Assassins really demonstrate how dire projectile accuracy is

[Mog Flintlock]

There's something incredibly frustrating about watching a generic assassin fire 6-7 crossbow shots at a sleeping target only 8 squares away without getting a single hit.

Assassins have a functional precision of 12 in these sorts of fights.

[Una]

The manual says: "If the range from attacker to target is greater than Precision/2 – 2 (half the Precision, minus 2) then the missile will deviate from the target. The amount of deviation is equal to the range x 1.25 / Precision."

I believe that every point of precision over 10 counts double. So this assassin would have 14/ 2 - 2 = 5 squares of full accuracy, then it deviates. Since this is at 8 squares, it will deviate.

Deviation is then said to be: "The amount of deviation is equal to the range x 1.25 / Precision."

So it's (8 * 1.25)/14 = 0.71.

I don't know what this means exactly. Does this mean that it will deviate 1 square (rounded up) or does this mean that 29% of shots won't deviate after all?

It also says: "The game will randomly determine whether the missiles deviate long or short, left or right, or some combination. The actual distribution is a bell curve – most projectiles will fall within the middle of the deviation range, but some will land at the extremes."

Long/short and left/right are just directions, but then the last sentence talks about range. So is it talking only about the extremes of direction, or... ? Who the hell knows. I think it means that if a shot would deviate 5 squares, some might deviate for instance 3 left and 2 up, while another might deviate 5 left and be "further away" from the origin square that way. This seems to make the most sense with the formulas described.

In our case we only deviate at most 1 square, so that doesn't matter.

So we're left with either 100% deviation (no chance to hit) or 79% deviation (21% to hit). Let's assume the latter.

Since the target is the only one in their square, they will be automatically chosen for the 21% shots that presumably land in that square. We then get the formula for calculating a hit as:
Attacker: DRN + (Size points in the square)/2 +2 if magic weapon
Defender: DRN + (shield parry value x2) – (Fatigue / 20)

I don't know if the shield of your target applies if they are sleeping or not, I assume it does since these are the sort of things that tend to happen in Dominions, but let's discount it for now (if the shield is hit, there's still a hit anyway, it just subtracts shield prot from the missile damage).

So attacker will be (against a human): DRN + 1 (I assume it rounds down, but we cannot tell)
Defender: DRN + 0

At +1, the DRN table on the wiki (https://illwiki.com/dom5/dominions-random-number#:~:text=When%20a%20random%20number%20is,open-ended”%202d6%20roll.) says we have 54% chance to make it stick.

So all in all, depending on our assumptions, we have either 11.34% or 0% chance to ever hit a sleeping target 8 squares away as a 12 prec assassin!

That does seem pretty dire doesn't it?

Incidentally, if it's light and the assassin has 15prec instead because of that, his perfect accuracy range would be 8, giving us a 54% hit chance which is easy to calculate because precision doesn't involve a roll, and the attack/defence roll at the end will always be that 54% in this scenario. Still not a great hit chance for a /professional assassin/ aiming at a target so close by, imo! And good luck getting the assassin to shoot form 8 squares to begin with, if the AI decides to shoot from 9 squares, we're right back down to missing almost all shots...

[Sombre]

I believe, though I could be wrong, that a shot could also deviate 2 left and 2 right, landing back on the targeted square.

[ulzgoroth]

The language about the deviation makes me think that the 0.71 value might be used as the standard deviation of a zero-centered bell curve for deviation.

Originally posted by Sombre:

I believe, though I could be wrong, that a shot could also deviate 2 left and 2 right, landing back on the targeted square.

The manual's description does not suggest they handle the deviation as a random walk.

Though such a model could have some merit, possibly.

[MadAsgardian]

The LA Jomon ninja's poison shurikens do feel about 10% accurate, but the ability to auto route a commander-less army is pretty powerful I guess.

[Sombre]

Generally speaking I think a good lesson from this is script your assassin to 'attack rear'

[ulzgoroth]

Originally posted by Sombre:

Generally speaking I think a good lesson from this is script your assassin to 'attack rear'

Which leaves assassins kitted with ranged weapons looking rather silly, though.

[Necroscourge]

Ideal assassins use nasty magic knives/shortswords anyway like heart finder or serpent kryss. Some chassis' like Abyssian Slayers just start with dual poisoned daggers

[Scrapulous]

Assassination in this game is imprecise in pretty much every dimension. It's not "go sneak in there and kill this specific person," which is how I typically think of assassinations, it's "go sneak in there and kill whatever single person you happen to meet, under whatever circumstances." If they find their target in a tavern, they will not settle for killing their 'target,' but insist on killing everybody in the tavern, too, occasionally dying in the attempt. According to the manual, they even go into the stable to kill the target's mount.

None of it really says "surgical precision." So the low ranged accuracy seems pretty on-brand to me. Maybe they should think about smearing poison on their targets' underwear instead.

[Loggy]

Based on reverse engineering the manual is missing a few important details. Notably, it rolls a d100, if you rolled a 1 or 2 you divide by precision by 3, or if you rolled 3-6 inclusive you divide your precision by 2. I tried to reimplement the algorithm in a way that would calculate some odds of hitting various stuff.

Originally posted by Una:

Deviation is then said to be: "The amount of deviation is equal to the range x 1.25 / Precision."

So it's (8 * 1.25)/14 = 0.71.

I don't know what this means exactly. Does this mean that it will deviate 1 square (rounded up) or does this mean that 29% of shots won't deviate after all?

If you pretend the 6% divisions above don't exist, ~28.5% of shots would not deviate at all, and the rest would deviate one square. All things considered, you'd have a ~36.5% chance to hit where you aimed.

Note that a "deviation" is "pick a random point in a 3x3 square" and successive deviations do this multiple times a row. So with 1 deviation, you are 1 in 9 to still hit where you aimed. With two deviations, you're actually still 1 in 9 to still hit where you aimed because probability is weird like that. (Specifically, no matter what the result of the first deviation, there's a 1 in 9 chance to pick the option that "undoes" whatever the first one did!)

A quick run of dist=8 prec=12 into my reimplementation of the deviation algorithm throws up a ~35% chance to hit the target square. Specifically, the number of deviations and the probabilities of each (after factoring in 6% failures) I got were...

0: 0.26857142857142857
1: 0.6847619047619047
2: 0.03666666666666667
3: 0.01

Is it right? I have no idea, I never gathered game data to test it against.

(You also have to not miss the target in the square, this is simply odds for the thing to go where the guy aims it)

[Swirler]

Originally posted by Scrapulous:

Assassination in this game is imprecise in pretty much every dimension. It's not "go sneak in there and kill this specific person," which is how I typically think of assassinations, it's "go sneak in there and kill whatever single person you happen to meet, under whatever circumstances." If they find their target in a tavern, they will not settle for killing their 'target,' but insist on killing everybody in the tavern, too, occasionally dying in the attempt. According to the manual, they even go into the stable to kill the target's mount.

Ah yes, the Gavrilo Princip method.

[boozermonkey]

Originally posted by Necroscourge:

Ideal assassins use nasty magic knives/shortswords anyway like heart finder or serpent kryss. Some chassis' like Abyssian Slayers just start with dual poisoned daggers

By my experience it’s mostly assassins with bottle of living water and bug or asp spam pretty much overwhelmingly.

[Mog Flintlock]

Originally posted by Sombre:

Generally speaking I think a good lesson from this is script your assassin to 'attack rear'

I'm aware :P. I was expansion testing as LA Marignon and forgot to script an assassin. Watching him fail utterly vs a sleeping target makes me wonder why he's even carrying a crossbow.