Kaiser Soze

we work mainly on internal forums right now, but recently we created a new Ai subforum and new alphas Ai, so we have started to talk about AI (and test it ). So we have lots of internal messages, some of them are Ai related.

Both of you have access granted.

In fact I'm currently reworking how the AI decides which unit is better to be built. I've found they do not work properly. In DC maybe (because most units are good vs all and they scale in strength during the game), but in Dc: Pro unit can be good agaisnt one thing, but suck against another. And I've found that many time the Ai decides to buld the wrong kind of unit. Mostly because it reduces the demand of the unit that are good against what the enemy has. So if the Ai can build 2 units with same armor and rating it chooses to built the unit that is more effective against the armor that the enemy has not units, and this leads to a disaster , so for me it's a must the change of that functions.

First simplified version of these function will be ready in a matter on days (1-2 days). I can pass you if you like. It will only focus a little more on building unit that are effective against what the enemy already has.

The problem is that while I was changing them I realized a few things. For instance, the unit rating and the weapon effectiveness are squad values. Meaning that the AI thinks that a squad with 1 tac is as powerful than a squad with 8 tacs. And try them with the same importance (regarding issues like building units for counter them or avoid unit that they can counter pretty well). So I'm trying to change that using the number of troop in the squad (and maybe taking into account the leader of the squad somehow). This way the AI will have a more realistic idea of the effectiveness of the enemy army and of its own army. This will lead to more changes (even in the unitstats.ai file), but I think I can handle them.

Also I've think that a unit with more than 1 weapon (like tanks) only uses 1 of them (the best available) in order to calculate its overall effectiveness (not testet yet, but I'm 90% sure). So the Ai may underestimate the efectiveness of units like predators or baneblade. I think this can be handled with the rating value of the unit. But I'll try to figure out how to deal with it.

Once I finished all these changes I can report them to you. But not sure if they will be of any use to you because I'm quite focus on DC: PRo and it has a few thing that the basic game does need (unit scales a lot more in DC: PRO, so the squad size it's a factor that changes during game and not a constant; we have a lot more specific counters and less "good vs all", so its important to build the right unit, etc.).

Well, more detector the better is not exactly true.
surely one detector may not be enough.
but spent resources on skulls instead of another things may be wasted them once you have enough.

The problem is that there is no magic number of detectors needed, so I don't think there is a point improving this function. Based on my test the Ai do not build too much detectors, so no problem. I just wanted to point this bug in case you have a fixed number of detectors you were aiming at, because the bug will make the AI build more than that number.
Right now the function always tries to have 1 detector unattached. and I've said, I don't see a problem with that.

Yep I think that too.
Because my next question will be than built just 1 detector doesn't seems to be enough sometimes

The problem with that is that is difficult (I'll need to calculate which buildings are present in order to figure about which upgrades are available).
and It will not know if the unit with the upgrade is killed.
So I guess I'll have to forget about changing the squads behavior using its upgrades.