We do appreciate that the current AI is confusing - and the issue has been raised. We're currently working on this system.
I'm really glad :-)
At the moment, when you aim at a target, the game simulates the shot in the background to see where the bullets will hit. If you have a 93% chance to hit (per bullet) and the enemy only has 2 or 3 HP, the chances are at least 2 or 3 bullets will hit, so the chance to kill will be 100% (remember, the game has already simulated the outcome).
I take your (and hippoblue's) points about the hit% being calculated for each bullet, and that actually makes a lot of sense to do that. However, whilst it's true that 'the chances are' that two or three bullets will hit, that isn't the same as it being a 100% certainty.
At 93% to hit, if you have three bullets being fired then the chance of no bullets hitting would actually be 0.00343% (1-0.07)*(1-0.07)*(1-0.07). Therefore percentage chance of at least one bullet hitting is 99.9657%. It is very likely that at least one bullet will hit, but it is not a 100% mathematical certainty. (I suspect what is happening is that the figure being shown to the player is rounded up as it's a standard whole number percentage. It's 100% to the nearest whole number.
Big deal you might think, so what if it's 99.9567% and not 100% - But it is potentially a big deal to the player when one of those '100%' kill shots doesn't kill.
At 99.9567% you're talking one occasion in every 29154 where every bullet of the shot will be a miss. Doesn't seem like much of chance, but let's say a player has 100 of these 100% kill shots in every play-through of the game. Now you're getting a missed 100% kill shot once in every 291.54 games (292 if we round up).
That's one player in every 292 players (who plays through the game) seeing a 100% guaranteed kill shot not actually make that kill. You know your projected sales figures, but you're now potentially getting quite a lot of players who have 100% kill shots not killing at some points during the game.
Those players potentially complain that the rng is broken - and you only need one of those players to upload the event onto YouTube before other players also potentially start deciding the same.
As far as solutions go, I think you've got two possibilities;
Either round the kill percentages down rather than up. While presenting a 99% kill chance when the figure is actually 99.9567% is actually less accurate than presenting 100% it also leads to a situation where the player's expectation is always met, the shot could hit, the shot could miss, but the player expects either outcome - that rare miss is going to be part of the player's expected 1%, overall the player will feel lucky if those 1% shots very rarely happen. (I think a 'lucky' player complains less than an 'unlucky' player).
Or simply don't present the kill % at all - I think it's enough to show the hit % and leave it at that. The player can extrapolate whether a shot is likely to lead to a kill or not.
- This week I have been mostly playing Chaos Reborn.