I've been asked several times if I could explain how the engine simulates certain plays, and recently the running play was brought up.  I can give the generic details, I think, without breaking the seal on the game to allow you a better understanding of what the engine is doing.  If it goes well, I may present a few more "behind the scenes" type posts with the caveat that when we work on the new update, it may change some of this.  But maybe we can get some ideas stirring on something we haven't thought of.

So here, in general, is how a rushing play is simulated:

1. Get the formation, style, and tendency from the playbook based on down and distance.  Just a random check here to pick the play based on the ones you have listed.
2. Set the offensive players on the field based on chosen formation, depth charts, and fatigue.
3. Set the defensive players on the field based on the offensive formation, depth charts, and fatigue.
4. Select play type from tendency - assume for this example that Run is chosen.
5. Select the rusher based on rush frequency settings. Just a weighted random selection based on the players currently on the field.
6. Determine if it is a wide rush or a rush inside based on style.  More aggressive will be more likely to run wide.  WRs always run wide.
7. Determine if rusher is stopped at line or behind.  Get offensive and defensive average ratings for rushing at location on field (more on that later) based on players on field minus any blocking factors for the ball carrier.  Select a tackler based on location that would be trying to make the tackle.  Factor in team ratings, ball carrier ratings, and defender ratings to get chance rusher is stopper.  If stopped, determine yardage.  If not stopped, continue on to next location.
8. Repeat this check for each new location with a wide rush getting one more check based on trying to turn the corner up field.

The engine breaks the field of play into different locations to progress through the steps of each play.  It's basically a grid from side-to-side and end-to-end.  The middle is where the play starts and then it goes out to the sides and then down the field.  So, for instance, a wide run would start in the middle and then move outside one location and then try to move down the field from location to location.  The locations down the field basically represent 0 or less yards (a stuff), 1 to 5 yards, 6 to 15 yards and 15 or more yards.  So if it was determined the runner was tackled in zone 3 down the field, we would then determine how many yards he got from 6 to 15 yards and that is based on the results of that initial check to see if he was tackled.  When we get into the 15 or more yards, additional checks are used to determine tackle chances incrementally down the field.  There is special code to check plays where the field is compacted (i.e. less than the 15 yards).

That is in a nutshell how the play progresses.  I think it's a pretty solid structure, but there are some places where we can make improvements.  For instance, when it factors in team average ratings and individual ratings it eventually just lumps everything together and produces a probability number to check for the result, where that could be broken down more into "smarter" logic to take into account mismatches in the teams.  More importantly, we should be able to show some of this play progression in the play by play and maybe come up with a way to show some sort of "good play/bad play" analysis like we do for our baseball defensive plays.

The one thing I'd like to emphasize about this is that while we don't track all the players' positions and every step they take, we can account for players' likely locations so even the team average ratings are taking into account the players most likely to be involved in the play.

Thanks Norbert.  This is great.  Would you please elaborate a bit on the following statement relative to 1)TE, 2)FB & 3 WR.

"...minus any blocking factors for the ball carrier"

thanks.

Thanks Norbert, that's great. 

Can you say how it would decide whether to factor a fullback's blocking rating in to the play? 

Posted by chalvorson on 12/29/2011 6:08:00 PM (view original):
Thanks Norbert.  This is great.  Would you please elaborate a bit on the following statement relative to 1)TE, 2)FB & 3 WR.

"...minus any blocking factors for the ball carrier"

thanks.

I am thinking that all players can block except the player actually carrying the ball.

Brad, I never asked just because it is too complicated for my old brain.  I just keep plodding along and get lucky on a few things ever once in a while.  I probably will never win a NC, but I still have fun playing the game. 

Depending on the location of the ball (i.e. the location of the ball carrier) it determines how much influence each player has based on the formation.  So a FB would have blocking influence near the line and some down field.  Pretty much the same for OL and TE, but a varying amount based on how far down the field and if rushing inside versus wide.  If the FB is running the ball, the RBs blocking would be a factor, but the RB generally has less blocking influence than a FB, so it's not quite the same as just flipping the roles.

I'm a little hesitant to go into too much detail because 1) it would take quite a bit of work to check how much influence each player has in each formation and in each location and 2) I don't know if after I get to dig more into the engine that I might not find some better ways to handle simulating these plays to where it might change considerably anyway.  So basically I don't want to spend a lot of time just to explain how some of this works when if I wait until I can actually start working on the engine I could use that time to not only dig into how it works but also improve it.

On what could be improved, for instance, I don't think the engine adequately accounts for the number of players blocking versus not blocking.  You should have an advantage in rushing if you have a FB blocking versus not blocking, or have a higher rated FB blocking versus a lower rated FB blocking.  But the way it is averaging in ratings for the team, I don't think it fully accounts for this.  It does affect it some, as when I run plays with various settings 1000 times, I can see the effects in the results.  But I think if it handled it more specifically rather than averaging in everything, we would see results conform closer to expectations more consistently (minus the whole "bad day" factor @bhouska  )

Posted by iamthetwo__2 on 12/29/2011 6:24:00 PM (view original):
Brad, I never asked just because it is too complicated for my old brain.  I just keep plodding along and get lucky on a few things ever once in a while.  I probably will never win a NC, but I still have fun playing the game. 

From one old goat to another, I'm still plodding too, sometimes shuffling, sometimes calling for 911.

Posted by iamthetwo__2 on 12/29/2011 6:24:00 PM (view original):
Brad, I never asked just because it is too complicated for my old brain.  I just keep plodding along and get lucky on a few things ever once in a while.  I probably will never win a NC, but I still have fun playing the game. 

Thanks, but that doesn't answer my question.  I honestly want to know, what are you guys going to use this info for? 

I feel like Norbert could give us every example of every type of play, but you are still gonna run the same plays, recruit the same players, practice the same, set the same lineups, etc, right?

(minus the whole "bad day" factor @bhouska )
 

LOL, thank you Norbert.  I hope you know we are on the same team, I just want us to end with the most fun game possible, and to me losing to a far inferior team is not fun.  That's all I'm saying.  If I knew exactly how to help you, I would.  But, all I can do is voice my opinion on what we're seeing now as results.  Sorry if it comes off as constantly complaining.

Posted by bhouska on 12/29/2011 6:21:00 PM (view original):
I appreciate the info, Norbert, but now I'd like to know from all the people asking for this info what they are deriving from it?  Are you gonna change your gameplans because of it?  Your lineup?  Your recruiting style?  Your recruiting targets?  Your practice plan?

I honestly want to know, I've seen lots of people asking for this info, how is it gonna help???

At the moment, it is merely a view at the workings of the engine. I don't believe this one post changes anyone's thoughts, but coupled with other insights, it is possible to build a "smarter" gameplan.

To be honest, I'm not sure how that would work yet, although i have some ideas. If we know that WRs always run outside, that leads to this question:

Norbert, when there is a run (let's say to WR2 side and TEs side) would more weight be put on those two guys' blocking rating than WR3 on the other side of the field?

And if so, would the defense have a counter system, ie the players on the other side of the field don't factor into the odds as much?

I'll add some additional information about the offensive team ratings and defensive team ratings for rushing.  Whether these are what they should be or not is a different story, but I think it would help you look at your running game.  You can look at these in light of the "team average" ratings I've mentioned above.

Offensive - Primary: Speed, Game Instinct, Blocking, Strength - Secondary: Technique
Defensive - Primary: Technique, Tackle - Secondary: Speed, Game Instinct, Strength, Elusiveness

Just an aside, I've never understood some of these ratings.  Like why is Technique separate from Blocking? Isn't Blocking really Blocking Technique?  Isn't Tackling really Tackling Technique? Wouldn't you combine those with Speed, Strength and Athleticism to tell how well a player blocks or tackles?  This is where I'm thinking once I start digging into some of these engine checks that it might make more sense to overhaul some of the logic rather than just try to tweak it.

Norbert, 

What do you mean by "Factor in team ratings"? Overall team rating? Why would that be taken into consideration since not all of your players are on the field?

If it is an average of all the players on the field, then I would imagine things like strength and blocking to actually be important for the WR's on the field, since that would lower your overall team rating?

I didn't ask for the information and I don't really want him to get any more detailed than that.  I just started the league but it will influence how I recruit (or whether I recruit) at fullback.  That was my main question.  Since you can't vary the practice schedule between fullbacks and runningbacks, I was considering not worrying about the fullback at all. 

It's probably less useful for someone like you who already knows how to win games, but for a newbie reading all these threads about how game planning doesn't matter, I wasn't sure how true that was. 

Sorry Norbert, didn't see your second post. That answered my question.

@bhouska using this information, you could set your depth charts in a way that is optimal for your style. If you are running with WR1, you can order your depth charts in such a way as to put your best blockers into position to block.