UFOextender accuracy issues.

[WhiskeyBob]

With this option enabled, it now shows hit %, but accuracy works in some "magic" way.
1. I've noticed that for hit ratios below 50%, a real probability to hit a target is far less than 50%.
2. But if my soldier standing right before an alien, hit ratio is about 100%, even if indicator states 33%. 
3. Closer to the end of game, aimed shots shown with 100+% hit ratios. I've tried to reload game tens of times, and it's really hard to miss with 110% hit probability.
Should this be considered as game defect, or such a change must be released as game patch?
I do have some shooting experience, and understand that such effects as derivation, or "bullet likes the sun" won't work in XCOM, but I could modestly advise to change shooting accuracy coefficients.

[volutar]

1. I've noticed that for hit ratios below 50%, a real probability to hit a target is far less than 50%.

It can't be truth. At least, for the nightly.

2. But if my soldier standing right before an alien, hit ratio is about 100%, even if indicator states 33%. 

Hit ratio shows chance to hit particular "central" voxel, not unit itself. In geometrical, angular terms, actual chance to hit is extremely hard to estimate. You can miss slightly, but still hit unit even on medium distance, but short distance you will still hit some other voxe of the target even you miss the target voxel. It's obvious.

3. Closer to the end of game, aimed shots shown with 100+% hit ratios. I've tried to reload game tens of times, and it's really hard to miss with 110% hit probability.

110% means "no miss". It's the max

[mrxian]

Here's what I observed.

The chance to hit is an estimate on the low side. Your actual chance to hit is larger.
The game (I assume) first rolls to hit the target.
If it hits, it hits.
If it misses, the game then determines where the miss goes. The game basically creates a cone in front of your unit and determines where in the cone the shot goes. The larger your original hit percentage, the narrower the cone is.
The shot can veer off course up and down and left and right.
The shot then flies from the gun until it hits something or leaves the map. It can hit it's original target or something else in it's path.

So if you try to shoot at something right next to you, that unit probably is in the 'miss cone' anyway, so you'll hit anyway.

It is also a bad idea to take low-chance shots if your own units are in the vicinity.

Image to explain more.

Note: This is based on my own observations of the game, so I may be completely wrong.

[Warboy1982]

technically correct, although a "hit" doesn't mean a hit, it means a much narrower cone. at extreme range, the cone may be wider than the unit you are trying to shoot, meaning that there is still a chance to miss.
in short, you can hit with 0% accuracy and you can miss with 110%.

[WhiskeyBob]

It can't be truth. At least, for the nightly.

Well, this is my own observations based on hundreds of reloads. Sure I can be wrong, since stats science can't even rely on millons sampling. 

and you can miss with 110%.

That's right, sometimes it happens.

[Hobbes]

I've made a few additions to your image to show better how stray shots work (which is the issue being discussed).

[WhiskeyBob]

Hit ratio shows chance to hit particular "central" voxel, not unit itself.

Last time I did uninstall C++ 10+ years ago, but I presume OpenXCOM uses pixel model, not voxel? Or volume measurements was implemented especially for OpenXCOM? UFO 1993 is too old for voxel algorythms. Maybe
If hit ratio is shown for a central voxel, then it must be harder to hit a whole body. But what I see now, it's working vice versa. 

[volutar]

Last time I did uninstall C++ 10+ years ago, but I presume OpenXCOM uses pixel model, not voxel? Or volume measurements was implemented especially for OpenXCOM? UFO 1993 is too old for voxel algorythms. Maybe

Key word - "Maybe".
Since Rebelstar games Gollops were using hit bit map, as alternative for visual map, with 1 as solid and 0 as space, to trace fire, it's quite obvious even in game (in RebelStar Raiders there was no visual/hit difference, since graphics was too primitive). With Laser Squad 2 (transformed into UFO: Enemy unknown) they made another step forward - into 3D, by making this hit map multileveled.

If hit ratio is shown for a central voxel, then it must be harder to hit a whole body. But what I see now, it's working vice versa. 

Hit ratio shows chances of target point being in the particular target voxel, which it choses by scanning the target model from the center to periphery, and picking the first visible. It might be close to obstacle, and thus you can miss even with 110%, just because it might deviate 1 voxel side, which is not enough to MISS the target, but enough to hit this obstacle.

Accuracy clouds (from above) for 0/25/50/75%. If you'll look closer, you notice feature of north-south/west-east uniformity. North-south direction cloud is wider (but it doesn't affect overall chance too much, since central "hit ray" stays the same).
So any trial will be in the inner "ray" with defined chance (but mapped for 110% being max), and all others are in outer cloud. Chances are not CLEAR, for many reasons, among them - it can be the any point in the line of target, but not inside of target (before of after the target, slightly off, but still inside of target LOFT, etc).



Target voxel picking algorithm is not vanilla in OpenXcom, it tries more voxels, so in some respect it's more advanced. But it suffers from the "close to obstacle" syndrome, which I planned to overcome since quite long ago. Instead of scanning for the first to central, better to pick the most "central" from the visible. For example, if you see only the head of enemy, algorithm will aim the neck voxel, which is closer to the center of the unit. Obviously it's better to pick center of the head, to get less chances to hit the obstacle, but hit the target. But it will require for maths and kind of filters to choose the center. I didn't have time for this, and frankly, currently it works quite OK, so it doesn't seem to be the task of any importance.

[kikimoristan]

how do you go fps look in openxcom?

[Warboy1982]

f10 to take a soldier's-eye view screenshot of the voxelized landscape

[Arthanor]

Now the real question is: Why do missed shots always go straight to the back/head of other operatives? ;P

The game almost seems to be thinking: "Well, he missed by a lot, what's pretty far that could be hit for maximum effect? Mouahahahaha!" *swear words ensue*

[kikimoristan]

The Avenger.

Man i wish i could play whole game like this. i don't even need textures just a sky hahaha. FPS openxcom would be funny.