New experimental version

[Nelphine]

That's awesome. That's better than the high men in my game with chaos rift/doomsday/Armageddon against them.

I'm assuming that level of unrest is a bug.

(And 10 vs 30 should go to the 10 anyway - 30 is too far for the AI to have any ability to predict what the target will be. Even 2 v 6, most doomstacks will in practice reach the distance 2 this turn, even if in theory it might take 2 turns. That leaves 3 v 9, but by then 9 distance is so far that things could change too much, especially if the doomstack can't reach distance 3 in one turn. So yes. I'm fine in all cases with 3 times the distance resulting in the AI going for the closer target.)

[Seravy]

No, not a bug. Max taxes for 55% unrest, +4 for another 40% is 95%. With no unrest reducing buildings and no units that stays at 95%.

[Nelphine]

I've been thinking about it, and I think I have to change my stance on lair priority. Right now, AI clear out most lairs. I think this is a gameplay problem.

As a kid the thing I remember most about this game was dungeon crawling. The other wizards were the enemy and more dangerous, but I didn't care about them. Lairs let me take my heroes and conquer monsters.

But right now, before your change, the AI clears out 75% of the lairs in the world (if not more. I think on lunatic its closer to 90%.)

Making them conquer MORE lairs will leave even less for the human player.

Additionally, AI already have lots of cheating in place. Largely, they don't need any of the rewards from lairs, and so that becomes one more snowball in their avalanche of early game advantages.


So I think I want to suggest that both land intercontinental AI attacks switch to a 50 priority for lairs.


Please note, this does not affect towers or nodes. Both of those have long term strategic value, so the AI should be taking them out.

(By the way, the lunatic game I just lost, the first AI had a node melded on turn 16. I think the higher priority is definitely working.)

[Nelphine]

Oh. Right I forgot unrest modifiers are not flat numbers but instead 10s of percent.

Can we change that so that its the lower of the flat number or the 10s of percent?

(So 6 orcs would only create 1.5 extra rebels, but 15 orcs, it would only create 4 extra rebels)

If we can't do that change, I'm fine with 95% orc rebels.

[Seravy]

I suspect fixing the "doomstack skips a turn" bug has more to do with lairs getting cleared faster. Assuming a doomstack was hitting one target every 3rd turn before, now it can do it every 2nd turn. Priority only matters if the AI has multiple things to pick from and that's not very likely if they aren't in a war. (and if they are, war targets still rate higher than lairs, not the units but the cities definitely)

[Arnuz]

I actually approve of faster lair clearing as long as that entails less cheating. It makes the AI strategy similar to the human's.

[Nelphine]

The above two comments are a good argument for reducing gold/power cheating - the ai could be clearing lairs 50% faster which is a lot of gold and mana rewards that the cheating will not need to provide anymore.

[Domon]

well, that's one way of doing it - but

- strategic combat means they will be able to clear a lot of locations that will not be clearable by the player for a long time. not yhat of a big deal since the player can often use tactical combat to achieve something similar (sprites for example)

- hunting and getting treasures and using these to fuel one's own development is fun, and all that fun is wasted on the AI. this is actually something to consider. letting the player clear more lairs makes the game more asymmetrical between player and AIs, but in a good way.

[Nelphine]

I agree with that second comment. Lair hunting is lots of fun, and I think the human should do more of it. (I'd go so far as to try to vastly reduce AI lair, and declare that's the human version if AI cheating bonuses.)

[Seravy]

Finished the Expert game I was playing. I believe Normal and Advanced are fine as is.

Expert I'm unsure about - Most of the games went as expected (tho since we defined goals I only played one) but there are a few older lost games I marked red, and I don't remember why. Maybe I was playing a bad wizard or ended up in a 4v1 war or something, idk.
As this is fairly inconclusive data and we recently changed AI behavior, I prefer not to change the difficulty level right now. To be honest the numbers between Advanced and Expert aren't that far apart so if I had to reduce it, I wouldn't know where to start.
Advanced is :
100
66
130
130
160
130
100
50

While Expert is
120
50
148
160
180
150
100
40

None of the categories have a higher than 30% difference but most are 20 or less. Also, we calculated almost every AI casting related stuff with the assumption the bonus there is ~50 on any serious difficulty level, so I rather not mess with that unless absolutely necessary, and that is the biggest jump I see in the table, all the others are more subtle.

Unfortunately I'll be busy tomorrow, so I have to do the 5.0 release earlier - no time to play more games.