Too Many Priority Levels a Bad Thing?

User avatar
Eric S
Hello World
Hello World
Posts: 5

Too Many Priority Levels a Bad Thing?

Post#1 » 27 Jun 2017, 04:03

I've spent a bunch of time prioritizing possible enemy combinations, trying to pick off the most vulnerable enemies first. Here is half of it:
Screenshot_20170626-205312.png
Screenshot_20170626-205312.png (140.38 KiB) Viewed 1796 times


My problem is that my bots spend so much time re-targeting to get a higher priority, that they never end up attacking anyone. (Example, match 704433)

I have a clause to continue attacking an enemy that I'm currently attacking, but it doesn't seem to be doing well.

I don't want to have to spend time tagging bots, but maybe that is the next thing to try.

Are there suggestions? Is it a lost cause?

mcompany
Autonomous Entity
Autonomous Entity
Posts: 872

Re: Too Many Priority Levels a Bad Thing?

Post#2 » 27 Jun 2017, 06:10

I know what you are doing and it is starting too much to look like this:
IMG_20170626_205519.jpg
IMG_20170626_205519.jpg (147.42 KiB) Viewed 1793 times


If so, can I please remove your inevitable headache and ask you to just try something a bit simpler
If not, then at most I'd recommend using the "selfAttacking" filter ("is attacked by me") for avoiding some of your target switching. After that, there isn't a lot you can do (even if you wasn't using that monstrosity)

As for whether or not that is actually a bad idea: it probably is pretty close to optimal, however, there are so many nodes and so much flexiblity that you can't do quickly with it that it is virtually impossible to maintain and improve

pier4r
Skynet
Skynet
Posts: 3390

Re: Too Many Priority Levels a Bad Thing?

Post#3 » 27 Jun 2017, 07:18

Nice picture eric. Good luck with maintainability and bugs.
http://www.reddit.com/r/Gladiabots/wiki/players/pier4r_nvidia_shield_k1 -> Gladiabots CHAT, stats, insights and more ;

User avatar
Ritter Runkel
Neural Network
Neural Network
Posts: 498

Re: Too Many Priority Levels a Bad Thing?

Post#4 » 27 Jun 2017, 13:15

Eric S wrote: Here is half of it:
Screenshot_20170626-205312.png

...

Are there suggestions? Is it a lost cause?


Imho the idea is fine. As long as you use the matrix for the trigger. Target locking should solve many things. And maybe you could think about two or three target locks at different positions. Then you can lock targets but switch when it is worth.

And you should make the aiming more bot-class-specific. Especially sniper and mg hate to switch target (but even there, there could be good reasons to switch).

One last idea is that you need to define when it's better to aim and pull the trigger and when it's better to push.

It's hard to analyze the replay. You played against one of the most advanced players. Switching targets could have several reasons. For analyzing it's better to watch replays of people you lost to with same elo +-300 points.

Watching high elo players is pretty good to get an idea what you can do with a SG or a sniper. It's super to learn about advanced tactics.

sollniss
Automaton
Automaton
Posts: 178

Re: Too Many Priority Levels a Bad Thing?

Post#5 » 27 Jun 2017, 14:24

You just have to do it right.
Attachments
Unbenannt.PNG
Unbenannt.PNG (198.56 KiB) Viewed 1766 times

Captain Spock (Castlevania)
Automaton
Automaton
Posts: 177

Re: Too Many Priority Levels a Bad Thing?

Post#6 » 27 Jun 2017, 16:55

Because of the aiming time, targeting is always a tough problem. As Ritter says, you only switch when it worth. There are several simple principles you can follow:
1. don't switch among targets at the same damage level.
If you lock an target, you would expect to continue attack until it run off the damage range. If you are a sniper, you can lock a target in all range. If you are a assault attacking a middle range, you can continue attacking until it flee to long range.
2. Prioritize low shield ally attacking ones
If you attack an 0% shield enemy, you may gain nothing because it is at the edge of your range. However, this is unlikely to happen if an ally is attacking it.
3. Prioritize higher threat ones.
For example, in an long range fight, you should push the snipers away first when it is possible.

This is the assault/machine gun attacking module when I was 2000.
Image
Hope you can find the principles in this example.

buzzefall
Algorithm
Algorithm
Posts: 67

Re: Too Many Priority Levels a Bad Thing?

Post#7 » 27 Jun 2017, 17:31

It is barely anything visible here, but I just wanted to show you the basic structure:
1. The module segmented. First level of segmentation is range. Thus there are 3 BIG segments: short, med, long range. In each segment you line up other priorities. This way in each range segment you can add more segmentation like
2. the second layer of segmentation is created by target lockers. for example you lined up targets for medium range. And somewhere in the middle you have put priority to shoot target alredy being shot - thus you have two parts of segment - locked fire, and target selection when it is BENEFITIAL enough. For example you will loose target that you shot at when it comes out of range or you was forced to retreat, no matter - you lost your target.

I marked by red color target-locked actions. Two last big branches are disconnected , wrong picture, assume it's connected (it's long range nodes)

Image

User avatar
Ritter Runkel
Neural Network
Neural Network
Posts: 498

Re: Too Many Priority Levels a Bad Thing?

Post#8 » 27 Jun 2017, 19:07

Buzz, Thats a nice explanation and a clean layout of your module
Do you use PC version or do you design it on a smartphone?

User avatar
Eric S
Hello World
Hello World
Posts: 5

Re: Too Many Priority Levels a Bad Thing?

Post#9 » 01 Jul 2017, 20:15

I'm not sure what was causing it to perform so miserably. I had in logic to continue shooting short/medium range enemies currently targeted by me. But I ran tests against my previous attack logic, and the new approach is far weaker. Perhaps it takes too many CPU cycles to check all the possible nodes? Or perhaps it is just a bad strategy. I had it prioritized by total HP of the enemy (shield + health by class). Obviously this can't take into account regenerating shields.

Back to the drawing board! Thanks for all the helpful encouragement!

pier4r
Skynet
Skynet
Posts: 3390

Re: Too Many Priority Levels a Bad Thing?

Post#10 » 01 Jul 2017, 20:58

The cpu cycles do not matter because the game is done in "virtual ticks".
http://www.reddit.com/r/Gladiabots/wiki/players/pier4r_nvidia_shield_k1 -> Gladiabots CHAT, stats, insights and more ;

Return to “Strategies”

Who is online

Users browsing this forum: No registered users and 1 guest