Raiders must have an invisible accuracy bonus.

[pfhorrest]

I've always noticed that sometimes, raiders tend to be crack shots no matter what their shooting skill or the condition of their weapons are. I've heard the same anecdotally from other players but the discussion always gets dismissed as "just unlucky RNG" or "bias in your observations." So, I've done an experiment with the same save file a total of 40 times and here are the results.

The scenario: My sniper, a careful shooter with a shooting skill of 12 and a 'normal' sniper rifle, versus a raider with a 'shoddy' sniper rifle and a skill of 3. The % chance for my sniper to hit the raider is 31%, and the % chance for the raider to hit my sniper is 3.4%.

In a whopping 28 of 40 trials (70%), my sniper was killed by a fatal shot to the torso or the head. In none of these trials was my sniper simply downed by a shot to the leg, which is surprising to me because it seems like a common outcome of facing enemy sniper fire. In only 12 trials did my sniper win, and in only 4 of these 12 trials was the enemy sniper killed outright.

In only 1 of the 12 trials in which my sniper was successful did it take less than four shots to incapacitate or kill the enemy sniper. In some of the trials in which my sniper was killed, my sniper fired eight or more shots without hitting once.

For reference, the storyteller & difficulty was Cassandra Extreme.

This test tells me that either raiders have an invisible accuracy bonus which is not told to the player through the mouse-over hit & chance, or that fights between specific pawns under specific circumstances are weighted heavily towards certain outcomes, whether this is arbitrarily decided or a factor of RNG, either way it is not reflected by the statistics or the equipment of either the raider or the colonist.

If anyone has done similar trials I would like to hear about it. If in fact one of the above is the case I would like that to become incorporated into the game via data accessible to the player somehow. If an enemy pawn has an overwhelmingly good chance to hit one of my pawns despite having seemingly awful gear & stats I would like to know that information when I mouse-over.

[Rock5]

Maybe, just like it's hard to down and not kill raiders when the storyteller decides you have enough people, maybe if the storyteller decides you have too many people it ups the chance that your colonists die. Did you switch storytellers?

[pfhorrest]

Maybe, just like it's hard to down and not kill raiders when the storyteller decides you have enough people, maybe if the storyteller decides you have too many people it ups the chance that your colonists die. Did you switch storytellers?

This is interesting, because in that save file I do have 8 colonists. The game's been Cassandra the whole way through.

[Alenerel]

maybe it is related to difficulty. you could try to check the same cassandra but in base builder.

[A Friend]

It's a conspiracy.

[Zhentar]

I've done a fair bit of looking at the code involved, and haven't seen any sign of any weighting based on faction or difficulty (except for the random insta-kill on downed for non player pawns).

But your numbers do seem pretty far off from where they should be. If you want to run more tests, try enabling the "Draw Shooting" option in the Den Mode view menu and see if it reveals anything enlightening.

[O Negative]

Yeah, 40 trials might seem like a lot. But, it's not, really. Still plenty of room for data skewing with only 40 trials.

I've been in situations where I was vastly outnumbered, and killed every last raider without a single injury (no killbox).

You've somehow angered the RNG Gods, I think

[RemingtonRyder]

In any situation where pseudo-random numbers are used, you can get results that defy probability. Random number generators (in computing) simply fish numbers out of a sequence.

In a fairly bad RNG, that raider can 'steal' the number you needed to hit him because each time a random number is needed it's the next number (frame) in the sequence.

In a slightly better RNG, each entity which can use random numbers has their own frame. The sequence of random numbers is the same, but it's being accessed in an instanced fashion so that all entities are equally affected by pseudo-randomness and can't 'steal' random numbers from each other.

Slightly better than that is having each entity's frame change over time even if they don't use a random number. That way all entities are still equally affected, but they are no longer pulling sequential numbers from the sequence.

[carbon]

The scenario: My sniper, a careful shooter with a shooting skill of 12 and a 'normal' sniper rifle, versus a raider with a 'shoddy' sniper rifle and a skill of 3. The % chance for my sniper to hit the raider is 31%, and the % chance for the raider to hit my sniper is 3.4%.

In a whopping 28 of 40 trials (70%), my sniper was killed by a fatal shot to the torso or the head. In none of these trials was my sniper simply downed by a shot to the leg, which is surprising to me because it seems like a common outcome of facing enemy sniper fire. In only 12 trials did my sniper win, and in only 4 of these 12 trials was the enemy sniper killed outright.

What's not mentioned here is whether there's is a mismatch in the amount of armor the two are wearing. If we're discussing a Nudist v. Power armor battle, then aiming is far less of a determining factor.

Anecdotally, I routinely find myself in prolonged 1v1 matches with raiders with inferior effective accuracy, and I almost always win (which is the whole idea). Now, I don't play on extreme, so maybe that might have a hidden influence.

Does anyone know for sure if it is possible for hit shots to be blocked by terrain under certain circumstances? I feel like I've seen the occasional situation where a sniper has been convinced they can curve a bullet around a rock outcrop half way between the shooter and the target, but all of the shots go into the rock outcrop. So something like...


_________|________
A________|_______B
_________|________
|||||||||||||||||||||

where A is the shooter, B is the target, '_' is open terrain and '|' is wall (usually happens with weird diagonal angles). Not sure if the OP has a weird case of this happening, but it is worth ruling it out.

[Zhentar]

Yes, projectiles (whether you rolled a hit or a miss) can be randomly intercepted by things in the path with at least a .2 fill percent, which includes rock chunks.

However, I would expect it to be symmetrical for two shooters targeting each other.

[daduhweewah]

I have never once thought raiders always hit, they shoot so bad most of the time, I am thinking "I'm fked" a lot but the raiders out numbering me 2:1 will just miss every shot and I will hit (though I always start myself off with all my pawns being able to shoot well, like 6+++ in shooting for each one)

And yes, as a poker player, 40 trials is nothing to even think this says something, it is such low volume that until you run 4000 trials (if not 40,000) you have no case. The roulette wheel at the casinos always have the pole with all the previous numbers that people think they can get a "read" on to know what to bet, but that wheel is essentially a random number generator, and that pole is said to be the most profitable addition to any table game in a casino ever made. (yea the roulette thing doesn't seem to have any relevance here but it usually carries the last 20-30 spins, which is no volume and can show a certain number 5-6 times sometimes- and that is basically on a 2.7ish % chance to "hit" a pawn with a shot since there are 38 slots on most wheels)

[bclewis]

Yes, projectiles (whether you rolled a hit or a miss) can be randomly intercepted by things in the path with at least a .2 fill percent, which includes rock chunks.

However, I would expect it to be symmetrical for two shooters targeting each other.

I would expect it to be weighted to benefit whichever person the obstacle was closest to.

In the extreme case, consider a waist-high sandbag, with the protagonist standing directly behind the sandbag, and the enemy at a distance.  In that case the protagonist should hit the sandbag with 0% probability, and the enemy should hit it closer to 50% of the time the bullet would have otherwise hit the protagonist.

[edit, more mathyness:  For horizontal shot I believe the angle between target and obstacle is arctan(h/d), where h is height of gun relative to height of obstacle, and d is horizontal distance from gun to obstacle.

As d approaches zero this goes to 90 degrees (shooting straight down); as d gets large it approaches zero. ]

[edit2: oh!, if you meant that's what you expect to happen the way game is currently coded, you may well be right ]

[Zhentar]

Oh, that's a good point - the first 3 or 4 tiles (I think 4 looking at the code, but I remember there being something odd about it and I'm not testing right now) from the shooter have a 0% chance of intercept, and past that distance doesn't matter. So objects close to a shooter will provide an asymmetric bonus not represented in the shot report hit % displayed.

[pfhorrest]

Well, 2 more permadeath colonies dead from a couple raiders with 1 or 2 shooting skill, awful sniper rifles, and <1% hit chances... instant headshot, body shot kills on colonists with power armor. And all within the first 8-10 shots fired. really makes me not want to play the game anymore. On another save file I used dev tools to give the highest tier sniper rifles to my colonists, each of whom have 11+ Shooting, and my colonists still miss far more than they are "supposed" to... I have some battles that just seem to always result in some colonists dying, and others where I can't lose even when reloading the save file over and over. Wish Tynan or someone familiar with the code could weigh in on this.

[pfhorrest]

One note - I'm using a 50% stat multiplier on "aiming time" for all of my games, could this be messing something up or causing these results?