Goal Line Blitz - American Football MMORPG

Dryzzt23
offline

Jul 10, 2009 14:00:30

Link

How about pre-snap audibles for both OFF and DEF?

It can be as easy as an "if X, then Y" type statement

TS_SuperStar721
offline

Jul 12, 2009 15:24:24

Link

So if there's going to be pre-snap offensive motion, if the defense has a play that's "Man Coverage TE" on the DPC, would they follow the TE with the shift? If not, motion could be a total pain in the ass for DC's.

In the example play with TE motion, if the LOLB is set to Man Cover TE, if he's not going to shift pre-snap to follow the TE, then the TE will be wide open every time, and OC's could use that as an exploit.
http://test.goallineblitz.com/game/replay.pl?game_id=1201&pbp_id=266547

Edited by TS_SuperStar721 on Jul 12, 2009 15:24:49

r8
offline

Jul 13, 2009 08:27:49

Link

Got some more new plays

Enkidu98
offline

Jul 13, 2009 09:48:18

Link

r8 thanks for sharing these.

Could you let Bort know about my suggestion for allowing us to create offensive (and defensive) packages?

If we're adding more and more plays, the job of playbook management just got harder. And if people expand the number of plays they use by calling them individually LOTS of folks are going to be bloating their AI's even further.

Would be much better if we could just create a 'package' of plays and reference the package in the AI like 'Select Play' instead it would be 'Select Package' and the plays from that package would be referenced.

Cuts down significantly on the work needed to do an AI, saves space. Makes everything look 'nicer' and saves time.

Would help the defense as well. Being able to create defensive packages would allow Defensive AI's to call the same set of plays in certain situations without having to enter in each of those plays every time. Again.. just call the 'package' and it references the package to see which play to call.

Stephen A
offline

Jul 14, 2009 14:48:56

Link

great idea on the packages.
that would be a great way to start simplifying the AI and save coordinators a lot of time and mistakes.

r8
offline

Jul 14, 2009 15:48:43

Link

Originally posted by astephe
great idea on the packages.
that would be a great way to start simplifying the AI and save coordinators a lot of time and mistakes.

Yes it is a great idea, Bort is absolutely sold on it and will start working on it very soon

Joe Buck
offline

Jul 15, 2009 09:48:02

Link

Does this awesome package concept restrict formations? I.E Can we have a "run left package" that contains the I-form and Strong I and Singleback left-side runs?

Enkidu98
offline

Jul 15, 2009 10:14:16

Link

I don;t know what Bort will do with it, but in my conception this is how it works.

A 'Package' is essentially a premade set of outputs.

So, if in the normal AI you could create an Input, with 20 Outputs and in those 20 Outputs you give each play a 5% chance of happening and you select the specific play you want to call... you have a 5% chance of calling each specific play? If you want to do that more than once, you have to create those 20 outputs all over again. Creating room for error, taking up a lot of space, making your AI take a long long time to load, etc.

Correct?

What the 'Package' would do, is add the option under your 'Outputs' to select 'Load Package' in addition to 'Select Play'. When you choose the 'Load Package' it gives you a list of the packages you have defined on a separate Tactics Tab (Like the DPC tab). These packages are completely user defined, and so you could call those 20 plays in a 'Package' but you only have to make that package one time, and whenever you refer to it in the AI, it pulls from that same set of plays and chooses one at random based on the percentages you applied to them. So you only have to create that set of plays a single time. Big time saver with your AI's.

So, if in my normal first down AI I create in input:

Input - 1st Down, Between my 20 and their 20, Any time, any score, etc (normal Input settings)
Output 1 - 20% Chance - I could say, run a Strong I run play. (Use the basic settings and choose Run, Strong I, leave the rest basic.
Output 2 - 20% Chance - I could run a Strong I pass play (Again, pass, choose strong I, use basic settings for the rest)

The reason I am suggesting the above is just to contrast with the next outputs.

Output 3 - 20% Chance - Because I REALLY like the Cross Up Play, I want to have a good chance of running it on first down. So I assign this Output and use the 'Select Play' function to specifically call that one play. Now I have a 20% chance to call the Cross Up.

Output 4 - 20% Chance - This time I want to run a certain set of running plays. So I use 'Select Package' instead of select play and I choose the '1st Down Runs' Package. This means there is a 20% chance I'll run a play from the package named '1st Down Runs'

Output #5 - 20% - I want to also have another set of passing plays (beyond those used above) so I use the 'Select Package' option and choose the '1st Down Pass Plays' Package. It means there is a 20% chance I'll throw one of these passes.

Previously, I created these two packages (so they would be in the list when I chose 'select package')

1st Down Runs Package:
The top portion of the Package has some of the basic information that a normal output would have. So you could define the behaviours of players in the play. Like Say you want the TE pass blocking left in all the plays in the package, you would set the TE Pass Block Left just like a normal output. Ideally the package would also contain the ability to set the depth chart for every position _except_ the QB.

The bottom portion is where you add plays. You could add any number of specific plays, and assign each play a % chance to be run. So lets say for 1st Down Runs you choose 5 run plays and assign each of those a 20% chance.

So when your AI calls the package (which it would 20% of the time, it then performs another roll to determine which of the 5 plays you have in that package it runs.

It then runs that play.

Its mostly just a way to help compartmentalise your AI's especially considering we're going to get a significant addition to the number of plays that can be run and sicne there will be occasions that you'll have many more choices from one formation, you can't very well just remove plays from the playbook (as you may want to run them too) so you need a way to be able to limit the playbook for certain situations (EG... Packages) That way you can keep all the plays in your playbook if you want, but limit them to being called in certain situations and since many of these situations would be repetitive in your AI, having you create it only once, and then just refer to it as a 'Lad Package' means you have to input quite a bit less in your AI.

Makes AI's smaller and easier to manage and when you need to modify something you modify it just once in the Package, and then its fixed for every Input that calls that Package.

Time Saver, Space Saver, Sanity Saver.

And the Defense should basically have the same ability. Create Defensive Packages. A set of looks that you may want to run in some situations, with the same personnel and basic settings, and since there are multiple 'inputs' in the AI where this may be applicable, you can call them by loading the package instead of having to create an AI output for each of the plays, multiple times in the AI.

This gives OC's and DC's a reat anount of flexibility as well, as they can change packages very easily so you could, as a DC for instance, have a set of packages you use against teams who really like to run HB Screens. But since another team likes to Air it out deep, you can have a defensive package set up for better long coverage. So, when you change your ai's you change just which package its point to for play selection and its done. You don;t have to go through lots of inputs and change even more outputs. Just change every instance of where that package is called and you are done.

Since packages can be assigned a percentage at the 'top level' AI You can use them interspersed with normal play calling.

So using defense again.

You could have your 'Basic Package' and this is your set of defensive looks you like to use against a formation. You assign this package say 50% so that half the time in that situation, you'll call one of your basic defensive plays.

Then you could have a 'HB Screen Defense' package, and you set that to 50% as well. So the other half of the time it calls a screen defense play.
But the next week, you face that different team, so you just change the 'HB Screen Defense Package' (Leaving the Basic Defense Package alone) to your 'Cover Long' package.

This allows your 'basic' defense to go unchanged and you only add in/remove plays that you specifically scouted the opponent for and you only have to do the AI changes in one output. Rather then having to change lots of outputs and do them on many inputs.

Edited by Enkidu98 on Jul 15, 2009 10:57:48

Edited by Enkidu98 on Jul 15, 2009 10:24:44

Joe Buck
offline

Jul 15, 2009 10:37:06

Link

I could kiss you, but I'm not going to embarass us like that. that's perfect.

Mightyhalo
offline

Jul 15, 2009 12:11:09

Link

Originally posted by sbuck143
I could kiss you, but I'm not going to embarass us like that. that's perfect.

Now if I can just find the potential sbuck exploit. Way too excited about the suggestion worries me a little.

SikoraP13 DTD
offline

Jul 15, 2009 23:08:20

Link

lol sweet more of my plays were added

tycoon34
offline

Jul 16, 2009 11:05:32

Link

Enkidu...youre the man.

Octowned
offline

Jul 16, 2009 16:55:38

Link

I've been suggesting that packages thing for months, guess it takes a name to get it noticed.. I'm just glad it is being done finally. The suggestion has been up since january 20th in epic suggestions and speaking with Bort.

The suggestion also included various options, such as an "on/off" check box next to each INPUT, and a drag/drop or up/down arrow screen (much like the depth chart) that lists your inputs only (to reduce processing time), so you can re-order your AI much more effectively, click "update" and have the processor re-do your AI behind the scenes. Clicking "move down" 30 times is... well, quite annoying.

Enkidu98
offline

Jul 16, 2009 17:38:18

Link

I'm hardly a name.

I think there are some basic differences between the two suggestions and one just clicked a little differently and sounds easier to implement?

I'm just happy its being done, that it may be partially my idea means little except 'YAY' I can make better Oand D AI's and they wont take so long to load on my damn laptop

headsigh
offline

Jul 17, 2009 13:10:53

Link

If it isn't too hard, do you think you could test the following code?

Originally posted by
_scramblevector CheckForScramble(float vision, _playtype play, int scramblesetting, player* passrushers) {
case scramblesetting of
3: //this is scramble often
visioncheckresult == VisionCheck (vision, passrushers, preset_low_difficulty);
WRscheckresult == WRsOpenCheck(play);
closestrusherdistance == GetDistance(GetClosestPassRusher(passrushers));

if (visioncheckresult && !WRscheckresult && (closestrusherdistance < preset_safe_distance)) {
return GetSafeDirection(passrushers); //either WRs were open, the pass rushers were still not close enough or he didn't see them.
} else if (WRsOpenCheck(play) { return NULL }//Ideally this should be checked before ever checking for a scramble, so this is just a failsafe. This means the WRs are open. QB should pass.
} else if (!visioncheckresult) {
return [0,0] // stays put, didn't see the rushers. Brace for impact!!
} else return [0,0] // rushers are still far enough, no need for a scramble yet. Check next <preset number of ticks>

1: //this is scramble almost never
visioncheckresult == VisionCheck (vision, passrushers, preset_low_difficulty);
WRscheckresult == WRsOpenCheck(play);
closestrusherdistance == GetDistance(GetClosestPassRusher(passrushers));
RNGfactor = rndnumber(0.5);

if (visioncheckresult && !WRscheckresult && (closestrusherdistance < (preset_safe_distance * RNGfactor))) {
return GetSafeDirection(passrushers); //either WRs were open, the pass rushers were still not close enough or he didn't see them. Note the RNGfactor meaning that he will only scramble if he is closer than a number that goes between half the preset safe distance and 0.
} else if (WRsOpenCheck(play) { return NULL }//Ideally this should be checked before ever checking for a scramble, so this is just a failsafe. This means the WRs are open. QB should pass.
} else if (!visioncheckresult) {
return [0,0] // stays put, didn't see the rushers. Brace for impact!!
} else return [0,0] // rushers are still far enough, no need for a scramble yet. Check next <preset number of ticks>

/* scramble sometimes could be done in a way similar to almost never scramble, just with a different RNGfactor */
}

It's to see if this would increase scrambling for players who have it on "scramble often". I dunno the specific perimeters, though I guess it's the QB's tactics being set to scramble often and his throw-away frequency to aggressive.

Warrior General Games