Turn off target pos fluctuation for cheating AI

[hoijui]

it was already perfectly explained why this is normal, and yes it is. it is how it is meant to be. it was explained why your view of it is biased: you are not using a suitable point of view, which would be the one of the attacking team, not the one of a spectator with full-view.
i do not really know how an inexact radar signal would look like in reality, but fact is, that how it works in spring, is the way it makes sense, game-play wise.

with your method:
10 samsons approach an HLT on which they have no LOS. the 10 samsons shoot, each one at a different position. the player can easily visually evaluate the center of the 10 shots, and force-fires there. the same could be done by a widget. -> fail!
same problem would be the visualization. would you just draw a bigger dot, with the center at the exact position?

[slogic]

> the same could be done by a widget. -> fail!
Allowing custom widgets is a fail. Without them user would need _manual_ interaction in this situation.

Also _free_ ability for units to auto fire at radar blips is a Spring invention. By default it should be off and everybody would be happy with balance cause value of units like "target facility" would increase.

> would you just draw a bigger dot, with the center at the exact position?
No, let it fluctuate around unit position. If you again afraid of position spotting then see info above. Also for moving units the original position is not evident.

[hoijui]

so you want to force two quite important things on mods, which now are options, just to make your method half way viable. of course that will never happen, and even then it would still be bad, cause in early till mid game, doing this manually is totally no problem, and would be done all the time.
just accept it will not be done, no matter how much you think it is better. if you do not accept this, make a feature request in the appropriate forum please.

[AF]

I question the benefit of what your doing for the cost of effort it entails

[slogic]

I don't understand lawyer slang, sorry. It is technical forum.

[AF]

And I do not like rudeness, no matter how innacurate it is:

Lamens translation:

I think that what your trying to do gives an advantage. But that advantage is not worth the hassle you need to go through to acquire it. I also think that the same amount of effort elsewhere would yield far greater returns.

What your doing right now is avoiding genuine AI work in favour of cheat hacks. You would get a far better result from a minor resource hack for a fraction of the cost in time and effort, nevermind implementing genuine AI behaviours and all the beneficial emergent behaviours that arise from them making the AI look smart

[kremmy]

Problem: AI's units firing at inaccurate radar blips from weapon range don't always hit.
Solution: Give AIs perfect radar.

Wait, what? No, that's bullshit, where's your head at? How is this a problem in the first place? Units that fire-at-will will fire at radar blips within their weapon range. This is expected and proper behavior.

Do you understand what the problem is now?

Problem: AI is making no attempt at properly microing units so they are effective rather than standing there firing at edge-of-weapon-range rader blips that they keep missing.

Solution: Make the AI do something other than have the unit stand there being useless.

This is firmly "AI is not handling units in an I manner, A or not."
I'd rather not believe that bad micro by the AI is being used as a justification for this attempt to allow deeper AI cheating. Somewhere along the line we need actual AI improvements, not offloading the work to the engine.

[slogic]

Problem: AI's units firing at inaccurate radar blips from weapon range don't always miss until target is in its LOS hit.

Problem: cheating AI is missing shooting at enemy after direct attack command was issued on him when trigger "enemy spotted within weapon range" worked while unit was approaching the target making no attempt at properly microing units so they are effective rather than standing there firing at edge-of-weapon-range rader blips that they keep missing.

Do you understand what the problem is now?

[AF]

This is one of the many advantages of scouting AIs, scouts have uses beyond merely detecting radar blips, both observational, and tactical

[Forboding Angel]

Why not jsut give the AI a maphack? Seriously, doesn't every decent ai in spring use maph4x by now?

[FluXy]

A decent AI for me is one that doesn't use the maphack.
It would be nice if they had an option for that if they support working without it.

[hoijui]

the AI already has a map-hack, so it can always see all units, even without radar. but when enemy units are not in LOS, the AI units will shoot at the estimated/wobbled unit position.

[gajop]

I might have not understood the entire discussion completely but from what I saw the main argument is that AI's
ability to use perfect map information malfunctions when picking targets? If so, that feels more like a bug than normal behavior.

Creating AIs that can play competitively against human opponents in RTS games is a problem that hasn't been solved yet, and further getting into AI creation is not trivial, especially not in Spring (sometimes odd and mostly undocumented interfaces, not to mention a rather complex game engine from the AI perspective), we should still allow for things like perfect map information.

On the other hand, I didn't know a cheating AI can get things other than information on demand:

As we concerned cheats, my point of view is: GiveMeMetal(), SetMyHandicap(), GiveMeEnergy() and CreateUnit() functions are the most ugliest cheats AI can have. But they do exist! While we're arguing about adding control to much more moderate cheat.

. This questions the ability of comparing multiple cheating AIs against themselves or even against humans.

It would be nice if you could allow only certain cheats, such as, say, creating a cheating interface similar to Cheat::enable(Cheats::PerfectMapInformation) - or whatever the current C/C++ interface looks like.
This would hopefully allow for the ability to disable certain cheats for some games, as you would probably want to disable free metal/energy/unit cheats for AI competitions of AI vs Human competitions.

[AF]

This depends on which RTS you speak of. I remember well seeing Foreboding and co loose game after game to Shard, resorting to Double teaming a single Shard with rush tactics to try and defeat it. Most FFA's with a single shard turn into an everybody vs Shard affair.

So really its more about the complexity of the game. Cheating only gets you so far

[gajop]

I wouldn't know much about Shard personally, I've only seen one replay of it on evolution (it doesn't come to linux out-of-the-box and i'm too busy to look at those cmake files you mentioned), which btw looks like fun, and even though it managed to win in a 3 shard vs 3 humans game, it wasn't that impressing since people seemed to have played poorly - probably lacking any experience in evo.
As far as other AIs go, the best of which work for me (E323AI, KAIK and AAI), being E323AI isn't able to competitively play against me (and I'm definitely no pro) - on some veh maps I tested playing against it, I am playing about 50% against 3 E323AIs (it could be that I haven't set some setting in some file that makes it better)

Anyway, regardless of how good AIs are, cheating is really needed to get you started - even if you did manage to create a bot that's capable of defeating good players without it. You mustn't ignore the fact that you were designing/building spring AIs for years, and Shard was even part of your bachelors diploma, wasn't it?
Besides, I'm sure most people would be content if they could play competitively against a limited cheating (perfect map information) bot rather than a weak bot that doesn't cheat.

[AF]

EvoRTS is a different beast to BA and XTA, and the mechanics of games like EvoRTS favour AIs much more than humans, whereas the games like TA have units which needs to be managed using special knowledge, and the way the game is played changes drastically over time, yet it is still possible to use these units in their conventional roles

This means an AI that could outsmart the elite players at XTA wouldnt be able to stay in that position for long before what was necessary to hold it changed in nature. However, in some games the AI can excel at a niche or type of game play that means it does not become obsoleted playing to the strengths of a human player.

So far no such niche has been found in TA AIs to monopolize on, resulting in algorithms that try to best a human at what they do best.

For example in Kernel Panic, the AIs have reached a nice level in that they monopolize on their instant reflexes and their omnipresent focus, which is a major advantage over a human with limited focus and reflexes

[Johannes]

Nobody really knows mechanics of EvoRTS well at all, I've seen it beat players due to it doing a rush they didn't expect possible and that's pretty much it. Way too early to say if the AI is playing in the correct manner, or how well it's playing.

[AF]

That'll be foreboding catching people off guard, remember, foreboding maintains the Shard support for Evo