Collection of helpfull Maxims

User avatar
Ritter Runkel
Neural Network
Neural Network
Posts: 498

Collection of helpfull Maxims

Post#1 » 04 Jul 2017, 23:17

Let's start a collection of helpfull Maxims for gladiabots. Whoever has a new maxime can copy the list and add the new one.

I start with these ones :ugeek:

  • When you push and one bot dies, improve your retreatment. If you get pushed and you die, improve your aiming. If your team evaporates after seconds, cry like a baby.
  • When you replay and press the pause button, think about what would you do, if you where a bot. Code the AI in a way that they behave as you want them to do.
  • If your bots make cool things and you don't know why, try to understand it.
  • Think about your AI architecture. It should be easy to read, maintainable, effective and powerful. If you miss one of this four goals, you won't archive the other goals.

pier4r
Skynet
Skynet
Posts: 3339

Re: Collection of helpfull Maxims

Post#2 » 04 Jul 2017, 23:59

Nice ones, I summon mcompany.
http://www.reddit.com/r/Gladiabots/wiki/players/pier4r_nvidia_shield_k1 -> Gladiabots CHAT, stats, insights and more ;

mcompany
Autonomous Entity
Autonomous Entity
Posts: 872

Re: Collection of helpfull Maxims

Post#3 » 07 Jul 2017, 07:49

pier4r wrote:Nice ones, I summon mcompany.

I guess I am being summoned for my old maxims a long time ago. Actually at some point, I wanted to actually write a full explanation of what is best on a small scale, partly to explain the complexity of the game on a small scale, and to explain the logic of those old maxims. However, if anyone haven't noticed, I've been rather inactive (or just simply inconsistent on daily playtime) lately and simply haven't found time to write it. I also wanted to write a whole explanation of what was not possible normally without tags (and how difficult most of that information is with tags, if possible) but I ran into the same problem with that. Anyways, the maxims:

1. When in doubt, score
2. When in doubt, move forward
3. If you are still in doubt, shoot

Do note that while I believe number 1 is definitely first, most or my AIs don't follow it for a couple of reasons. Also, looking back, I think 2 and 3 might need to be switched, but still, my implication that the main three optimal options that are available being moving, shooting and scoring still holds true

Miojo
Script
Script
Posts: 36

Re: Collection of helpfull Maxims

Post#4 » 18 Jul 2017, 08:57

mcompany wrote:
pier4r wrote:Nice ones, I summon mcompany.

I guess I am being summoned for my old maxims a long time ago. Actually at some point, I wanted to actually write a full explanation of what is best on a small scale, partly to explain the complexity of the game on a small scale, and to explain the logic of those old maxims. However, if anyone haven't noticed, I've been rather inactive (or just simply inconsistent on daily playtime) lately and simply haven't found time to write it. I also wanted to write a whole explanation of what was not possible normally without tags (and how difficult most of that information is with tags, if possible) but I ran into the same problem with that. Anyways, the maxims:

1. When in doubt, score
2. When in doubt, move forward
3. If you are still in doubt, shoot

Do note that while I believe number 1 is definitely first, most or my AIs don't follow it for a couple of reasons. Also, looking back, I think 2 and 3 might need to be switched, but still, my implication that the main three optimal options that are available being moving, shooting and scoring still holds true


Funny thing. For me is:
Flee>Shoot>Move>Score
(flee, shoot and move may change order depending on situation... score only after game has "ended")

Some more:
-Only flee when attacked.
-Stop being agressive when your bots get to 25% hp. You probably did something wrong but that can save a loooooooot of games.

Ein Stein
Automaton
Automaton
Posts: 131

Re: Collection of helpfull Maxims

Post#5 » 18 Jul 2017, 09:12

Miojo wrote:Funny thing. For me is:
Flee>Shoot>Move>Score
(flee, shoot and move may change order depending on situation... score only after game has "ended")

In T7W, SYP, and Kingmaker, I've seen a lot of "scoring while defending" players. That is, the shotgun tries to score when the other bots prevent enemy from pushing. Do you run this strategy Miojo? Or you just push like in other maps?
We cannot solve our problems with the same thinking we used when we created them.

User avatar
LuBeNo
Autonomous Entity
Autonomous Entity
Posts: 532

Re: Collection of helpfull Maxims

Post#6 » 18 Jul 2017, 13:05

He pushes and is very good at it.
Image
My algorithm of life: if(self.tired) sleep(); else if(self.hungry) eat(); else follow(Jesus);

User avatar
TheGnoo
Automaton
Automaton
Posts: 175

Re: Collection of helpfull Maxims

Post#7 » 18 Jul 2017, 14:11

Miojo wrote:-Only flee when attacked.


I have to disagree with that. If you're pushing and end up nearly 4 ennemies, you better run away before they start shooting at you, imho.

My maxim: "A few games a day keeps the Ritter Runkel away". It means that playing just a few games allow you to slowly improve your AI, but if you play too much, you will mainly face the top10 and it'll be frustrating.

User avatar
Ritter Runkel
Neural Network
Neural Network
Posts: 498

Re: Collection of helpfull Maxims

Post#8 » 18 Jul 2017, 19:03

TheGnoo wrote:My maxim: "A few games a day keeps the Ritter Runkel away"


Ha ha ha. That's why it's so hard to find a match against you. :lol:

I like your attacking style, it's a pitty that we don't meet more often :geek:

User avatar
TheGnoo
Automaton
Automaton
Posts: 175

Re: Collection of helpfull Maxims

Post#9 » 18 Jul 2017, 21:18

I'm actually back in the business ! I lost faith for a few weeks when I lost my AI, but this game is so wonderful that I had to come back :)

Another maxim I apply:

"Don't be sneaky if you want a better AI"

This means that you should keep your sneaky strategies for later, to win future tournaments for example. Until there, you will just give your idea to your opponent for a couple Elo points, and you don't improve your AI. That's not worth it.
And this is why I'm waiting so much for real tournaments, so we can prepare some strategies that come out of nowhere and have some delicious replays.

pier4r
Skynet
Skynet
Posts: 3339

Re: Collection of helpfull Maxims

Post#10 » 18 Jul 2017, 21:26

Ah in terms of useful concepts taught by the game, here is a list and I add more when I remember them:

https://www.reddit.com/r/pireThoughts/w ... h_concepts
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: Collection of helpfull Maxims

Post#11 » 18 Jul 2017, 22:30

TheGnoo wrote:Don't be sneaky if you want a better AI


That's a really good one. At the beginning I spent to much time for some sneaky tactics without improvement of my retreatment or aiming. I got some wins against top players but they fixed after every lose and I had to find something new. I improved when I deleted almost every map specific stuff and worked on a simple generic AI.

pier4r wrote:... Reddit ...

Nice collection, didn't knew that you already collected the maxims.

Miojo
Script
Script
Posts: 36

Re: Collection of helpfull Maxims

Post#12 » 19 Jul 2017, 01:47

TheGnoo wrote:
Miojo wrote:-Only flee when attacked.


I have to disagree with that. If you're pushing and end up nearly 4 ennemies, you better run away before they start shooting at you, imho.

My maxim: "A few games a day keeps the Ritter Runkel away". It means that playing just a few games allow you to slowly improve your AI, but if you play too much, you will mainly face the top10 and it'll be frustrating.

I have a lot of "100% shield flee" nodes for situations like that... If I flee before they attack then they can push me much easier. And 1 bot distracting 4 is a very good situation, that's why sg is so good.

User avatar
TheGnoo
Automaton
Automaton
Posts: 175

Re: Collection of helpfull Maxims

Post#13 » 19 Jul 2017, 04:23

Miojo wrote:And 1 bot distracting 4 is a very good situation, that's why sg is so good.


Couldn't agree more. SG is definitely the most interesting bot of this current meta.

pier4r
Skynet
Skynet
Posts: 3339

Re: Collection of helpfull Maxims

Post#14 » 27 Jul 2017, 00:19

About deploying new matches, improving one's best score (in absolute value. The relative score actually is more meaningful but harder to compute)

The following is, of course, interleaved with sessions of reworking one's ai.

- If you are at your best (in terms of score), try to improve: deploy
- if you are under your best, try to improve: deploy.

- if you dominate, prove it again: deploy
- if you don't dominate, you need data to work out edge cases: deploy

The holy masters in this are: ritter, chris and mcompany, plus others that are less present in the top league (like IRRobert, Arcueid & co)


in other words: when I'll dominate, I will spam more than chris.
http://www.reddit.com/r/Gladiabots/wiki/players/pier4r_nvidia_shield_k1 -> Gladiabots CHAT, stats, insights and more ;

pier4r
Skynet
Skynet
Posts: 3339

Re: Collection of helpfull Maxims

Post#15 » 04 Aug 2017, 09:54

- multiplayer career: test driven development, until one dominates. Like a laboratory, don't bother much to test with yourself as you will optimize only against yourself and not against the active playerbase.
- tournaments (not the unavoidable ones): production. Because there is no second, third, fourth chance in a tournament as there is in the ladder.
http://www.reddit.com/r/Gladiabots/wiki/players/pier4r_nvidia_shield_k1 -> Gladiabots CHAT, stats, insights and more ;

pier4r
Skynet
Skynet
Posts: 3339

Re: Collection of helpfull Maxims

Post#16 » 04 Aug 2017, 20:45

Another way to develop one ai (or parts of it) is to go via experience.

Either one tries to anticipate simulations, planning and developing a module, or one is lazy, one knows that a module may end with many modes to handle a situation, but just waits for defeats to fill those nodes.

Since Alpha 5.3.1 I do like this, for this I need defeats, so I fill nodes in modules that I designed long ago. Unless I find the design meh and I redesign it (but my Ai is pretty stable, in the macro design, since 5.2)
http://www.reddit.com/r/Gladiabots/wiki/players/pier4r_nvidia_shield_k1 -> Gladiabots CHAT, stats, insights and more ;

Jeffrey Harpen
Automaton
Automaton
Posts: 119

Re: Collection of helpfull Maxims

Post#17 » 11 Aug 2017, 23:39

Flee from the closest, attack the weakest shield - this one came quickly when I started to analyse my first matches, so can be helpful for the beginners.

Always know when to run - and this is what I am still working on. ;)

User avatar
Ritter Runkel
Neural Network
Neural Network
Posts: 498

Re: Collection of helpfull Maxims

Post#18 » 12 Aug 2017, 06:42

Jeffrey Harpen wrote:Flee from the closest, attack the weakest shield - this one came quickly when I started to analyse my first matches, so can be helpful for the beginners.

Not sure about this. I did several tests and flee from furthest worked better because you are faster to go out of range and minimize damage per second. Problem is zig zag going when fleeing. So now I'm working with flee from any attacking me (center of attacking enemies). Fleeing is smoother now.


Jeffrey Harpen wrote:Always know when to run - and this is what I am still working on. ;)

Oh yes. That's the difference between 1000 elo and 1700. And it's difficult but worth every minute you spend with optimization. :mrgreen:

User avatar
Kanishka
Skynet
Skynet
Posts: 1420
Contact:

Re: Collection of helpfull Maxims

Post#19 » 12 Aug 2017, 07:06

Ritter Runkel wrote:Oh yes. That's the difference between 1000 elo and 1700. And it's difficult but worth every minute you spend with optimization. :mrgreen:


I'm 1700. Now what?
Fixes break an AI more than bugs do. :ugeek:

Gladiabots Off-Topic Chat


My Stats: Kanishka_RN3;Kanishka_MiPad

User avatar
Ritter Runkel
Neural Network
Neural Network
Posts: 498

Re: Collection of helpfull Maxims

Post#20 » 12 Aug 2017, 07:13

Improve aiming :lol:

Let me try to rephrase in a Maxime:

  • First solve the tutorials and single player missions to get a basic AI
  • Improve retreating until you're at master league
  • Improve aiming when you reach GM
  • Improve harvesting and edge cases to get to the top 10
  • Don't try to optimize too much for specific maps, that only hides other bugs which needs to be fixed

Return to “Strategies”

Who is online

Users browsing this forum: No registered users and 3 guests