User Pass
Home Sign Up Contact Log In
Robbnva
offline
Link
 
This is just one of the examples, I used to HC another pee wee team where I only had 4 plays chosen and had the same problem but I can't get access or proof of that AI anymore cause the idiot owner sold the team back despite being in 4th place.

Anyway on to the Guppies game yesterday.

Auto Adjust VERY QUICKLY
10 plays chosen each with a 10% chance of being called
both boxes checked on the offense AI screen

HB Sweep Weak averaged 5 YPC - called 3 times
HB Dive averaged 2.97 YPC - called 18 times
Weak I HB Slam averaged 4.47 YPC - called 16 times

if the Sweep Weak was getting 5YPC why would it get called less than the Dive which got called the most and was the least productive


http://goallineblitz.com/game/game.pl?game_id=863192
Edited by Robbnva on Nov 12, 2009 09:39:36
 
Forbin
offline
Link
 
Auto adjust is based on success rating.

You had 1 weak-i pitch left go for 12.5 yards, and a quick perusal of the PBP looks like the rest of them went for very short yardage, which is counted as a failure.
 
Robbnva
offline
Link
 
ok well when I have time tonight using monsterkill I will look this over more closely, but I have seen isseues with this all season.

Mods don't lock as I will provide more info later but nothing to see here ATM
 
Forbin
offline
Link
 
I can say that the adjust in packages seems to work great for me - when set to "Very Quickly", it works TOO well. (I'd almost never suggest using very quickly for packages, unless the package has very few calls)

Had one game earlier this season where I expected the weak slam in SB, and the weak dive in SB to both be decent gainers. Those 2 plays ended up being called something like 40 times total. lol

That was the last time i used "Very Quickly" with a heavy use package. Quickly works well, and medium works by the 2nd half. Anything less than medium is useless, imo.
 
Hikariu
offline
Link
 
AA is working for my package.
 
TrevJo
offline
Link
 
1Q, >2:00
HB Dive 4 for 15
Weak I HB Slam 2 for 13.5

1Q, <2:00
HB Sweep Weak 2 for 13.5
Weak I HB Slam 1 for 4.5 (cumulative 3 for 18)

2Q, >2:00
HB Dive 3 for 7.5 (7 for 22.5)
Weak I HB Slam 2 for 10 (cumulative 5 for 28)

2Q, <2:00
HB Sweep Weak 1 for 1.5 (cumulative 3 for 15)

3Q, >2:00
HB Dive 3 for 8.5 (cumulative 10 for 31)
Weak I HB Slam 4 for 14.5 (cumulative 9 for 42.5)

3Q, <2.00
Weak I HB Slam 1 for 5.5 (cumulative 10 for 48)

4Q, >2:00
HB Dive 7 for 20.5 (cumulative 17 for 51.5)
Weak I HB Slam 4 for 18 (cumulative 13 for 66)

4Q, <2:00
HB Dive 1 for 2 (cumulative 18 for 53.5)
Weak I HB Slam 2 for 5.5 (cumulative 16 for 71.5)

Weak I HB Slam: I can see why Weak I HB Slam got called so much. By chance, it got called early, and it did well for you early, and didn't really do badly until the very end of the game.
HB Dive: This one I don't get. It seems like this one should have been called a lot less, certainly not the most. Makes no sense to me. Was it getting a lot of first downs or something? Are first downs even factored in to play success rate?
HB Sweep Weak: Not sure what to think although Forbin's first post may be the key. Two failures and one big success... I could see how that would lower the percentage overall. But to not get called at all for the whole second half seems strange. That does make me wonder if a big success (the 12.5 yd gain) is not factored in as much as it should be, compared to a moderate success (i.e. a 4 or 5 yd gain).
 
tragula
title
offline
Link
 
Originally posted by TrevJo

Weak I HB Slam: I can see why Weak I HB Slam got called so much. By chance, it got called early, and it did well for you early, and didn't really do badly until the very end of the game.
HB Dive: This one I don't get. It seems like this one should have been called a lot less, certainly not the most. Makes no sense to me. Was it getting a lot of first downs or something? Are first downs even factored in to play success rate?
HB Sweep Weak: Not sure what to think although Forbin's first post may be the key. Two failures and one big success... I could see how that would lower the percentage overall. But to not get called at all for the whole second half seems strange. That does make me wonder if a big success (the 12.5 yd gain) is not factored in as much as it should be, compared to a moderate success (i.e. a 4 or 5 yd gain).


No one knows what is successful for the AA. It is possible that a run for 3yds and 50 yds have the same weight for the AA. We certainly don't know what is the effect of a first down a TD or a turnover.

Analysis based on the scout tool is just not enough. You need to see if the play failed, and the avergage is not enough. Also a play may have being called from another package/input.

For the HB Sweep Weak you must take into consideration that when a play is successful the other plays it compete with are lowered down. Before the second half started their where about 20 successful plays called from two inputs. It is quite possible that the combined % for these plays is boosted to above 50% dropping the remianing 8 plays below 6.25% each. Put in a couple of failure for the sweep and it may be way below 5%, and have a very faint chance to be called. Even after the play succeeded once, it's % change will look like 5% -> 6%, not enough to take over the huge success that came from the dive and weak slam.

Unlike build related issues, Bort is making a mistake on not making this information public.


Edited by tragula on Nov 12, 2009 16:03:45
 
PING72
offline
Link
 
Originally posted by tragula

No one knows what is successful for the AA. It is possible that a run for 3yds and 50 yds have the same weight for the AA.


I don't think so. I've seen it a lot where early in the game a run is broken for a big gain due to broken tackles even though the defense run was designed to stop it. Even though it then goes for negative yards (especially if a QB rollout) repeatedly after that it keeps getting called over and over.

IMO there is a yard/play component to it that can sometimes screw you.
 
tragula
title
offline
Link
 
Originally posted by PING72
I don't think so. I've seen it a lot where early in the game a run is broken for a big gain due to broken tackles even though the defense run was designed to stop it. Even though it then goes for negative yards (especially if a QB rollout) repeatedly after that it keeps getting called over and over.

IMO there is a yard/play component to it that can sometimes screw you.


Maybe.

My guess is that the decision is always made after the play with no memory (makes more sense). However it is possible that a big gain, make such a huge impact that it takes many bad plays to turn the play off.
 
Forbin
offline
Link
 
I don't have a link, but I'm pretty sure Bort commented on AA before, and said that it is basically just a yes/no thing. If play is successful, increase %. If play is not successful, decrease %.

What determines success? Only Bort knows, I think. Probably something like 3 or 4 yards on 1st down, and a 1st down on 3rd & 4th down. Don't have a guess for 2nd down, but it may just be the same as whatever the setting is for 1st down.

Any completion for positive yards is likely a success, on 1st & 2nd down, though.


^ Pure conjecture.
Edited by Forbin on Nov 12, 2009 18:00:57
 
TrevJo
offline
Link
 
I hope the system is not that simple. Yes/No is not adequate when some results are really good and some results are really bad.
 
cosmoxl
offline
Link
 
Originally posted by Forbin
I don't have a link, but I'm pretty sure Bort commented on AA before, and said that it is basically just a yes/no thing. If play is successful, increase %. If play is not successful, decrease %.

What determines success? Only Bort knows, I think. Probably something like 3 or 4 yards on 1st down, and a 1st down on 3rd & 4th down. Don't have a guess for 2nd down, but it may just be the same as whatever the setting is for 1st down.

Any completion for positive yards is likely a success, on 1st & 2nd down, though.


^ Pure conjecture.


I think success is 1/3 the yardage on 1st down, 1/2 remaining yardage on 2nd down, and 1st down on 3rd down.
 
RMiller517
offline
Link
 
Thread locked by moderator.
 


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