XCOM EU Weapon Accuracy Question

[elric132]

Hello.

âWhen you have a target selected and a hit percentage is displayed, is this # completely accurate or is it a rough estimate?

âAre there any factors NOT taken into account in this #? (In other words could it be wrong because some information is not being used/included in the calculation displayed to the player)

âI am tempted to start tracking the actual # of hits versus the hit percentage #s displayed. I understand this # is a probability so variation and exceptional streaks can occur. But there have just been too many occasions where my actual hit #s are far below the % displayed and very few such streaks that favor me. Some of the negative streaks have been long enough to make me question what is going on. Again, I understand snake-eyes will come up on occasion (so to speak), but what I am consistently experiencing is just hard to believe.

âThanks for your help understanding what I am seeing.

[JDDysart]

As a college professor once explained it to me, the so-called Random Number Generator is actually a "PSEUDO" Random Number Generator; meaning that with a large enough sampling It behaves like an RNG, but there actually is a pattern to it if one pays enough attention for a long enough time.

 

Then there are the additional factors, not only just the chance to hit, but any bonuses (or penalties) that may affect that 'chance to hit' on one side, as well as the other side's 'chance to be missed' along with any bonuses (or penalties) thereto. Generally the 'chance to hit' is labeled "the offensive" side, while the 'chance to be missed' is labeled "the defensive" side.

 

Should you be familiar with the concept of "Save Scumming" - generally if you take the same actions in the same sequence you tend to get the same results; to get different results you need to figure out how to mix up the sequence (and sometimes change out the offensive unit, the defensive unit or both) in order to get different results. XCOM can be frustrating in that when you use four offensive units in an attempt to take out four defensive units, and when done in one particular sequence the offensive side fails ALL four attempts; but, in a different particular sequence, all four "otherwise exactly the same" attempts may ALL be successful.

 

"Under the Hood" of Random Number Generation:

• The RNG uses a 'seed' value that is fed into a complex numerical operation that produces a "pseudo random number" value, returning said value (that is then used as a 'seed' for the next "pseudo random number").
  
• XCOM effectively generates a 'stack' containing several of these "pseudo random numbers" at certain intervals.
  
• Then retrieves a number from this 'stack' each time the game encounters a need for a random number.
  
• At the end of said interval, any remaining numbers in this stack are discarded, and another 'stack' is generated.
  
• And please do note: chances to 'hit' or 'miss' are not the only equations that use random numbers (from said 'stack')!

When the above mentioned "interval" is too short or said 'stack' is too small, the results will appear to be a 'lucky streak' (good, bad, or otherwise).

[elric132]

Hi, thanks for your response.

​I know about RNGs, my bachelor's degree is in cs and I spent ~20 years working in IT including ~10 as a programmer.

​When you target an opponent a box appears with a ? icon at the bottom. If you click on it, it displays modifiers to your chance to hit (offensive and defensive). I think it is all of them ASSUMING you have dissected your opponent. This may explain some of the inconsistencies I noticed. If true, it also implies you cannot get an accurate number for opposition that cannot be dissected (which I think is true for some of the robotic opponents, someone can fact check me on this).

​In reference to "save scumming", it isn't necessary to the degree you suggest assuming you turn on the option that generates a new (quasi) random result each time you repeat the action called "save scumming" :-) under the 2nd wave settings.

​Since my initial post I tried keeping a simple log (I broke the hit %s into 10% levels and tracked hits and misses) over several missions to see if the displayed percentages were in the ball park of what they should be. They seem to be. I was a little frustrated with this result because it failed to explain what I at least thought I was seeing as I played. But something else occurred to me. Maybe what I was seeing were peculiar patterns versus an overall inconsistency.

​As a (simplifide) example, if I flip a coin a 1000 times and the results come out roughly 50-50 you might assume the behavior is what was expected. This is similar to my first test. But what if I told you among those 1000 flips on 10 occasions I got 50 consecutive heads and 50 consecutive tail streaks. This might be what I am seeing. I am going to again use some simplified tracking to see if this is the case.

​In regards to your "under the hood" section, I understand and agree with all of it. But let me suggest that if the game consistently produces "streaks" versus a more even distribution the majority of the time then it is fair to say there is a problem.