I doubt a warlight ai will ever be as good as a chess ai so i don't think we should worry about a scenario of that kind

It could potentially be a really solid AI, but I agree with you insofar as no one's actually going to spend the time to develop that great of an AI for WL.

I want to mention again that Warlight is based off the idea that a player cannot pay to gain any advantage in any games, and that if the AI eventually becomes superior to some humans, the autopilot can be used by an unskilled player to play and win entire games.

I like the autopilot idea, I just think it should be restricted to using the Prod 1.0 AI - which can take over for you anyways if you get booted.

A game option restricting players from using the autopilot feature would be nice too.

Edit: looks like this is fixed now

Members (or Fizzer/Norman), try this test case:

1. Open a Small Earth game where it's your turn but the opponent has already moved.

2. Make Autopilot, Wunderwaffe Run (not Run+Commit)

3. Commit manually.

4. Next turn loads, repeat steps 2&3

It crashes every time for me. Does it crash for anyone else?

TimerDelay::system.Exception Error (0): Exception: Error: RefreshIncome() Card not found
at warlight\shared\viewmodels\live\livestates\buildingturnphases\DeployingPhase.as:RefreshIncome()
at warlight\shared\viewmodels\live\livestates\BuildingTurnState.as:LoadOrders()
at warlight\shared\viewmodels\live\AutopilotVM.as:()
at warlight\shared\viewmodels\live\AutopilotVM.as:()
at vmutils\VmManager.as:()
at haxe\Timer.as:()
at haxe\Timer.as:()
(sp=6)


Btw, when I refresh, it works fine. It happens only when I run Wunderwaffe 2 times in a row

Edited 4/11/2016 13:52:37

Ok, there are 5 AIs currently available. So we need 5 membership accounts (i know you guys have a lot of alts with "M")
1. Join 1v1 ladder with such account
2. play ladder only with autopilot
3. check rating :)
Anyone willing to do this?:P

I'm on the ladder now, playing 3 games at a time, all using bots. Wunderwaffe when possible, when it refuses to work I use 2.0 or Cowzow, whichever looks smarter.

You shouldn't be able to use AI's in ladders in my opinion, while interesting seems like bs to me.

Edit: unfair if the bot plays better than a noob on the ladder, so the player uses it when against similar skilled noobs.

Also a waste of time for players who play the ladders for competition against human opponents who think and do interesting plays, not just to play an AI.


Besides that its not all bad, single player level sharing is cool, especially when the AI is improved now and isnt as easily manipulated into certain moves.

Edited 4/11/2016 17:34:51

Autopilot AI seems to go against the whole idea of Warlight imo, simply because people will abuse of it, and at one point or another it will end up with only AIs playing against each other.

Nice to see Turtles agree in this point. :)

The smart person can beat any computer but its possible for an average person to lose to it.

Autopilot AI seems to go against the whole idea of Warlight imo, simply because people will abuse of it, and at one point or another it will end up with only AIs playing against each other.

What's wrong with it?

It's quite easy to counter.

If anything it works as suggestions on picks and such.

@Motoki, you mean in WL? At the moment - yes, in the future - possibly not.

I doubt an AI could beat a good player when they try in the future, the task of creating an AI which could even play at a high level of play is not going to be done by anyone.

I just dislike that there isnt a setting in games to turn the autopilot feature off and that they can also be used in ladders (reasons stated in previous post).

The problem with Warlight AIs is that it's to complex to look by brute force into the future so minimax approaches don't work. The only thing I see possible are... let's call them "local lookaheads" helping you for stuff like finding the optimal expansion for a certain bonus.

Since WarLight's victory condition is very easy without real upsets (gain more income than the opponent by expanding, breaking bonuses and prevent bonuses geting broken) it's actually not that a big deal to not be capable to look into the future since there are no situations like me taking almost all your pieces but I'm still checkmate. For this reason a WarLight AI has to substitute the ability to brute force into the future with well thought through evaluation functions and a ton of other heuristics.

It would be quite interesting if effective Warlight AI was developed AND it was allowed in tournaments, ladders, and other competitive games.

I would expect the game then to drift away from standard map+settings (like the current "standard" Strategic 1v1) and towards changing the game each time you play: different maps, different settings, different challenges.

In short, an effective Warlight player would be one who adapts quickly and learns new strategies in every game. In an ever-shifting tactical landscape, I can't see an AI being more effective than a human (at least not in the foreseeable future).

Although that's a bit of a different direction for some, I, for one, would welcome that kind of Warlight (it's what I enjoy most, anyway - not perfecting my play on one template and one map, but challenging myself to develop different strategies with every new couple of games).

It's a little like the Battles in Ender's Game, for those who are familiar with that book!

Edited 4/11/2016 18:29:38

^ love that book

Quo vadis Fizzer, quo vadis WarLight :(

While I am eager to see better AI, I am against automatons which can do jobs instead of humans. If one wants to play against AI, or with help of AI, there are practice games.

Even uses of AI for booted/surrendered players should be avoided, or they could play for the same amount of time as designed in settings (which can't be more than 2 turns, currently is not implemented). Everything else is delaying of the game, artificial result, ignoring in-game deals (cooperative player suddenly become enemy), etc. In ladder games, use of AI should be prohibited or minimized to small amount of turns.

The fact that in team game, when one player is booted or surrendered, and he becomes AI, if AI becomes better or even with starting AI, the outcome can go to advantage weaker player over better ones.

Fortunately, for those who don't like AI replacing players, there's an easy setting for that: make sure that booted/surrendered players in your games do NOT become AI.

I agree, however, that making the AI "too good" is potentially a problem for competitive games. An option to use a "weaker" AI for those situations would be welcome.

@M.Poireau Yeah but autopilot should be able to be disabled too. :/

I'm watching the new AI use Blockade cards... and the uses generally seem to hurt the AI more than they help. (They usually Blockade high-value bonuses, even if they're not threatened by their enemies. Odd!)