1700+ AIs (Alpha 5.2)

User avatar
NullPointer
Autonomous Entity
Autonomous Entity
Posts: 539

1700+ AIs (Alpha 5.2)

Post#1 » 29 Nov 2016, 13:57

Hi everyone,

I wanted to share a couple of my AIs that brought me to 1700+

Since we can't have private matches (for now), I'd like to see if anyone is interested in trying to beat them in the Training mode. Keep in mind that beating them with a counter AI that exploits a certain behaviour of these AIs won't be as interesting as beating with an AI you would use in Ranked matches.

-- Seven Wonders --

This one is still undefeated for almost a month now, that's about 50 to 80 wins in a row.

Formation:

white ---- red

A ---- D
B ---- C
C ---- B
D ---- A

All Snipers.

Image
Image
Image
Image



-- Set Your Priorities --

1 loss in 1 month because one Node was attacking out of range when not supposed to, but this was fixed.

Formation:

--SN
----- SN Mid Support
---------- SN Mid
----- SN Mid Support
--SN

All Snipers.

Image
Image
Image

I have a few videos to share too, but didn't have the chance to upload them yet.
Last edited by NullPointer on 22 Dec 2016, 19:18, edited 4 times in total.

User avatar
NullPointer
Autonomous Entity
Autonomous Entity
Posts: 539

Re: 1700+ AIs

Post#2 » 29 Nov 2016, 14:02

High resolution here: https://ufile.io/36724

Any feedback is appreciated :D

pier4r
Skynet
Skynet
Posts: 3390

Re: 1700+ AIs

Post#3 » 29 Nov 2016, 14:23

Thanks for sharing! I just observe that your AIs are way more specialized than mine (I use one AI per map, not per position/bot) but at the same time way simpler. I sometimes reach 100+ nodes that are difficult to handle.

PS: your post shows that who has good tactics and points is not avoiding to play, indeed for the lack of challenge you are even sharing your tactics! Kudos! I share mine in the website linked to my profile, but those are not so strong.
http://www.reddit.com/r/Gladiabots/wiki/players/pier4r_nvidia_shield_k1 -> Gladiabots CHAT, stats, insights and more ;

Christian
Algorithm
Algorithm
Posts: 56

Re: 1700+ AIs

Post#4 » 29 Nov 2016, 14:30

Sniper AI´s tend be quite small. They do equal damage over all distances so no need to optimize damage output regarding range. Distances are only needed for reacting to threats. Other bot classes tend to switch targets depending on range much more often.

ConsciouS-0nE
Autonomous Entity
Autonomous Entity
Posts: 740

Re: 1700+ AIs

Post#5 » 29 Nov 2016, 16:28

Null=0 vs me 2
0/2

I wish could fight you even more
Youre really good sir. Huge respect for the old man

Programmers>amateurs>..>..>. Me d*f$k ami doing here.
Im not the best player, but atleast im something.I'm 8
Science is madness! 3:)

User avatar
NullPointer
Autonomous Entity
Autonomous Entity
Posts: 539

Re: 1700+ AIs

Post#6 » 29 Nov 2016, 16:52

pier4r, the main problem with huge AIs is maintenance and, if you're using the same AI for multiple bots, a tweak for 1 bot may be a new bug for another.

I then switched from an individual complex AI to simple co-op AIs that would show enough complexity if put together and, if I find any bug, the fix would be isolated for that bot only. I also agree with Christian that Sniper AIs are small on its own already.

Conscious-One, your unorthodox AIs are amazing haha

ConsciouS-0nE
Autonomous Entity
Autonomous Entity
Posts: 740

Re: 1700+ AIs

Post#7 » 29 Nov 2016, 17:55

Im not creating ais serously untill i Make my second account.

'Accounts' when theyre implemented
6.0 tho
The real game starts to heat up
For now just for fun
Im even suprised the stupid ideas
Can reach apex points if done properly.
Im not the best player, but atleast im something.I'm 8
Science is madness! 3:)

pier4r
Skynet
Skynet
Posts: 3390

Re: 1700+ AIs

Post#8 » 29 Nov 2016, 18:03

NullPointer wrote:pier4r, the main problem with huge AIs is maintenance and, if you're using the same AI for multiple bots, a tweak for 1 bot may be a new bug for another.

I then switched from an individual complex AI to simple co-op AIs that would show enough complexity if put together and, if I find any bug, the fix would be isolated for that bot only. I also agree with Christian that Sniper AIs are small on its own already.


Completely agree but I always thought that having Ais for individual bots would require more maintenance.

Consider that my AI is based on invariants so I do like

Code: Select all

if not carry
  if enemy at range X
    if self type Y
      ..all the code for a bot here ...


So fixing for type of a bot is not difficult. The problem is that I canot have 3 bots aiming at the same target at the start (for example seven wonders) because I know I should have another AI that instead of aiming for closest (I'm making up) is aiming furthest so together they aim at the same target. I always refrained to do this because once I found a fix for a class type, then I have to propagate it on all the AIs.

Maybe I'm wrong and that's why I mostly suck.
http://www.reddit.com/r/Gladiabots/wiki/players/pier4r_nvidia_shield_k1 -> Gladiabots CHAT, stats, insights and more ;

User avatar
NullPointer
Autonomous Entity
Autonomous Entity
Posts: 539

Re: 1700+ AIs

Post#9 » 29 Nov 2016, 19:14

Propagating fixes through all AIs is very time consuming, you're right, but it still takes less time than maintaining one huge AI trying to behave in 2 different ways depending on which bot it's assigned to.

I played against your AIs, they are very good, I would just advise against this approach of 1 AI fits all, you will go much further. Even after the changes GFX is planning to encourage generic AIs, I think we will be seeing the top 10 using many different AIs anyway, even for the same class. The only difference is that they will be less especialised.

And this makes total sense to me. If you're playing Battlefield, with humans, you have same classes playing differently to achieve the same goal together, someone has to rush in while someone else flanks, otherwise it's not a flank.
Last edited by NullPointer on 29 Nov 2016, 21:34, edited 1 time in total.

pier4r
Skynet
Skynet
Posts: 3390

Re: 1700+ AIs

Post#10 » 29 Nov 2016, 20:50

Yes if you play as human, a simple game as shadow gun dead zone (or even older ones like call of duty1 and so on) requires different behaviors given different positions, weapons and so on.

So it makes sense that the decision tree for a bot does the same, but I just expected that the maintenance cost would have been higher. I will consider your point.
http://www.reddit.com/r/Gladiabots/wiki/players/pier4r_nvidia_shield_k1 -> Gladiabots CHAT, stats, insights and more ;

User avatar
Chunkymonkey
Automaton
Automaton
Posts: 139

Re: 1700+ AIs

Post#11 » 29 Nov 2016, 21:24

Wow thanks null pointer! Yours are very elegant compared to mine. And your points about specialization vs generalization are very interesting. =)

User avatar
GFX47
Dev
Dev
Posts: 2914

Re: 1700+ AIs

Post#12 » 29 Nov 2016, 21:49

Real sub-trees should help maintaining your AIs.

User avatar
Ritter Runkel
Neural Network
Neural Network
Posts: 498

Re: 1700+ AIs

Post#13 » 30 Nov 2016, 06:47

:o

I'm a little bit sad that you posted your AI's before the next update comes.

Now people will copy them and when I play against s.o. I don't know if I play against your bot or some other bot (and yes, it makes a difference)

I know that your bots are close to perfection but my joy to improve my bots to beat them just dropped almost to zero.

:cry:

User avatar
Ritter Runkel
Neural Network
Neural Network
Posts: 498

Re: 1700+ AIs

Post#14 » 30 Nov 2016, 06:48

Btw, very nice and clean approaches

mcompany
Autonomous Entity
Autonomous Entity
Posts: 872

Re: 1700+ AIs

Post#15 » 30 Nov 2016, 07:17

Well, at the very least you can go into the testing room and see if you can somehow defeat this setup with something of your own

User avatar
Ritter Runkel
Neural Network
Neural Network
Posts: 498

Re: 1700+ AIs

Post#16 » 30 Nov 2016, 07:21

This I also see when playing against null pointer.

Today a had a five hour flight and some of this time i used to think about new strategies. Now I wanted to test them but I realized that the multiplayer will be full of hobby null pointers soon.

Sorry but that was disappointing, even when it was definitely not the intention.

Christian
Algorithm
Algorithm
Posts: 56

Re: 1700+ AIs

Post#17 » 30 Nov 2016, 07:55

I guess right now the community is still too small and enthusiastic to just copy an entire AI for cheap victories. The whole fun of the game is watching the own AI´s performing. If I am willing to copy someone elses, why bother playing at all?

I haven´t noticed remarkable improvements on enemy ai´s on these maps yet. On the contrary, "all Sniper" setups have become rarer on these in the last days.

ConsciouS-0nE
Autonomous Entity
Autonomous Entity
Posts: 740

Re: 1700+ AIs

Post#18 » 30 Nov 2016, 07:57

I havent played a single game after null shared his ais lol
A lot will copy them, im waiting.

pier4r
Skynet
Skynet
Posts: 3390

Re: 1700+ AIs

Post#19 » 30 Nov 2016, 08:26

Ritter Runkel wrote:
Today a had a five hour flight and some of this time i used to think about new strategies. Now I wanted to test them but I realized that the multiplayer will be full of hobby null pointers soon.



I disagree.

(a) you overestimate how many people read this forum with consistency
(b) you underestimate how lazy are people (if they want to win, they will just create a second account and feed each other. It is easier). For example I appreciate the sharing and I will consider the approach (one AI per bot on a single map), but the fun is designing my AI not making a copy.
(c) Sharing for me is great. The bright side is, once a strong tactic is public, then there will be always someone trying to find an even stronger one. And that's the fun part.
http://www.reddit.com/r/Gladiabots/wiki/players/pier4r_nvidia_shield_k1 -> Gladiabots CHAT, stats, insights and more ;

User avatar
NullPointer
Autonomous Entity
Autonomous Entity
Posts: 539

Re: 1700+ AIs

Post#20 » 30 Nov 2016, 13:00

Ritter, you find ways of beating my AIs more than anyone else, I'm sure you can beat these. You just couldn't beat them yet because we rarely find each other in the game. This would actually give you a great edge against anyone who copies it.

I'm now wondering how long it will take for these AIs to become average. I had AIs in the past that kept me at the top and now they would be easily destroyed.

Return to “Strategies”

Who is online

Users browsing this forum: No registered users and 1 guest