Big AI flaw with dead leaders factions

[sacredceltic]

Again, I think you misread me.

Yes, AI troops of dead leaders, when they don't have allies to keep fighting for, prioritise occupying villages. Fine. I'm not debating this.

I'm talking of the behaviour of OTHER AI factions, that still have a LIVING leader, toward the factions above, whose leader is dead.
Rather than focusing on WINING (remember ? That's the point of playing...) against STILL LIVING enemy leaders (these are the only ones you still want to win against...), they waste time fighting remaining troops from dead leaders factions. That doesn't make sense.
A human wouldn't do that because it's a waste of time and energy. So AI shouldn't either.

When AI has the choice of several ennemies, it should prioritise the enemies that still have a living leader or that are allied to factions that still have living leaders. Giving priority against factions or alliances that have already LOST is silly.

[sacredceltic]

So far as I know the AI does not even distinguish between different sides. It sees only "enemy unit". I don't think it even sees or considers allied units at all.

Yes, well, then that's precisely the issue that should be corrected. AI is completely dumb. But not quite.
I know objectives are set to the AI, and that's very good. Surely there's a way to distinguish between enemy troops that have already lost and the others, on which fighting priority should be set. That seems so simple and obvious that it is how it should work.
I don't see a technical problem with solving that issue. It's plain.

You clearly ignored most of iceiceice´s post and the "developer" tag on his profile.

Being a good developer and being a good strategy analyst are 2 different things, requiring different skills.
The flaw I'm discussing about is OBVIOUS from a strategic point of view. It's just that developers forgot to think of the case.

I didn't elaborate on iceice post, because it starts with this idea that AI would be seeking "revenge", which is extravagant, since you claim yourself that AI doesn't distinguish between sides. How could it seek "revenge" then? against which side? Be consistent!
Additionally, iceice doesn't want flawed code to be rewritten out of fear it would break something (by the way, several different AIs are now available in Wesnoth, so I can't see how cloning the best of them, in order to add one more and correct it would break anything: if you love the flawed AI, you can stick to it, while others may choose to dump it)

I'm fed up with this constant rant about not changing anything, and that things are as they are, each time a flaw is spotted and an improvement is suggested. It's clearly not the way to go forward.
It's amazing to see how a group of people, initially interested in technology and gaming, can become so conservative.

[tekelili]

Again, I think you misread me.

You are right, I didnt understand you at first instance. However AI behavior that you claim to change, is still no clear for me. AI can not guess oponent skill and game style. Trying avoid fighting units from sides without leader alive, could be easyly exploited by human players. It also has not much "sense" avoid killing units under your front line that could stole villages and in last instance surrond and kill your boss.

[sacredceltic]

AI can not guess oponent skill and game style.

It's not a question of game style. It's a question of logic and setting the correct objectives to the AI.

It also has not much "sense" avoid killing units under your front line that could stole villages and in last instance surround and kill your boss.

Remaining troops of dead leaders have, indeed, the capacity to move from village to village and ambush your weaker troops, but usually not the capacity to surround your leader, unless you are very risk-prone to let you surround by a faction without leader (which is by definition weaker).
But you're right, the case might arise. However, I'm not saying that AI should not kill remaining troops from dead leaders AT ALL, but that their fighting priority should be set to the still living leaders, which is clearly not the case currently. That's what a human would do.

I can't see how distinguishing factions having a dead leader or not is a problem, since the status panel delivers that information. So it's available to the AI.

Of course, as human beings do, if no living leader's troops are at range, AI troops may attack dead leader's remaining troops, to take a village or secure a spot. I'm not debating that. But that it is still a low priority, compared to the high priority of WINNING against remaining LIVING leaders which is the condition of victory. Victory is indeed the number 2 priority after keeping the leader alive, but it is still the priority over anything else than that.
AI knows what objectives and priorities are, AI can know which leader is alive or dead, so it's just a matter of setting it properly. I don't see a coding issue here.

[tekelili]

AI can not guess oponent skill and game style.

It's not a question of game style. It's a question of logic and setting the correct objectives to the AI.

You keep doing same mistake: reducing analisys of AI behavior vs other AI. However AI code has to work in any scenario, included 3v3 maps vs human players. There is no way to weight AI tactic calculation (only evaluates its current turn) in base a human player leader dead. ¿Why AI should avoid attack a mage on swamp and prefer focus on a dwarf on hill because has still leader alive? Human players would prove such AI behavior stupid in 5 minutes.

[sacredceltic]

I never said AI priority flaw was against AI. I said it's again dead leaders remaining troops.

I'll explain again :

There are 3 players. 3 humans, you and me + Bill. Bill kills my leader and I have few remaining troops on the terrain. Now that I am dead, what is your PRIORITY? Beating Bill or eliminating my remaining troops?

If you waste time eliminating my remaining troops (which do nothing but occupy villages. This poses the same problem to you and Bill, so you can just as well discount the villages that my remaining troops still occupy), Bill may occupy more still unoccupied villages or even your own villages and subsequently get more troops.

Now if Bill is a AI player, it will try to fight ANY troop it meets, making no difference, in terms of priority, between my troops (with a dead leader) and you (with a living one), although you are CLEARLY a more potent danger to it.

Anyway you see the problem, whatever the number of players, AI behaviour, which doesn't set priorities between live leaders and dead ones is a BIG flaw.

I systematically beat AI players, even much stronger ones with much more troops, because they focus on beating every troop they meet, whatever their faction (with live leaders or not), and doing so, they mobilise a lot of troops, while I focus on stealing their villages and regrouping to beat them, while ignoring remaining troops from dead leaders (I just let them alone, as long as they're not a menace to my leader, and I quietly retake villages they leave behind them. Actually, they serve me to slow down the progress of my AI opponents)

[tekelili]

I never said AI priority flaw was against AI. I said it's again dead leaders remaining troops.

I'll explain again :

There are 3 players. 3 humans, you and me + Bill. Bill kills my leader and I have few remaining troops on the terrain. Now that I am dead, what is your PRIORITY? Beating Bill or eliminating my remaining troops?

You still try to set a narrow favourable scenario for your tesis. I gave you a different example that you avoided: 3v3 Waterloo Sunset in wich 1 team has 3 human players and other team is controlled by 3 AI.
And btw, I am an expert player beating AI and I know very well how exploit its weakness. If your idea could improve AI behavior, I would be suporting it.

[zookeeper]

Yes, it'd be great if someone improved the AI. When you run across people who are capable and willing, let them know.

[sacredceltic]

You still try to set a narrow favourable scenario for your tesis. I gave you a different example that you avoided: 3v3 Waterloo Sunset in wich 1 team has 3 human players and other team is controlled by 3 AI.
And btw, I am an expert player beating AI and I know very well how exploit its weakness. If your idea could improve AI behavior, I would be suporting it.

No I don't.
I don't know what "waterloo sunset" is. I imagine a scenario...But that is irrelevant. I'm talking of generic AI behaviour which should apply to all scenario and games with at least 3 sides, as long as the condition of victory is the defeat of opponent leaders (which is generally the case)
In the case you describe, you mention only 2 sides (= 2 alliances of 3 factions), so it is not concerned by the flaw I'm raising, which is about cases when there are more than 2 sides (3, 4, ...8, n sides) and 1 or more sides are beaten (according to the victory condition above) because their sole leader, or all their leaders (in the case of alliances) have been killed. Let's call it a DEAD side, to make it simple, whether it is composed of 1 or more allied factions.

The remaining ALIVE sides (these with at least one still-living leader within their alliance) should focus on winning against the other ALIVE sides to achieve the victory condition. But that is not what AI players do (whether they are member of AI-only alliances or not). What they do is continue their systematic behaviour of preying on whatever troops they meet, as long as they're enemies, regardless of wether these troops belong to DEAD sides or not. Doing so, they waste lots of time and energy, which enable human players still ALIVE to systematically beat them, the way I do.

This is a major flaw, because it makes playing against AI players BORING with more than 2 sides, since humans are 100% sure to win by focusing on ALIVE sides while AI players don't. If only that behaviour was changed, through setting the right goals in LUA, playing against multiple AIs would be far more thrilling.

[tekelili]

I don't know what "waterloo sunset" is.

That is pretty much problem here. It is a debate beteween a guy that evaluates how your change in AI behavior could affect any scenario, because he knows very well all BfW enverioment, and a guy that is only interested in discuss about his narrow sperience in BfW scenarios.

[sacredceltic]

again, you didn't read me properly from the start. Try reading what I last wrote...Your scenario with 2 sides is irrelevant for the case I raised.

[tekelili]

Waterloo Sunset (as any 3v3 map) has 2 Teams and 6 sides
Edit: Your tactic is stupid vs human players, because any unit is very used very clever and means a big threat vs AI. Units from sides without leader are very valuables as are not paying upkeep.

[sacredceltic]

That's where you don't READ me. No, 3vs3 is not 6 SIDES. It's 2 SIDES with 3 PLAYERS each. If you're allied, you're on the SAME SIDE. You SIDE with your allies, not with yourself...

Again, the flaw I'm mentioning arises in situations where there are MORE than 2 SIDES as in the following examples :

a) 3 different alliances of n, m, and p numbers of players, and 1 alliance is DEAD (ie all the leaders that compose the alliance are dead)
b) 5 players in a free for all and 1,2 or 3 or more players are already DEAD (their leaders have been killed)

If you tried to properly read me for once...

[Ravana]

Side is one [side] object. What you mean is team. Calling team "side" does not help your argument.

[tekelili]

If you tried to properly read me for once...

I will try as many times as claimed, but you still have to convince me about (not that your behavior is better for some scenarios) it is better for any scenario.