Warrior General Games

Package auto adjust seems straight forward, +/- on the plays called within the package.



Does the Playcalling AI Input Auto Adjust only adjust the outputs for that specific input? If I'm set at 60/40 run/pass will the auto adjust run more passes if they're successful?

Originally posted by Kegatron
Package auto adjust seems straight forward, +/- on the plays called within the package.

Does the Playcalling AI Input Auto Adjust only adjust the outputs for that specific input? If I'm set at 60/40 run/pass will the auto adjust run more passes if they're successful?
Yes but using both can be a disaster, depending on how you set up the packages.

Let's assume you have 4 pass plays in a package. Against a certain team, only 2 of those plays work, but they make big gains because of a tagging glitch. If the 2 bad plays get called first, the Ai auto adjust might adjust away from the entire package, instead of letting the Package auto adjust sort out the 2 good plays.

do you recommend using the one in the package or the Input? I haven't used either yet as i'm still trying to figure out teh simple things lol

So I use both but, TJ is right. At first it (setting the AA to the high + side) just made everything a wreck. What I learned is to highly limit things to make the Auto Adjust work better.
My packages tend to have an average of 4 plays per package. The 'Auto Adjust' inside each package then only has 4 outcomes to choose from... limiting the 'bort craziness' that can happen. The AA inside each package >should< just work on the plays inside that package so if one play is working very well, technically the AA should choose that play more often. Where you can WRECK it is when you set your %'s on each play in the package. If you go with 25% on all 4 plays then AA will work fine. If, however, you make one play 40%, one 30%, one 20% and one 10%, then AA inside the package is going to fight what you want to have happen. That's when it makes ya mad.

As for the AA in the AI itself, on the inputs, this setting should adjust which 'package' gets used most (not the plays inside... just the package). Again... if you go with 4 inputs and set each to 25% then AA tends to work as intended (note... nothing is perfect but this causes a LOT less issues). However, if you go with 4 inputs and go with the 40-30-20-10% as above, then you get the same problem and that's when things get blown up.

If using 1 Play in a package or 1 input in a Output, then AA isn't needed. However, this also makes it much easier for a DC to counter you as they'll come to know what play is coming on any given down and distance. The lowest number of plays I have in a package is 2. The rest is all about the equality of the possibility of the %'s you choose.

Last bit for you. Separate Packages and Inputs in your mind. Packages should mostly be about a specific type or style of play. Examples; Package A has 4 run plays in it, all to the right side or maybe all off tackle right side. Package B has 4 run plays right up the gut but to the left of center. Package C has 4 short pass plays in it.
Perhaps all 3 Packages has the exact same Formations... whether you only use one formation of 4 different ones they would be the same inside each package.
Now... we think Inputs and we use those 3 packages for the input for 1st and 10.
That gives you 12 total plays the AI has to chose from. Now your Auto Adjust comes into play. If all %'s are equal (25% for each play in the packages and 33% for each Input) then it should work as intended... ignoring any play that fails too often and favoring those plays that work more often. If you don't go with equal %'s then you might not want to go with Auto Adjust because you are WANTING the higher % plays to run more and AA could wreck that. If just starting out as an OC, I highly suggest using the AA and making sure your plays and inputs are equal in their % chances. As you learn what works and doesn't, then you can move to change and take more control of those things. IMHO.
Hope this helps you.

Lordy, love me a good Theo reply. Thank you!

Currently I’m running the package plays all at 100%. I’ve experimented with a couple at 200%, but other than that I’ve left it pretty straight forward.

I currently only have four inputs with more than one output.

First down- A/B package …running/passing
Second down <7- A/B package …running/passing
Second down 3.5-6.5- A/B package …running/passing
Second down 0-3- A/B package …running/passing


The only thing I’ve messed with is the run/pass % in the output. So I won’t mess with AA on the single output ones, but what should I start with here? Adjust Medium?

Oh! Can't do that or you invite Bort's wrath!!! To explain;

So let's say you have a 1st quarter Output made for 1st and 10.
Inside the output is 4 Inputs.
Inside those 4 Inputs are ONE Package in each one; each Package with 4 possible plays inside from which to choose from.

#1, The %'s from the Inputs must add up to 100%... COMBINED... not each. Otherwise the Bort machine takes total control and that can lead to some really bad juju. So, using what I posted just above, your Output has 4 Inputs so you can go 25% in each (which equals 100% total) or two Inputs with 30% and two with 20%... or 40-30-20-10%... anything that adds up to 100. For sake of argument, let's say you only have 3 Inputs in that Output. In that case, if you wanted EQUAL possibility for each Input to happen, you would give each Input a 33% chance. Yes... that only adds up to 99 but it doesn't go PAST 100 so the AI will function as you want it to... and not as Bort 'lord-knows-what'.

#2, Same principal for plays inside those Packages. The % should not exceed 100 COMBINED. So 4 plays at 25% each or 2 at 30% and 2 at 20%... ect, ect, ect.
Anytime you have more than a total % chance of over 100 then things can go sideways.

As for setting, again, if your chance (% settings) are equal across the line (in both the Packages and in the Inputs)... and they add up to 100% total... then since they're balanced %-wise you should go with Very Quickly or Quickly. If, however, your % settings differ (either from possibility of play inside the package or possibility of a specific Input happening) then I would leave it in the middle and let the % settings do their thing.



I think I can give you a decent example of when and how to use and not to use Adjust Very Quickly.

So, let's say I have a Output for 4th down and 1 yard or less to go.
Inside that Output I have 2 Inputs.
Inside one Input I have a Goal Line Run package.
Inside that package I have 4 Goal Line plays selected; HB Blast, FB Dive, QB Rollout Weak, and HB Sweep Strong.
I want the HB dive to be the most called play so I assign that a 40% setting. FB Dive gets a 10% chance. QB Rollout Weak and HB Sweep Strong get 25% chance each. This adds up to 10% so my Package is good to go. But, because the %'s vary I don't want Bort messing that up so I go with the middle 'Auto Adjust' setting. (FWIW, I never go with anything below that... have never seen the need to.)
Inside the other Goal Line Package I have 2 plays selected; FB Screen Weak and HB Screen Strong.
The Fb Screen gets a 50% chance and so does the HB screen. So here I would go with Auto Adjust = Very Quickly because I WANT Borts' AI to recognize if one play is working much better than the other... since both plays have an equal chance of happening. If it works as intended then the Bort machine will select the play that works better.

Now... concerning the 2 Inputs... I might do as follows;
Input #1, with the Goal Line Run Package selected, would be the one I want to have happen most often so I give it a 75% chance setting. The other Input has the Goal Line Pass Package so it would get a 25% chance setting. That adds up to 100% so everything is now set for that Output.

Dang, swear I read that if you left them at 100% the sim would read a four play package as 25% 25% 25% 25%

LOL. Nah... other way around. TBPH, it took me a season's worth of all of these kinds of mistakes before I finally got it. Then several seasons of tryin stuff, askin questions, and watchin replays until I think I saw dots in my sleep... until I started to be able to put a winning product on the field.. at least I wasn't gettin thrashed by CPU teams anymore! Now I can handle most DC's at least on a level playing field... but the best are still pretty hard to beat. So while I know enough to be non-dangerous to myself I'm probably not a truly serious threat to the very elite DC's... at least not yet.

Originally posted by Kegatron
Dang, swear I read that if you left them at 100% the sim would read a four play package as 25% 25% 25% 25%

Sorry, I haven't read all Theo's tl;dr, but this is absolutely true.

Fwiw, personally...

For inputs I've always had them add up to 100 (ie 60% pass package, 40% run package) since it makes the most sense logically when trying to change things.

For actually packages I've always just left everything at 100%, always. Sometimes I'll set something to 150% or 200% if I'm trying to run it especially, but that's not something I do a ton (generally if your plays aren't running you have too many plays, ldo.

Theo, sorry for the callout but you're verifiably mistaken here.

Originally posted by WiSeIVIaN
Sorry, I haven't read all Theo's tl;dr, but this is absolutely true.

Fwiw, personally...

For inputs I've always had them add up to 100 (ie 60% pass package, 40% run package) since it makes the most sense logically when trying to change things.

For actually packages I've always just left everything at 100%, always. Sometimes I'll set something to 150% or 200% if I'm trying to run it especially, but that's not something I do a ton (generally if your plays aren't running you have too many plays, ldo.

Theo, sorry for the callout but you're verifiably mistaken here.

Huh. That's not how I understood it from Bort but I could be wrong, for sure. I've been running it the way I posted for many seasons now and it works perfectly fine with no issues. I wonder why two different 'rules' for basically the same thing (thinking game building/coding wise)? Can't say more... gotta head off to work. Will return later today and see if I can find that Bort thing.

Originally posted by Kegatron
Package auto adjust seems straight forward, +/- on the plays called within the package.



Does the Playcalling AI Input Auto Adjust only adjust the outputs for that specific input? If I'm set at 60/40 run/pass will the auto adjust run more passes if they're successful?

So Keg....


Free OAI

To head it off, I've ran an OAI since forever that has auto adjust in both the packages and the AI itself. My OAI is available (thread in bhall OC forum, or just let me know here with OC GM access) if you'd like a copy either barebones skeleton, or filled with some plays that got me a loss in last years WL playoffs. OAI construction is it's own PITA thing, which is different than OC'ing, and honestly most make their OAI too complicated to be able to easily adjust which makes them never adjust and be worse OC's.


My general understanding, which is from some combination of ancient bort quotes and dotball legend is...

Very Quickly = Max 15% change for successful/unsuccessful plays
Quickly = Max 12% change for successful/unsuccessful plays
Medium = Max 9% change for successful/unsuccessful plays
Slowly = Max 6% change for successful/unsuccessful plays
Very Slowly = Max 3% change for successful/unsuccessful plays

This is important because in "rare" packages like a 4th and short package with 2 plays, you really need it to be on "very quickly" to have any chance at doing anything at all for you. While on something like a 1st down run/pass AI input that's going to run 30x a game, you probably want it on very slowly so as to not lose all control I like to view auto adjust as a helping hand.


How it works

So lets say you have 5 plays in your output, and lets assume these are all set to even (100% each, which makes them 20% each). If one of your plays runs and is very successful, with "quickly" auto adjust, that becomes 32% while the rest of the plays split the difference (32%-17%-17%-17%-17%). Then lets say one of the 17% plays is run and is somewhat unsuccessful to a tune of a -6% penalty, that 6% will be given was +1.5% to the other 4 plays. It would then become (33.5%-18.5%-18.5%-18.5%-12%).

After a few plays you can see that there is a meaningful impact with low play outputs and high auto-adjust, cuz math's.


The problem

So the problem here, is that Bort is kinda retarded, so what is very successful vs successful vs neutral vs unsuccessful vs very unsuccessful is a bit cloudy and base heavy in theory. On 3rd down if you catch the ball short of the first down, what is that? What is a 3 yard run on first down? What is a 3 yard run on 3rd and 4?

Generally it seems to maybe almost sorta - in theory - be something like the following, based on maybe nothing or maybe something.

Any pass play (including screens).
Completion = Neutral at worst
Completion of 5+ yards = Successful
Completion of 20+ yards or TD = Very successful
Incompletion = Unsuccessful
Interception or Sack = Very Unsuccessful

Just noting that screens that are TkFL do not seem to have an unsuccessful marker possible as long as they are caught, which can lead to screens "taking over" in auto adjust if you have them in packages with high occurrence and high auto adjust. This has been the experience of many.

Any run play (doesn't seem to vary by down/distance, but who the hell knows).
0-3 yard run = Neutral
4+ yard run = Successful
11+ yard run = Very Successful
TkFL = Unsuccessful
Fumble = Very Unsuccessful


Wise's Current Auto-Adjust

So what does the 3rd best OC in GLB use?

Inputs (these are run/pass ratios)
1st down = Slowly
2nd and 7+ = Slowly
2nd and 4-6.5 = Slowly
2nd and 0-3.5 = Slowly
Every other input I have is only 1 package output

Packages
(A) 1st/2nd Run Plays (6-10 plays) = Quickly
(B) 1st/2nd Pass Plays (8-10 plays) = Quickly
(C) 3rd and 7+ (also late game passing)(3-5 plays) = Very Quickly
(D) 3rd and 4-6.5 (1-2 plays) = Very Quickly
(E) 3rd and 0-3.5 (1-2 plays) = Very Quickly
(F) 4th and 0-2 (1-2 plays) = Very Quickly
(G) Redzone 0-5 (3 plays) = Medium (noting that close in the redzone is pretty TD automatic, so this is just for player distribution)
(H) Redzone 5-10 (3 plays) = Very Quickly

===========

tl;dr Wise uses too much auto-adjust, bort called retarded for no reason, wtf does auto adjust do?

Originally posted by WiSeIVIaN
So Keg....


Free OAI

To head it off, I've ran an OAI since forever that has auto adjust in both the packages and the AI itself. My OAI is available (thread in bhall OC forum, or just let me know here with OC GM access) if you'd like a copy either barebones skeleton, or filled with some plays that got me a loss in last years WL playoffs. OAI construction is it's own PITA thing, which is different than OC'ing, and honestly most make their OAI too complicated to be able to easily adjust which makes them never adjust and be worse OC's.


My general understanding, which is from some combination of ancient bort quotes and dotball legend is...

Very Quickly = Max 15% change for successful/unsuccessful plays
Quickly = Max 12% change for successful/unsuccessful plays
Medium = Max 9% change for successful/unsuccessful plays
Slowly = Max 6% change for successful/unsuccessful plays
Very Slowly = Max 3% change for successful/unsuccessful plays

This is important because in "rare" packages like a 4th and short package with 2 plays, you really need it to be on "very quickly" to have any chance at doing anything at all for you. While on something like a 1st down run/pass AI input that's going to run 30x a game, you probably want it on very slowly so as to not lose all control I like to view auto adjust as a helping hand.


How it works

So lets say you have 5 plays in your output, and lets assume these are all set to even (100% each, which makes them 20% each). If one of your plays runs and is very successful, with "quickly" auto adjust, that becomes 32% while the rest of the plays split the difference (32%-17%-17%-17%-17%). Then lets say one of the 17% plays is run and is somewhat unsuccessful to a tune of a -6% penalty, that 6% will be given was +1.5% to the other 4 plays. It would then become (33.5%-18.5%-18.5%-18.5%-12%).

After a few plays you can see that there is a meaningful impact with low play outputs and high auto-adjust, cuz math's.


The problem

So the problem here, is that Bort is kinda retarded, so what is very successful vs successful vs neutral vs unsuccessful vs very unsuccessful is a bit cloudy and base heavy in theory. On 3rd down if you catch the ball short of the first down, what is that? What is a 3 yard run on first down? What is a 3 yard run on 3rd and 4?

Generally it seems to maybe almost sorta - in theory - be something like the following, based on maybe nothing or maybe something.

Any pass play (including screens).
Completion = Neutral at worst
Completion of 5+ yards = Successful
Completion of 20+ yards or TD = Very successful
Incompletion = Unsuccessful
Interception or Sack = Very Unsuccessful

Just noting that screens that are TkFL do not seem to have an unsuccessful marker possible as long as they are caught, which can lead to screens "taking over" in auto adjust if you have them in packages with high occurrence and high auto adjust. This has been the experience of many.

Any run play (doesn't seem to vary by down/distance, but who the hell knows).
0-3 yard run = Neutral
4+ yard run = Successful
11+ yard run = Very Successful
TkFL = Unsuccessful
Fumble = Very Unsuccessful


Wise's Current Auto-Adjust

So what does the 3rd best OC in GLB use?

Inputs (these are run/pass ratios)
1st down = Slowly
2nd and 7+ = Slowly
2nd and 4-6.5 = Slowly
2nd and 0-3.5 = Slowly
Every other input I have is only 1 package output

Packages
(A) 1st/2nd Run Plays (6-10 plays) = Quickly
(B) 1st/2nd Pass Plays (8-10 plays) = Quickly
(C) 3rd and 7+ (also late game passing)(3-5 plays) = Very Quickly
(D) 3rd and 4-6.5 (1-2 plays) = Very Quickly
(E) 3rd and 0-3.5 (1-2 plays) = Very Quickly
(F) 4th and 0-2 (1-2 plays) = Very Quickly
(G) Redzone 0-5 (3 plays) = Medium (noting that close in the redzone is pretty TD automatic, so this is just for player distribution)
(H) Redzone 5-10 (3 plays) = Very Quickly

===========

tl;dr Wise uses too much auto-adjust, bort called retarded for no reason, wtf does auto adjust do?

He's actually using a copy of your AI already.

I used his outline, but I didn't have any AA setting on. This helps a lot, thank you Wise

I use both. Optimum is to use two plays in a package. Put two packages in the input/ output. If you really want to maximize auto adjust you do this in the quarter interrupts to avoid getting stuck on the wrong play when your opponents defense adjusts and of course to avoid the repeat play penalty

Originally posted by Kegatron
Package auto adjust seems straight forward, +/- on the plays called within the package.



Does the Playcalling AI Input Auto Adjust only adjust the outputs for that specific input? If I'm set at 60/40 run/pass will the auto adjust run more passes if they're successful?