Node Reactivity Depth Discussion(+10ms per node)

pier4r
Skynet
Skynet
Posts: 3339

Re: Node Reactivity Depth Discussion(+10ms per node)

Post#41 » 16 Mar 2018, 19:47

Tortuga wrote:It is very possible that soon, if not already, one or several people will find an almost perfect AI, which will remain within manageable levels of size.


I don't agree with this. Alpha 7 with no random maps and way less construct already shown that there was a continuous evolution of tactics, only the refinements required more and more effort to be seen (like a logarithmic progress).

I guess we are way far from a perfect AI. And in the game we were never near in my opinion, maybe around June 2017 after 6 months of alpha 7 we did a lot of progress, but still far away from the optimum.

For example in my AI there is so much to do (clean up and improve) that it will take thousands of games before I do enough progress and in the meanwhile much will change again.

One can see this over and over by the fact that who dominated in the past, when coming back after a while, struggled and not a little or anyway did not keep such a big distance from all the others. (you can see this from the leaderboard stats)

For example there was TcT, 84% victories, dominating in Feb 2017. Two months after he was shredded to pieces (and indeed he did not play much to avoid losing points).
Same now. In two months the today tactics will be overly obsolete. This is the nice part of the ghosts, that they will show this as long as they are compiled for the current resolution engine.
http://www.reddit.com/r/Gladiabots/wiki/players/pier4r_nvidia_shield_k1 -> Gladiabots CHAT, stats, insights and more ;

User avatar
Tortuga
Script
Script
Posts: 11

Re: Node Reactivity Depth Discussion(+10ms per node)

Post#42 » 16 Mar 2018, 20:19

This can be explained by the changes in the rules that have taken place in the elapsed time. As for the fixed maps, I think that the same strategies are always applied in a cyclic way, in stone paper or scissors style.

pier4r
Skynet
Skynet
Posts: 3339

Re: Node Reactivity Depth Discussion(+10ms per node)

Post#43 » 16 Mar 2018, 20:55

Alpha 7 (and alpha 8, that did not change much) was there fixed from the 21 jan 2017 to mid of July 2017. At first I too I thought "ok after 2-3 months, the best tactic is found". No, likely we find the next step of best - climbing on a logarithmic curve. But then came Miojo quite late, after nullpointr, tct, t800, ritter, wozza, me, mcompany, buzzefall, milkhunter, castlevania, lubeno and what not (see leaderboard history) and just surprised everyone with a very refined tactic that was never there, for 6+ months and 300k games.

Also I am pretty sure that at the moment, aside from the sg bug, milkhunter and miojo (and my AI if I fix all the conflicts) have a lot of potential waiting to explode. Chris will be there too, just a bit slower.
http://www.reddit.com/r/Gladiabots/wiki/players/pier4r_nvidia_shield_k1 -> Gladiabots CHAT, stats, insights and more ;

sparr
Script
Script
Posts: 12

Re: Node Reactivity Depth Discussion(+10ms per node)

Post#44 » 29 Sep 2018, 06:41

I like the idea of bloated AIs executing more slowly. Many other programming games give bots a certain number of CPU cycles at a time.

I do not like the idea of changing the base timing by class. This will wreak havoc with timers/counters that can currently be reused across classes.

ElvenMonk
Neural Network
Neural Network
Posts: 238

Re: Node Reactivity Depth Discussion(+10ms per node)

Post#45 » 29 Sep 2018, 06:54

First one would mess timers as well.
Not to mention that adjustments are now possible with 25ms increments only, but I hope both tickets are slowly moving out of scope for this project.

Return to “General Discussions”

Who is online

Users browsing this forum: No registered users and 1 guest