver.  These need to be completed at a higher rate.  If there's pressure, that's fine and i can understand missing with regularity.  But if the QB is sitting in the pocket like a 7 on 7 drill, he completes a high percentage of his passes to uncovered receivers regardless of distance.  
4/24/2013 9:28 PM

After reading the conversation I have the following thoughts:

Momentum of the play - If the S is covering deep he would be running downfield, it would be hard to stop, turn run upfield to tackle a WR who caught the ball short

LB in coverage - except for RB and TE, LB who defend WR usually don't run with them but use superior GI to position themselves to make the play (short)

Inside vs outside - passing should have same designation as running - inside covered by ILB and S, outside by CB and OLB for short and medium passes. S cover both for long and deep passes. This would get more specific match-ups.

For short and medium receiver/TE patterns higher comp% would be for the "possession" receiver who would catch the ball stopped, lower comp%, but higher YAC for the speed receiver who catches the ball running.

4/24/2013 10:48 PM
Slider, can you please share this with Norbert via the forums? I cannot because I don't have an active paid team.


Hi Norbert,

I'm a beta tester for GD 3.0, but since I have no active paid for teams I cannot send this to you via the forums.

I just read the recent posts about coverage and there is a pretty big loophole in the logic.

The issue is that I can setup my formation so that a star WR is the only one that has "influence" in an area and then have all the other receivers drag the defenders to other areas. As far as I can tell, there is no way to counteract this with defensive settings because your approach to determining where the defenders cover is basically always like a zone defense.

I did this setup and although my QB and WR are not that great, I just throw deep to the one WR every play for some test games. In a test game, I just scored 160 points against the #1 team in this way while throwing for 1628 yards (1614 to one WR).

My team is Texas A&M Kingsville in Craven if you want to see how its setup.

While using this loophole might be fun for awhile, I think it would ruin the game.
4/25/2013 2:47 AM
i noticed another guy do it too. he would throw for 300-400 yards a game but only have a comp% of like 20. i set up a DEF that was all Dime 100% pass and Long. In test games he would still pass for 300 yards and 3 TD but id beat him with a running OFF.
4/25/2013 9:16 AM
the Javelinas were always a handful for Portland State back in the day

they had a running back Johnny Somebody used to kick our ***
4/25/2013 10:48 AM
Posted by slid64er on 4/24/2013 10:48:00 PM (view original):
Slider, can you please share this with Norbert via the forums? I cannot because I don't have an active paid team.


Hi Norbert,

I'm a beta tester for GD 3.0, but since I have no active paid for teams I cannot send this to you via the forums.

I just read the recent posts about coverage and there is a pretty big loophole in the logic.

The issue is that I can setup my formation so that a star WR is the only one that has "influence" in an area and then have all the other receivers drag the defenders to other areas. As far as I can tell, there is no way to counteract this with defensive settings because your approach to determining where the defenders cover is basically always like a zone defense.

I did this setup and although my QB and WR are not that great, I just throw deep to the one WR every play for some test games. In a test game, I just scored 160 points against the #1 team in this way while throwing for 1628 yards (1614 to one WR).

My team is Texas A&M Kingsville in Craven if you want to see how its setup.

While using this loophole might be fun for awhile, I think it would ruin the game.
If this is a problem, it is more of a problem with the completion chances deep and these other cover issues we've been discussing.

The defense is not actually a zone, you will have a CB (or technically 1 CB worth of influence) on that guy deep as well as Safeties.  So, yes, you can pull defenders to certain areas, but shouldn't you be able to?  This guy will not be defenderless and you won't pull everyone off of him.  The CB coverage is man-to-man and the S coverage is zone.

It's difficult to hone in on the right chance to make a completion deep.  If a team throws deep every once in a while, you need enough completions that make it worthwhile, but if a team throws deep every play, you can't have too good of a completion chance or everyone would throw deep every play.  If we can figure a way to handle deep passes where the more selective deep pass is handled differently than the constant deep pass, then we should be able to handle this case where a team is passing deep a lot without penalizing the deep pass in general.

A lot of these changes may require a bigger change in the engine, so don't be discouraged if I can't work all of these scenarios out to perfection. I think I can get the current code to work better with some tweaks and adjustments and leave it open for a more intense update on coverage/passing later down the road.


4/25/2013 11:17 AM
As far as what ratings affect cover results and general receiving success, I think it's pretty obvious that SPD isn't the only factor.  With the logic broken out into steps and different checks, as opposed to 2.0 where it lumped everything into one big check, we have to start answering these questions.  So one part of this discussion I'd like to just focus on that.

Currently in 3.0, we can break out the cover match up by the target location, the target, the defender(s) in cover, and the step of the play.  We also take into account the pressure on the QB.  To select the defenders in cover, we look at all the defenders with influence in that location compared to the influence of the offense in that area, which tells us if there is more defensive cover in the area and the likelihood of a target being double covered.  We then pick the defender based on all those factors as well as hooking up CBs to WRs and LBs to RBs/TEs when possible.  Additional defenders will most likely be a S unless we have extra CBs or LBs.  Now we have our offensive player (the target) and the defenders and can look at the cover match up.

For the match up, we try to look at one rating (actually one mix of ratings) for each player based on the location, their role, and the current step.  Right now, the general idea is that targets in the medium, long, and deep areas in the first few steps of the play are mostly looking at SPD. Defenders ratings will depend on their role - CB/LB covering medium, long, and deep would also be looking mostly at SPD, while S which would already be in that zone would rely more on GI.  For the short zones, the offensive player is mixing in GI/ELU more than SPD and the defense is looking more at GI.  As the play continues on, at some point, those WRs running long are no longer outrunning the defenders and are now just established in the long zone and the ratings switch more to the GI/ELU than SPD.

This is up for discussion, but at some point I just have to decide something and go with it and it can't always be the perfect solution.  I can see the debate that SPD could affect a receiver's chances in the short zone, but I also think short passes and getting open can be a lot about just knowing where to go.  There are a lot of receivers we could point to that are examples of not very fast but get open on the short routes and slants.  It's very difficult to switch between the two philosophies for short routes and be able to also adequately give the defense their fair shake as well.  For game purposes, what do you guys think about making calls in these situations where we lock certain ratings in to be ideal for certain situations over others.  I'm even okay with locking in ratings for different positions for different situations as long as we can define those to where users know what to expect.

I'm going to add players in cover to PBP. Even if those defenders don't make sense or are wrong, it will let us know a little more about what is going on and if there is something off, it will be easier to see.

4/25/2013 11:33 AM
could there be some sort of AI learning load?
as in, if the O runs the a play, and then the same play again, and then again, on the 3rd of 4th consecutive play of exactly the same setup there would start to be decreasing probablilities of success. these odds would fall more and more with the 5th and 6th and 7th consecituve play with exactly the same settings, would be best to be non-linnear so that you can have the D mroe or less dialied-in to shut down an offense that literally repeats the same play over and over.
4/25/2013 12:07 PM
Posted by govolswoo on 4/25/2013 11:33:00 AM (view original):
could there be some sort of AI learning load?
as in, if the O runs the a play, and then the same play again, and then again, on the 3rd of 4th consecutive play of exactly the same setup there would start to be decreasing probablilities of success. these odds would fall more and more with the 5th and 6th and 7th consecituve play with exactly the same settings, would be best to be non-linnear so that you can have the D mroe or less dialied-in to shut down an offense that literally repeats the same play over and over.
Well, this gets in to a very debatable topic.  We could allow teams to adjust to repetitive patterns, but then you have to look at getting game planning for "free".  This also brings up the dilemma of real life versus game play.  In real life, if a team lined up in shotgun every play and threw deep every play, the defense would certainly adjust. In GD, if I set up a 5-2 and Short cover against someone that throws deep out of shotgun every play, what do we want to happen?  That's an extreme example, but it should demonstrate the problem.  In real life, obviously this defense would adjust.  In GD, should the game plan play out and just be mismatched or should it adjust to mirror real life.

So in the "always throw deep" scenario, if we keep the same success chances as every other deep throw, you can get too good of success from passing, while if we cut down the success for repetition, you get in to the area of getting success on the defensive side for doing nothing.  I'm sure we can come up with a compromise.

4/25/2013 12:38 PM
I'm all for KISS.  I think you set up one ratings mix for receivers and one ratings mix for defenders regardless of zone.  Speed, GI and Elu are all important regardless of distance from the line of scrimmage.  As it stands, since speed is minimized in the short zones, LBs can effectively cover WRs based off of GI.  It's one thing to know where you have to be or what a receiver is going to  do, it's another thing altogether to get there to defend the play, especially if there's no safety influence in the area.  You bring up a point about slower receivers getting open short, but they aren't being covered by LBs, and if they are that's why they're open.  And if that WR did have speed, would't the WR be even more effective?

If the defense is getting burned enough, maybe they should put the same amount of CBs on the field as there are WRs.  LBs are not meant to cover WRs so why neutralize WR advantages to make this an effective strategy?  We have different formation available to us on offense and defense, there should be advantages and disadvantages if there are personnel mismatches.  

And I'm not just talking about passing.  I've been torched by a trips running team using a 4-4 with all 4 LBs on the line.  You would think 8 in the box vs 5 blockers (possibly 6 if the TE was set to block) would stuff the run and open up the passing game, but neither happened.   


4/25/2013 2:13 PM
You still have to figure in the QBs ratings.  There's going to be a point where someone running down field and wide open is still going to result in the QB missing him.  I've even seen professional QBs miss a "wide open" receiver.

I also think there is some difference between a receiver that uses positioning and technique to get open and one that uses speed to get open.  The issue might be it's not an either/or check but rather whether or not a receiver can use his strength in getting open.  So maybe it's more like a receiver could use SPD or GI to get open and somehow we'd have to know which is being checked to use the defender's comparable ratings.  For instance, in 2.0, it lumps SPD, GI, ELU, TECH, and IQ all together so you can get a SPD receiver being covered well by a slow defender with GI, or vice versa, and that's not right either.

It would be more complicated and might be something I can't implement within this update, but I could look at doing something like I do for STR and SPD at the outside line where I compare the two different skillsets for the result.  So we could always look at the SPD matchup and the GI matchup and if WR wins either, they have the advantage.

4/25/2013 2:52 PM
Posted by norbert on 4/25/2013 2:13:00 PM (view original):
You still have to figure in the QBs ratings.  There's going to be a point where someone running down field and wide open is still going to result in the QB missing him.  I've even seen professional QBs miss a "wide open" receiver.

I also think there is some difference between a receiver that uses positioning and technique to get open and one that uses speed to get open.  The issue might be it's not an either/or check but rather whether or not a receiver can use his strength in getting open.  So maybe it's more like a receiver could use SPD or GI to get open and somehow we'd have to know which is being checked to use the defender's comparable ratings.  For instance, in 2.0, it lumps SPD, GI, ELU, TECH, and IQ all together so you can get a SPD receiver being covered well by a slow defender with GI, or vice versa, and that's not right either.

It would be more complicated and might be something I can't implement within this update, but I could look at doing something like I do for STR and SPD at the outside line where I compare the two different skillsets for the result.  So we could always look at the SPD matchup and the GI matchup and if WR wins either, they have the advantage.

I've seen misses as well.  And if there's any kind of pressure, I think it's fine as is.  But when the receiver is wide open due to a broken coverage or huge mismatch and there is no pressure, the QB more often than not hits that receiver.  I don't expect them to hit the receiver all the time, but under those conditions I expect a decent throw.  I just feel there should be a bigger bump for a receiver being wide open and a harsher penalty for a receiver being covered.

I think you're overcomplicating this.  Why can't you use SPD, GI, ELU, TECH and IQ all together?  All you have to do is weight the important attributes higher and find a balance.  It's not like a receiver is open only because he's fast or only because he uses positioning and technique.  Superior receivers would use everything to their advantage.  By limiting what a receiver can do based on field depth, you're taking away other advantages the receiver may have.  Along the same lines, if the defender has a decisive GI advantage, he should be able to cover more ground.  But if you only look at speed, you take away the GI advantage and what may make the defender superior overall.  Finding the balance is the key and I really don't think it should be an either/or scenario.  Mike Singletary may have had a high GI, but in no way would he cover Desean Jackson consistently.  Usain Bolt has great speed, but his GI would make him a bad cover corner.    
of 5

Terms of Use Customer Support Privacy Statement

Popular on WhatIfSports site: Baseball Simulation | College Basketball Game | College Football Game | Online Baseball Game | Hockey Simulation | NFL Picks | College Football Picks | Sports Games

© 1999-2014 WhatIfSports.com, Inc. All rights reserved. WhatIfSports is a trademark of WhatIfSports.com, Inc. SimLeague, SimMatchup and iSimNow are trademarks or registered trademarks of Electronic Arts, Inc. Used under license. The names of actual companies and products mentioned herein may be the trademarks of their respective owners.