Would be useful for events/etc. that don't want to allow autopilot.

Like say, RP Clan League, where a good number of players actually would lose to autopilot.

I tried to ask fizzer to have some kind of limitation to the autopilot, for ladders as an example. He said that hardly the AI could beat a human player. But really, it seems to me that it can lol....

Prod 2.0 might not be that great against humans, but if we're talking Wunderwaffe or Cowzow (and taking into account that these AIs will only improve from here on out), this is seriously going to destroy competitive events with weaker players by giving them an incentive to just autopilot their way through.

Warlight's dangerously close to "pay to win" here.

Warlight's dangerously close to "pay to win" here.

Couldn't agree more! I think the autopilot feature should only be used in practice games, so there's no incentive to winning with Cowzow or Wunderwaffe, or another AI. I doubt anyone could get a ladder trophy from using the autopilot feature though, although that may well be a possibility in years to come

I seriously can't see how this feature is good for good players, it might only be used in very specific situations like when an opponent stals or when you play a huge map and you're too lazy to do all this huge ammount of moves in your turn. For a skill based only purpose, picks and moves made by the AI are still very bad as well as predictable.

However for lower skill players, I can really see the AI winning half or more of the games already, simply because it is picking way better then it did, and also making moves smarter. Against a newbie it's really possible to use it to win.

I'll support any limitation request of use of the autopilot that you do even though I'm a member. A game where you can pay to win isn't funny.

I think the danger here is in the future with AI improvements it could get to the point where the AI can beat players, I think it's better to be "safe than sorry" and put in a way to disable the auto-pilot in the template options.

Perhaps the counter argument is that a host can make a template that disables the auto-pilot in a game the auto-pilot was added for, ie. deliberatly long and drawn out matches, especially Real-time, that your opponent is just waiting to bore you to death.

Perhaps in the final feature (this is pretty much beta) the auto-pilot will just be a scaled back version of an AI, incapable of doing anything smart, simply attacking or expanding in all directions (to help finish off an end game, which it was designed for). I think this would meet both sides of the problem. For now it's nice to see how all the AIs work though :)

I agree with everything above. ^^ +1.

It'd be great if we had an AI account on every ladder (or multiple version of the AI) to:

a) figure out what % of strat players would lose to the AI

b) have a consistent-skill benchmark we can use to compare people that aren't on the ladder at the same time

Mhmm. Autopilot was indeed a bad decision. Great for trolling diplos (and for playing lotteries when you can't be bothered to manually issue 39 attacks), but otherwise bad :/

Yeah I think Fizzer is greatly overestimating the skill of bad to average players. Maybe enable autopilot with Prod 1.0 in ranked games, and the others only in practice games.

An option to tick off the autopilot would be fine.

I agree with beren, keep the prod 1 available for all, but take off the other ones. Prod 1 is the useless AI, but can be always useful in staled games or big maps.

Atm my 1v1 ladder run and 2v2 ladder run is by selecting the best available choice from the AIs, which does still take some skill to recognize why some variations of orders are better than others.

If anything, I think the AIs can at most help people understand expansion a lot better. From what I can tell, prod 2.0 and Wunderwaffe are pretty good at expanding but when it comes to fighting, Wunderwaffe does some questionable stuff. Wunderwaffe seems convinced that if there's a +1 advantage, it should delay and attack despite there being no real reason for the opponent to attack (which can lose the stack the AI tried so hard to build up). I don't honestly see how an AI could become sophisticated enough to fight correctly. There's a ton of variables I think about when setting up fighting orders, and every situation is different (there's no real generalizing pattern or rule)

I actually expect to win at least 50% of the games I'm playing on 1v1 ladder/2v2 ladder using the AI, but only because the AI can properly expand and until my ELO is high enough to be paired against players who can also properly expand I should win.

Maybe 3 options?

Autopilot Settings: Fully Enabled, Prod 1.0 Only, Disabled

I don't honestly see how an AI could become sophisticated enough to fight correctly. There's a ton of variables I think about when setting up fighting orders, and every situation is different (there's no real generalizing pattern or rule)

Tree search + ML. :P If someone has the time to adapt AlphaGo's core principles to WL, we're looking at a new Ladder #1 soon.

Edited 4/13/2016 19:17:57

AlphaGo-like principals are too expensive to be aplied in WL any soon. Maybe in a few years, but until then the AI will get better and better with manual incrementations done by their creators.

For Prod 1, maybe it should be allowed by default but removable from the template by the creator.

The goal of autopilot is to win a game that is already won, prod 1.0 will likely lose from a winning position since it overcommits to expansion (expands with 5s-6s and is very passive) Prod 2.0 and Wunderwaffe should not beat players who have enough game knowledge. Ps was able to defeat me from 3 wastelanded picks when I played using AIs. If your opponent knows how to expand efficiently, then an AI should not beat them. I actually have fun playing with the AIs so you'll see me stream from time to time.

prod 1.0 will likely lose from a winning position

True. That's how the single-player challenges (esp. Crazy and Insane) work out.

Prod 2.0 is definitely not a threat to me; I haven't played Wunderwaffe or Cowzow yet.

But this isn't about decent players; I'm bringing this up because there's a good number of players in RPCL2 that I can easily see losing to an AI.

Won't using the AIs teach them how not to lose to the AIs though? I don't disagree with you, I like it when everything in WL is customizable, so turning off autopilot should be a thing.

It would; the AI's are great for building basic skills.

But that would still throw off a competitive event by giving members free wins that aren't representative of their actual skill level.

Edited 4/13/2016 21:18:57

If the AI can beat 10-15% of players, we already have a pay to win feature at the lower end. What's the point in playing the ladder for someone rating below 1300 if someone can just beat them by using the AI?