When you do the PBP Bebug it comes up with this:

[INBlk:51-53 (0.5) OUTStr:50-53 (2.5) OUTAgi:26-36 (2.5) IN:EqualBlocking OUT:OffenseStrong ]
[INBlk:49-53 (0.5) OUTStr:45-50 (2.5) OUTAgi:28-41 (2.5) IN:EqualBlocking OUT:EqualBlocking ]
O'Brien throws to a covered Arthur at the UWW 9. Arthur makes the catch.
[BLK:41-41 (-1.2) RESULT:DefenseStrong]
[TKLAvoid:OutsideShort 57-44 (0) RESULT:TackleAvoided]
[BLK:36-41 (-1.0) RESULT:EqualBlocking]
[TKLAvoid:Medium 57-56 (0) RESULT:StrongTackleAttempt]
[TKLBrk:Medium 33-51 MHITS:4 RESULT:StrongTackle]

Is offense listed first? How do you come up with these ratings such as first line InBlk: 51-53 (0.5)? How do we interpret these ratings? What does "Agi" mean?
Hopefully, these are not just dumb questions.
Also, the PBP does not show offensive/defensive formations for each play, unless I am missing it somewhere.

I hope to post more about testing and especially the debug info.  I was mostly using that for my own purposes, so it's sort of in short hand, but I can explain it for those that want to also use that as part of testing.

For each play, we iterate through multiple steps until the play is dead.  I hope to post more on the plays similar to the "anatomy of ..." posts we've had for the 2.0 engine.  But basically, for rushing, the rusher moves through each area of the field where blocking happens, a tackle attempt can happen and if a tackle attempt happens, the rusher can avoid or break the tackle.  If not tackled, he continues on.  For passing, the QB checks targets based on passing distribution and once he finds one that he can throw to him and then the catch and YAC are worked out similar to rushing.  Passing has many other components like sacks and defended passes.

Basically all of that debug pbp is spitting out info about the various match ups.  Anytime we have multiple players involved, or in some cases one offensive player (like the rusher) against one or more defensive players, you will see the "0-0 (0.0)" format.  The textual info shows various results on the play.  Many times in the engine when simulating through a play it determines the relative effectiveness of certain events.  For instance, when checking how well the offense blocked the defense in various situations, it will determine the relative sucess as DefenseBreakthrough, DefenseStrong, EqualBlocking, OffenseStrong, OffensePush, which is least successful for offense to most successful to defense.  This gives much more info to use in the engine than the old numerical results.

So to parse the debug info shown:
[INBlk:51-53 (0.5) OUTStr:50-53 (2.5) OUTAgi:26-36 (2.5) IN:EqualBlocking OUT:OffenseStrong ]

INBlk - means this refers to the blocking match up inside the tackles at the line.
51-53 - this is the average relevant blocking ratings of the offensive players involved and the defensive players involved. Always in the format of OFF-DEF
(0.5) - this is the difference in influence (how much players are involved) between the offense and the defense.  In this case, the offense has half a player (0.5) more influence in this match up.

[INBlk:51-53 (0.5) OUTStr:50-53 (2.5) OUTAgi:26-36 (2.5) IN:EqualBlocking OUT:OffenseStrong ]

OUTStr - the blocking strength match up on the outside of the line
OUTAgi - the blocking agility (or you may call it speed) match up on the outside of the line

[INBlk:51-53 (0.5) OUTStr:50-53 (2.5) OUTAgi:26-36 (2.5) IN:EqualBlocking OUT:OffenseStrong ]

IN:EqualBlock - for the inside blocking, the offense and defense were relatively equal in this step.  Note that this doesn't mean the ratings are equal, but that the results of the blocking was equal.  There is a range of results that can happen based on the ratings and influence match ups and that is what would be "tweaked" if we need to adjust some results in the play.
OUT:OffenseStrong - for the outside blocking the offense held strong, they didn't overwhelm the defense (OffensePush), but they pretty much are on the winning side of the blocking to the outside.  This takes into account both the strength and the agility (speed) match ups.

The single instance BLK entries you see in the debug will be blocking away from the line, or down field blocking, but a similar breakdown.

For tackling check:
[TKLAvoid:Medium 57-56 (0) RESULT:StrongTackleAttempt]

TKLAvoid - means this is the check to see if the ball carrier avoids the tackle attempt (i.e. makes the tackler miss)
Medium - this is the location of the ball carrier. Locations go from Line (TFL), Short, Medium, Long, Deep and they represent the general concentration areas of the defense.
57-56 (0) - similar to blocking, this represents the ball carriers tackle avoidance rating versus the defender(s) tackle avoidance rating.  0 means its a 1-on-1 tackle attempt.  A -1 would mean there is an additional defender in on the tackle attempt.
StrongTackleAttempt - the result of the tackle attempt.  Can be TackleAvoided, WeakTackleAttempt, GoodTackleAttempt, StrongTackleAttempt and really represents how well the defender got to the ball carrier.

[TKLBrk:Medium 33-51 MHITS:4 RESULT:StrongTackle]

The first part is just like the TKLAvoid only for the tackle break match up.
MHITS:4 - this represents a value to indicate the ball carrier is losing momentum which can be caused by avoiding or breaking tackles. It factors in to how well the ball carrier can perform actions further down field.
StrongTackle - this is the results of the tackle. Values can be TackleBroken, WeakTackle, GoodTackle, StrongTackle.

How do you know which players are involved? It seems like you'd have to know that before you knew if the average relevant blocking ratings were accurate.

Posted by scrodz on 2/15/2013 5:59:00 PM (view original):
How do you know which players are involved? It seems like you'd have to know that before you knew if the average relevant blocking ratings were accurate.

For the debug info, I wasn't so concerned about the players involved because I check that using another method that I didn't want to clutter the pbp with.  I can expand that debug info if you guys want to see more information, but just realize that it will take up quite a bit of real estate in the PBP if I start adding player names to it.  Also, that rating also takes into account the per play adjustments for TECH which affects how consistently players play each play, so you could have the exact same players in two different plays with different block ratings due to the TECH adjustment.

If you want to see any other "behind the scenes" type numbers in the debug info, let me know and if I can include it I will.

Haven't reread the Rushing stream, but if the design is to go blocking level by blocking level I have a few questions.  Is the levels determined if the defense is in a rush defense vs pass defense?  Does Game Instinct come into account for Line Backers and Safetys who would step up in run defense?

Just wondering if too much is given to a running back for breaking the first layer of tacklers and why rushing totals are so high, or perhaps LBs and Ss are slightly misplaced on the field after the play starts?

"Also, that rating also takes into account the per play adjustments for TECH which affects how consistently players play each play, so you could have the exact same players in two different plays with different block ratings due to the TECH adjustment."

Thanks! That was what I was looking for.

Since one of the main complaints about 2.0 was the inability to accurately determine how the engine treated events, I would like to be able to delve into the code a liitle closer during the beta to compare some of the match-ups and the result. To fully understand the code - I have some questions.

1) "So to parse the debug info shown:
[INBlk:51-53 (0.5) OUTStr:50-53 (2.5) OUTAgi:26-36 (2.5) IN:EqualBlocking OUT:OffenseStrong ]"
Question: for the portion - INBlk:51-53 (0.5) - I realize that the "51" is the offensive players involved and "53" is the defensive players involved and that the offense has a  "(0.5)" player advantage - but what part does the 0.5 play? Is it a multiplier, some other offset, or something else. In this example - the 51-53 (0.5) seems relatively even and thus produces an "IN:EqualBlocking" result. But the OUT blocking seems to favor the defense for strength 50-53 (close) and for Agi 26-36 (advantage DEF) except that the offense has the 2.5 advantage giving it the "OUT:OffenseStrong" text. "OUT:OffenseStrong - for the outside blocking the offense held strong, they didn't overwhelm the defense (OffensePush), but they pretty much are on the winning side of the blocking to the outside.  This takes into account both the strength and the agility (speed) match ups."  What would be the final score given the 26-36 disadvantage to the offense with the 2.5 player advantage to arrive at the OffenseStrong decision? Which player attributes were used? You can tell be straight up - I'm not afraid of the math.

2) "For tackling check:
[TKLAvoid:Medium 57-56 (0) RESULT:StrongTackleAttempt]"
Question: So here we see the Offensive ball carrier has broken past the line and is in the middle portion of his run. He comes up against a defender and they are relatively even in ratings - BUT the defender rates a StrongTackleAttempt - the highest rating for this even 1 v 1 match-up. Why strong? Which attributes?
[TKLBrk:Medium 33-51 MHITS:4 RESULT:StrongTackle] : definitely advantage defense and the tackle is made. Which attributes make up this rating?

An example from a recently SIMed game and some questions:
[BLK:39-38 (-1.9) RESULT:EqualBlocking]                                       Off 39 - Def 38 BUT Def 1.9 player ad and Equal block???

[TKLAvoid:OutsideShort 28-52 (0) RESULT:TackleAvoided]       Off 28 - Def 52 BUT tackle avoided???

[BLK:44-36 (-1.0) RESULT:OffenseStrong]                                       Off 44 - Def 36 BUT Def 1 player ad and Off strong???

[TKLAvoid:Medium 32-40 (0) RESULT:TackleAvoided]                 Off 32 - Def 40 BUT tackle avoided???

[BLK:44-39 (0.2) RESULT:EqualBlocking]                                        Understandable

[TKLAvoid:Long 28-40 (0) RESULT:StrongTackleAttempt]             Off 28 - Def 40 AND Strong tackle???

[TKLBrk:Long 25-38 MHITS:2 RESULT:GoodTackle                       Understandable

Another play:
[BLK:0-50 (-4.5) RESULT:EqualBlocking]                                           Don't get this one at all.

I will answer this in more detail in the following week, but for now I want to explain that the results are not based on just adding up the ratings and producing a result, but rather pull from a distribution of results.  This is so we do not have the same play over and over again.  This is basically where the "tweaking" would be for the results.  For instance, an equal match up might have a few blocks where the defense win and a few where the offense win and some that end up even.  The stronger the defense, the more they win, but it doesn't mean they win all the time.  It could be that the mix is just wrong and I might actually be able to go stronger in limiting the results.  For instance, with the blocking result where the offense was 8 point favorite, but the defense had a 1-man advantage, I could limit it so that OffenseStrong doesn't come up.

For the blocking results where it showed 0-50, that's a case where the offense doesn't have any blockers and I have to figure out where that is happening and if that is what it should be.  If it is, then I think I can always set that to the defense strong.  if it isn't, then I have to find out what offensive players should be involved and why they aren't.

" The stronger the defense, the more they win, but it doesn't mean they win all the time."

This is a scary statement to the coaches who want ratings to generate the outcome. This to me means there is some randomness in the game, which is what we really do not want again. Make the number variable based on attributes. Use of appropriate attributes such as athleticism, durability, stamina and technique could provide variability to players core abilities. Use of the player advantage/disadvantage will produce variable results based on coaches settings. I implore that you attempt to eliminate all random decision outcomes of any type.

Posted by katzphang88 on 2/18/2013 1:12:00 AM (view original):
" The stronger the defense, the more they win, but it doesn't mean they win all the time."

This is a scary statement to the coaches who want ratings to generate the outcome. This to me means there is some randomness in the game, which is what we really do not want again. Make the number variable based on attributes. Use of appropriate attributes such as athleticism, durability, stamina and technique could provide variability to players core abilities. Use of the player advantage/disadvantage will produce variable results based on coaches settings. I implore that you attempt to eliminate all random decision outcomes of any type.

You can't have non-randomness in a simulation.  The degree of randomness is what I think matters.  For defense for each play, "win all the time" doesn't mean if they don't that they lose either.  They can get a neutral outcome, like for blocking the "EqualBlocking" result.  For each advantage/disadvantage range, there are corresponding result ranges.  I might have to tighten the result ranges down more, but I would still have some randomness.  The reason this is better than the 2.0 engine is that the probability checks are more in increments of 10% rather than 1%, plus there are some mechanics in place to help reduce the chance of an "all failed rolls" scenario where the better team just falls on the bad side of the result checks.

I will look at using more discrete results based on the advantages/disadvantages, but I still can't just add up the ratings and look at settings and then produce the yards on the play.  Even with the little bit of variation there would be in that due to different settings and adjustments for TECH, we would wind up with many plays that look exactly the same.  The only way to bypass that would be to create some other thing that represented how well a player did on a particular play, but then that would still be random and we would just be pushing the randomness down a level.  The random factor for TECH is probably about as much randomness that I want to push down to the player level (as opposed to the result level which we are talking about).

I will also talk a little more about the advantage/disadvantage mechanics in the following week and hopefully that can spur more discussion.

There has to be a little bit of randomness. I don't want GD 3.0 to be like a roll of the dice but more like poker. If you have a good team like a top 10 team (AK) going up against a decent team like a top 25 team (KJ), hopefully you will win quite a bit, but there is a chance there will be an upset that they get lucky and win- such as if they put in a good game plan and catch a break or two.

Posted by katzphang88 on 2/18/2013 1:12:00 AM (view original):
" The stronger the defense, the more they win, but it doesn't mean they win all the time."

This is a scary statement to the coaches who want ratings to generate the outcome. This to me means there is some randomness in the game, which is what we really do not want again. Make the number variable based on attributes. Use of appropriate attributes such as athleticism, durability, stamina and technique could provide variability to players core abilities. Use of the player advantage/disadvantage will produce variable results based on coaches settings. I implore that you attempt to eliminate all random decision outcomes of any type.

Football is a random game.  The best team doesn't always win.  That's why we play.  The way you describe is that the season should just be constant recruiting cycles and whoever got the best recruits, gets a trophy and we repeat.

I understand that its frustrating to put in all this work and have a team you didn't feel was worthy beat your team, but it happens.  I know it happens in CFB, but the best example I have is KU-TCU in basketball this year.  If that game happened in this game, people would quit.  TCU had no business being within 20 points of KU.  But they won.

I always like the quote "You don't have to be the best team in the country, just the best team on the field."

However, with the new formations and control, you should be able to take advantage of your strengths and their weaknesses to insure that you win all the game you should.  If you have a good interior line, and theirs is poor, you are now more capable than in the 2.0 to exploit that.  This fact in itself should really help curb the "random" game results.

I agree that in real life, upsets can occur, but usually are carried through by our emotionality of being human and rising to the occasion and motivated to play over our heads for a period in time. This is a computer numbers game - the numbers should speak. As a coach in this game - I recruit to get better numbers so I can win. That is how the recruiting is designed.

Ahrens wrote: "However, with the new formations and control, you should be able to take advantage of your strengths and their weaknesses to insure that you win all the game you should.  If you have a good interior line, and theirs is poor, you are now more capable than in the 2.0 to exploit that.  This fact in itself should really help curb the "random" game results."

This actually helps prove my point. The new formations, game planning, and the ability to make attribute match-ups in my advantage SHOULD provide my team with a decision point advantage. BUT any "random" result negates that after the fact of game planning. Some events as penalties, field position, injuries will be somewhat "random" for placement in the game. My biggest concern is that I have seen games where I play similar SIM teams, 50 - 60 points per player average difference, and one I win with my better players performing as they should, and the next they barely scrape by. Since these SIM teams can get emotionally up - why the difference? Game planning is essentially the same, but it is how the 2.0 game treats the attribute match-ups that is off and why gameplanning was not as important in 2.0 as 1.0. I can see how individual plays may be altered by the outcomes of random decision points, but I can't see in the longer view of the whole game that it should be that much different. That is why in the above example of the blocking and tackling code, consistent outcomes of each match-up should be expected when the offense or defense has the advantage of the  DefenseBreakthrough, DefenseStrong, OffenseStrong, OffensePush categories.