Hit Rate

[GoldenAvariel]

The RNG has always been my arch-nemesis, but the number of misses I am experiencing are extremely frustrating. I've gotten where I overslot accuracy on all of my attacks on every characters to try to minimize that problem.

 

And while I knew how it felt, and my group jokes that using Buildup or Aim was a guaranteed miss on snipes and assassinations, I had no stats to back me up so I had not come here to ask if anyone else had that same feeling.

 

Finally, last night I did a copy+Paste from the combat log in-game into Notepad on my desktop. It took a while because Copy was only getting the log I could see and not all of the text I highlighted, but I finally got it done. Only the last 12 minutes were available to be viewed so that's all I have to go on. In that time period I performed 132 attacks and all had a 95% chance to-hit vs Freakshow ranging from white minions to red bosses. In that time period I missed 12 times (3 of them vs the boss within a 10 second window which nearly got me killed since I was playing a blaster) - certainly not as bad as I thought but still around a 9% miss rate instead of 5%. Of course 12 of the 120 hits were streakbreaker hits that didn't roll at all but I went ahead and counted them as rolled to-hit instead of going with a 10% miss rate on attempts that actually used the RNG.

 

I'd like to do further analysis over longer periods of time, but I cannot figure out how to get the combat log to write to disk. I've tried all the options I could find but haven't found that they work, or perhaps I'm looking in the wrong place. Is there a way to do that?

Edited December 22, 2021 by GoldenAvariel

[Troo]

High accuracy builds and attacks can feel like they miss significantly more than 5% of the time.

 

I believe my worst streak was 7 out of 10 assassin strikes missing.

It was hypothesized that accuracy somehow looped, and if you had too much it might behave like you had very little.

 

 

Edited December 22, 2021 by Troo

[Luminara]

There's a combat log parser somewhere around here.  I've seen references to it, but haven't pursued it.

 

For posterity, mixing single-target powers with AoE/PBAoE powers throws a monkey wrench into the works.  The streak breaker prevents misses for single-target powers and for sequential checks in AoE/PBAoE powers (meaning, the next critter in the target list for the power), but it does absolutely nothing to prevent a miss when mixing single-target with AoE (meaning, using a single-target power, followed by an AoE/PBAoE, or vice versa).  You can, in fact, miss an infinite number of times by staggering the vectors.  I frequently encounter miss streaks of this nature, typically three but sometimes extending to five or six, because I rely heavily on both vectors on all of my characters, and I suspect that's the norm for players, since so few sets are perfectly suited for either.

 

Frequent tabbing between targets can also create the illusion of miss streaks.  You'll use a power on target A, immediately Tab to target B and queue a power, miss target A and tab back to that target while the power on target B is animating, queue power on target A again, tab to target C, miss on target A and immediately tab back, et cetera.  Because the streak breaker is only countering your misses, not your misses on specific targets, you can continue to miss on target A until your head explodes from the irritation as you hammer on the keyboard and scream, "DIE DIE DIE FUCKING DIE!".  The streak breaker is working, but it's applying the break on target B, then target C, and so on.

 

These two flaws in the hit check mechanic are what lead to posts of this nature.  You can have +1000% Accuracy and +1000% ToHit and still miss several times in a row when using single-target and AoE/PBAoE consecutively, and you can also miss a specific target several times when you're changing targets with each power.

 

This information dump brought to you by the letter 4, the color & and sponsors like you.

[Troo]

2 minutes ago, Luminara said:

until your head explodes from the irritation as you hammer on the keyboard and scream, "DIE DIE DIE FUCKING DIE!". 

 

breathe, you're in a safe place.

[Doomguide2005]

I think @Carnifaxhas a parser that has a link posted to it somewhere.

[Carnifax]

https://www.carnifax.org/

 

Seems to be still alive (haven't checked in a while).

[Albion]

8 hours ago, Troo said:

 

breathe, you're in a safe place.

There is no such thing, Batallion will hunt us all down.

[aethereal]

It is my understanding that the code simply calls the operating system random number generator, it doesn't do anything screwy like implement its own random number function.  It beggars the imagination to suggest that the random number generators in any major OS are so badly flawed that they would do something like "produce a noticeably skewed distribution of percentile values towards higher numbers."  Like...  that would be a HUGE problem that would affect many, many, very seriously more important pieces of software than an old MMO.

 

Once you've gotten a random value from the operating system, all you do with it is basically multiply it and/or mod it so that it falls into the range of values that you want.  Managing to bias the results there would be an impressively bad bug.

 

It seems overwhelmingly likely to me on the basis of this that anyone who is finding bias towards missing in either their experience of the game, or in parsing combat logs, is detecting either:

1.  Flaws in human cognition that make them create false patterns from random things.

 

2.  Flaws in exactly what is logged and how, or what how the logs are parsed.

 

rather than

 

3.  Flaws in the generation of random numbers between 0 and 100.

[roleki]

I don't know the proper terminology for the phenomenon, but whatever it is called, I suspect people who have invested resources towards Accuracy/ToHit are more likely to notice every MISS! message and will attach undue significance to each one.  If I'm rolling IOs and miss twice in a 4-power attack chain, I'll think, damn, that's unlucky.  If I spent 2M on a Kismet +Acc and miss 2 times in the span of four attacks, I'll kill you and everyone who ever met you.

[Lunchmoney]

21 minutes ago, roleki said:

I don't know the proper terminology for the phenomenon

Perceptual bias I think is the term you are looking for.

It certainly is the problem when ever a thread like this starts up. To properly check the rng you need thousands, if not millions, of results. Not a 100 or so. And it's been done, it's been checked, and it's fine.

 

Edited December 23, 2021 by Lunchmoney

[Peacemoon]

We really need to a way to reach 100% accuracy I think. I don’t think the 95% rule really serves any purpose other than to frustrate players.

 

I would be quite happy even if 95% - 100% was subject to diminishing returns so that in order to get to reach 100% you’d need a lot more accuracy to get there.

 

Anything to stop me missing my drive by attack on a level 1 Hellion as a 50+1 full incarnated/IO’s super mega hero!!!

 

[Carnifax]

The only weirdness I've seen with ToHit rolls is Burn on my tank, which throws lots really weird low-chance misses while still seemingly working as intended.

 

Posted it in both the Tank forum and the Discord and none of us came up with a valid reason as to what was going on.

 

Burn still works brilliantly on him mind, it was just strange.

 

But the parser was the thing which alerted me to that (I thought it was a coding bug my me initially but going through the raw logfile it seemed like it was parsing correctly). 

 

All other powers were 90-95% hit rate like I'd expect. 

[Black Zot]

Never trust any percentage chance any game gives you that's not 100% or 0%.  It's ALWAYS lying, one way or the other.

[srmalloy]

8 hours ago, aethereal said:

It beggars the imagination to suggest that the random number generators in any major OS are so badly flawed that they would do something like "produce a noticeably skewed distribution of percentile values towards higher numbers."

It depends on how the RNG works and how a program uses it to produce random numbers. For example, many system-level RNGs produce a pseudo-random 32-bit integer. One of the more common shortcuts used by coders unfamiliar with how RNGs work to get a limited random number range is to take the generated integer and MOD it to the range they want. This can inherently create bias. For example, if you have a RNG that creates a four-bit random integer from 0-15, and you use it to get a 0-9 number using MOD, you get the following results:

 

RNG	Result	RNG	Result
0	0	8	8
1	1	9	9
2	2	10	0
3	3	11	1
4	4	12	2
5	5	13	3
6	6	14	4
7	7	15	5

Note that 0-5 appears twice each in the result, while 6-9 only appear once each. Now, this is a very simplistic example, and one that would be guarded against by any decent RNG library, but it shows how not paying attention to the generation and subsequent calculation can distort the return.

[Bill Z Bubba]

5 hours ago, Peacemoon said:

We really need to a way to reach 100% accuracy I think. I don’t think the 95% rule really serves any purpose other than to frustrate players.

 

I would be quite happy even if 95% - 100% was subject to diminishing returns so that in order to get to reach 100% you’d need a lot more accuracy to get there.

 

Anything to stop me missing my drive by attack on a level 1 Hellion as a 50+1 full incarnated/IO’s super mega hero!!!

 

As an old pen and paper role player, I'll always be against this. A one in twenty chance to fumble has always seemed appropriate to me. Then again, I do love chaos.

[aethereal]

48 minutes ago, srmalloy said:

It depends on how the RNG works and how a program uses it to produce random numbers. For example, many system-level RNGs produce a pseudo-random 32-bit integer. One of the more common shortcuts used by coders unfamiliar with how RNGs work to get a limited random number range is to take the generated integer and MOD it to the range they want. This can inherently create bias. For example, if you have a RNG that creates a four-bit random integer from 0-15, and you use it to get a 0-9 number using MOD, you get the following results:

 

RNG	Result	RNG	Result
0	0	8	8
1	1	9	9
2	2	10	0
3	3	11	1
4	4	12	2
5	5	13	3
6	6	14	4
7	7	15	5

Note that 0-5 appears twice each in the result, while 6-9 only appear once each. Now, this is a very simplistic example, and one that would be guarded against by any decent RNG library, but it shows how not paying attention to the generation and subsequent calculation can distort the return.

It can't create bias on the scale that people are looking for.  It only create a large bias because you're working with unrealistically small numbers.  If you take an actual 32 bit integer and mod by 10, it will, yes, be biased -- we'll be missing one 7, 8, and 9 out of about 4.3 billion.

[TheZag]

Every time the game needs a random number the devs call me up and i pull one out of a hat and tell them.  I cut 95 through 100 from a bigger piece of paper so they get pulled more often.

 

But RNG can be broken more easily then you think.  Final Fantasy 10 on PS2 has 32 million random seeds that determine how much damage each weapon swing does and what monsters attack and drops and all things random (obviously).  The HD re-release of FFX only has 256 random seeds.  The game still feels just as random as the original but you only get 1 of 256 pre determined games when you fire up a new file.

 

I doubt that is the case with city of heroes RNG and there has been plenty of testing so we can be reasonably confident that our 95% accuracy is as close to an actual random 95% as possible.

[FallenHero]

[DoctorDitko]

Someone has to do it:

 

 

Real randomness is clumpy!

 

 

Edited December 23, 2021 by DoctorDitko

[Luminara]

5 minutes ago, DoctorDitko said:

Real randomness is clumpy!

 

So is poop.

 

Usually.

[merrypessimist]

2 hours ago, Luminara said:

 

So is poop.

 

Usually.

 

That's a pretty crappy analogy.  <.< >.>  

 

Related to the actual topic and not just a terrible feces based pun:  Always remember that RNGesus hates you if you don't maintain a shrine to the holy percentile die.

 

Luminara is also awesome and not in any way crappy.

[Ironblade]

50 minutes ago, merrypessimist said:

Related to the actual topic and not just a terrible feces based pun:  Always remember that RNGesus hates you if you don't maintain a shrine to the holy percentile die.

A friend of mine had a d20 that repeatedly rolled badly so he smashed it with a hammer.

Then he put the fragments in the pouch with the other dice - as a warning to them.

[Peacemoon]

18 hours ago, Bill Z Bubba said:

 

As an old pen and paper role player, I'll always be against this. A one in twenty chance to fumble has always seemed appropriate to me. Then again, I do love chaos.

I’m not against having a chance to miss but CoH has always had quite an arbitrary system of it compared to any other RPPG I’ve played, with its 95% cap on accuracy across the board. 
 

Chance to miss is usually something you deal with against higher level enemies, not something that has to always be a factor against even level mobs, or -2 mobs, or heck even -49 level mobs. At some point it’s okay to say “alright, this is a guaranteed hit”.
 

Instead of fixing this core design flaw (in my opinion at least), we have layers of silliness added with ‘streak breakers’ and inherent accuracy boosts for low level characters to countenance what can be quite a frustrating play experience. 

 

I mean you can justify it by saying if you roll a 1 on a D20 you should always fumble and miss, but I’m not a fan of that concept and CoH isn’t a slow dice rolling type of game. 

I remember in WoW having to cap my ‘to hit’ and ‘expertise’ stats in order to reduce miss chance and and parry/dodge chance against enemies, and they removed it for good reason because it became tedious, and people didn’t know how much they needed vs  not having enough/having too much and trying to find that sweet spot without wasting precious stats. 
 

I’m not saying remove accuracy of course, but that’s my counter thoughts. I just don’t understand why always having a guaranteed chance to miss is so important.

 

 

[Sovera]

I'm a long time advocate of getting rid of the 5% miss chance. We can build to 100% and avoid it, or we decide it's not important and don't do it.

 

Many expansions back WoW had this and the Hit stat was the first thing any guide advised to cap in order to never miss. That is all I ask.

 

5% misses are not fun, they add nothing to the game, it does not make the game harder or more interesting. I'm not even asking for mobs to have this applied to them so that 50% defense would mean we never get hit. I'm fine with always having a 5% chance the mobs will hit.

 

 

-I- don't want to have the 5% infuriatingly there from level 1 to level 50 incarnate where I drop a meteor on a level 1 hellion and it misses. It can get so bad as to have a 30 second difference on a pylon test which is ridunkulous.

[Bill Z Bubba]

32 minutes ago, Sovera said:

it does not make the game harder or more interesting.

 

Funny that this is exactly what I said about the Cimerorans getting fixed.