AI CHEATING Implementation for Devour/destroy effects

No, ai doesnt cheat (at least not on pc side right now, i have no idea what happens to the tested console ai)

TL;DR: Make devour and insta kill effects a little less random, make statistics (of those effects) of a single player looking more fit to the actual % chance of the effect so that it would minimise the extreme unlucky/lucky cases

in practice: if there is 50% to devour for this certain spell, have a counter:

(separate counters for player and for ai, global counter or per game, whatever concept is more fit for devs)

the counters of that rng effect (in here example of 50% devour) would keep track of consecutive devours / consecutive devour fails for the respective player/ai and force break that consecutive strike after a fixed record, for 50%devour lets say after 3 devours or 3 fails it should force do the other thing, or: instead of forcing a 100%/0% chance tweak the % chance to smaller or bigger after the effect becomes consecutive, make the tweak stack until the consecutiveness gets broken

similiar with all the destroy/kill/annihilate whatever its called effects, beside the devour

good or bad idea? :stuck_out_tongue:

(yes yes i know generally ai shouldnt cheat but i think some kind of “RNG breaker” for the most extreme effects could only improve the gameplay, since… you know no player cares about statistics count and probability when their own luck totally doesnt fit the average of those statistic)


As someone who has logged a LOT of hours playing, there is absolutely no chance a player and the CPU are getting the same amount of devours, kills from death mark, etc.

Kerberos is not even remotely appealing to use on an invade team. The spell only ever devours when it doesn’t matter for me. Absolutely never in a close match when I legitimately need it to happen. Versus the CPU, it seems like the exact opposite. It only doesn’t devour when it wouldn’t matter much.

I tried using a team with Kerberos just yesterday. I didn’t play a lot of games, because the 10 or so I played went like I expected. I casted Kerberos’ spell 15 times, and devoured 1 time. That 1 time was the last troop that was about to die anyway.

I didn’t see an opponent with Kerberos during this test, but I ran into one opponent with Black Manacles. They casted BM twice, and it devoured one of my troops each time. My 50% triggers almost never, but 20% is downright consistent for the CPU. This isn’t recall bias, I write stuff down.

Just giving my input, because this definitely isn’t working how it’s supposed to. I can’t imagine how much it sucks for people without things maxed out. I feel constant pressure to kill anything with a % to devour before it can get mana. And while these troops should be treated as major threats, they should trigger less.


yeah, the numerous voices like yours is what inspired me to make this feature request…


Given that there are more 1 action kill solutions than there are not, it is really something of an annoyance. And you can’t protect every team member from a 1 kill solution given that the AI tends to get 50% more often than not. (I did something stupid and wrote about it… went out to see how often AI 1 kill devours would devour whole teams which clearly isn’t 50% or 20% or what nots)…

I have never understood 1 kill cards or spells or what not in games like this. But hey… it could be worse, you could be eating your team like Black beasty does :slight_smile:

Bad idea.

Let’s not get into the recall bias posts that this thread will inevitably attract.

But to the OP idea: no thanks. Whether said ‘counter’ tracking the randomness is visible or not, that’s just not how probability works. It shouldn’t have to skew the rolls based on whether you’ve been (or feel) lucky or unlucky previously.

If the counter is visible: how? UI doesn’t have much spare space and I doubt this’d be easy to show especially in small mobile screens. And what would it show? Your 50% devour from Kerberos is two casts away from a guaranteed roll? Or your spell won’t work for the next three casts because it worked too many times recently? Likewise we’d have to display the AI counter as well: how’s that next battle looking when you know the AI assassin hero has a guaranteed insta kill last troop (because 10% didn’t roll enough lately). You’d avoid said battle forever: too risky.

If the counter isn’t visible, the same issues arise plus worrying about keeping track of it yourself, which adds more impenetrable things to try and think about.

Whether visible or not, this’d lead to unnecessary changes in behaviour over needless complexity and knowing or worrying that a bad roll was guaranteed to come your way soon. Hard to code and easy for players to think is malfunctioning.

1 Like