TROY

Broken RNG

Recommended Posts

I've railed at the RNG in this game until I'm blue in the face.  Well, I'm shooting for Black.

Once again, I'm gonna say it.  The odds of failing 6 times in a row, at something that you have a 70% chance of success at are so incredibly small that its insignificant.

And yet, it happens in this game, over and over and over.  4, or 5 or 6 failures in a row. With 70 and 80% chances for success.

I've already heard all the same tired ol crap about each new roll being unaffected by the previous, and all the crap about it working properly and I'm telling you its not.

I've never played another game where success or failure with a regular % chance fails so often with such high success chance. Never.

I really really wish somebody would take this seriously, and EXAMINE the RNG in this game, and/or whats actually affecting our chance to do something that isnt reflected in what we see as a success chance. 

I played the ol dice rolling table top D&D games for a lot of years, and an RNG "should" be similar to simply rolling dice.  Ive never experieinced such high rates of failure with such high chances for success as I have in this game, from the very early versions til now.   

 

EDIT:  I want to add something to this to qualify my position, so it doesnt seem like mere crybabying.   In this game, things like cloth, and sewing kits, and matches; they are finite resources. Especially on the harder difficulty settings.  It really really sucks to fail 4 times at starting a fire with an 80% success chance when youre freezing and only have a few matches left. It really really sucks to waste a third of a sewing kit, and nearly a whole bedroll of cloth trying to repair your wool socks. (this was my 6 fails with 70% chance of success.  If all this stuff was unlimited, and it was only time I was losing, I wouldnt care. 

And a second edit.   Not 10 minutes of gameplay later, maybe 20 of realtime, I've just now had FOUR failures at repairing boots, with a stated 85% chance for success. 

Nice. cause leather is wicked easy to get.

 

Edited by TROY
  • Upvote 3

Share this post


Link to post
Share on other sites

I had my thesis on pseudo random generation, and what I can add to this post, which I can validate is that the only way a player can genuinely feel a fair form of randomness is to not actually allowing true randomness to occur.

Having a theoretical 85% chance at success with true random can still infuriate that one unlucky sod who fails four times at the same attempt with the same odds.  That's the harsh reality of true randomness, unfortunately as has been studied and time and time again implemented into endless games - true randomness is a poor approach, at least when it's a core and primary mechanic the user must directly engage with.  If it's used for room texturing on the other hand - that might be perfectly acceptable.

Randomness for primary mechanics should always be firstly pseudo, which it almost always is by default since true randomness is excessively resource expensive, secondly it should be partially predictable, and thirdly controllable, to achieve the prior two points.  Something as simple as bad-luck protection can eliminate all such issues.  And generally some form of good-luck protection is also implemented.

But to address the original thread - yeah, if it feels like it the odds are completely messed up and make no sense, that's because it's an unbiased random approach and won't make any sense, often times by acting biased, even though it absolutely isn't, it doesn't care that this is a game and perception of the player is a valuable consideration, because as far as random operations are concerned, such things don't matter.

Edited by luponius
  • Upvote 5

Share this post


Link to post
Share on other sites
9 minutes ago, luponius said:

the only way a player can genuinely feel a fair form of randomness is to not actually allowing true randomness to occur.

As Apple discovered with random shuffle on iTunes. True randomness quite often resulted in the same song being played several times in succession. They had to revert to pseudo-randomness to keep the users happy.

  • Upvote 3

Share this post


Link to post
Share on other sites
1 minute ago, JAFO said:

As Apple discovered with random shuffle on iTunes. True randomness quite often resulted in the same song being played several times in succession. They had to revert to pseudo-randomness to keep the users happy.

Yup, that's one clear example right there, and you can't blame the users nor apple for the blunder, but the complaints were justified and the solution correct.

  • Upvote 1

Share this post


Link to post
Share on other sites

Thanks, @luponius for that nice comment. Very informative indeed.

I didn't know it was possible for a computer to generate truly random number. I thought the only way to generate "random" numbers was using seeds and calling a function using them.

  • Upvote 1

Share this post


Link to post
Share on other sites
35 minutes ago, slackhideo said:

Thanks, @luponius for that nice comment. Very informative indeed.

I didn't know it was possible for a computer to generate truly random number. I thought the only way to generate "random" numbers was using seeds and calling a function using them.

It's quasi random to be fair, a pattern is always present, but that's the case for everything - there's physics that explains the universe - a complex mathematical pattern of sorts.  Using background system noise, hardware temperature and general sensors in the system can provide a decent form of true random.  For example sampling your CPU, GPU, RAM and Northbridge temps, network utilization and packet loss as offset to generate a value.

Since hardware and network fluctuations have a limited, predictable range (generally) and their pattern is very haphazard, it can be utilized to that effect.  It is however ridiculously expensive (edit: I don't mean financially, but computationally) to do this, and without very good reason, should be avoided.

Edited by luponius
  • Upvote 2

Share this post


Link to post
Share on other sites
41 minutes ago, slackhideo said:

I didn't know it was possible for a computer to generate truly random number. I thought the only way to generate "random" numbers was using seeds and calling a function using them.

Well it isn't depending on how you define "random". What they meant with "pseudo random" was that it wouldn't even trying to be random, like making sure that you can't fail 3 times in a row with 70% chances.

 

3 hours ago, luponius said:

Yup, that's one clear example right there, and you can't blame the users nor apple for the blunder, but the complaints were justified and the solution correct.

The randomness wasn't really the problem, the problem was that the shuffle didn't do what it's mean to do: shuffle the list. If you shuffle the list there's no way one song is played 2 times in a row.

Edited by FINDarkside
  • Upvote 1

Share this post


Link to post
Share on other sites
Just now, FINDarkside said:

Well it isn't depending on how you define "random". What they meant with "pseudo random" was that it wouldn't even trying to be random, like making sure that you can't fail 3 times in a row with 70% chances.

No, pseudo random is that it isn't actually random.  It's a human-perceived random which follows a mathematical pattern, and can be reconstructed with total precision, therefore not really random, using something such as a seed - an input variable that can be thought of as an offset.

Regardless of whether a pseudo (aka a cheap mathematical calculation) is used or a true random (generally a more expensive, hardware dedicated approach), without utilizing some type of bad-luck or good-luck prevention, can have some very bothersome results from a game-play perspective.  Case in poitn would be the example you've given.

Considering that fire starting is core to survival, it's probably something where the entire mechanism could be given a deeper look, not just the percentages and how they're implemented.  On Steam I suggested a sort of interactive system where the player directly participates in the activity to boost the chances or almost completely negate them in starting the fire itself.

  • Upvote 1

Share this post


Link to post
Share on other sites
6 minutes ago, luponius said:

No, pseudo random is that it isn't actually random.  It's a human-perceived random which follows a mathematical pattern, and can be reconstructed with total precision, therefore not really random, using something such as a seed - an input variable that can be thought of as an offset.

Nothing really contradicts what I said. I just said that apple didn't really replace anything with "pseudo random", because it was already pseudo random in the first place.

  • Upvote 1

Share this post


Link to post
Share on other sites
25 minutes ago, luponius said:

It's quasi random to be fair, a pattern is always present, but that's the case for everything - there's physics that explains the universe - a complex mathematical pattern of sorts.  Using background system noise, hardware temperature and general sensors in the system can provide a decent form of true random.  For example sampling your CPU, GPU, RAM and Northbridge temps, network utilization and packet loss as offset to generate a value.

Since hardware and network fluctuations have a limited, predictable range (generally) and their pattern is very haphazard, it can be utilized to that effect.  It is however ridiculously expensive (edit: I don't mean financially, but computationally) to do this, and without very good reason, should be avoided.

It's very interesting. I'm a computer engineer, but I don't know much about this area. It's nice to know what are possible to achieve.
But yeah, I can imagine how expensive is to do such a thing.

  • Upvote 1

Share this post


Link to post
Share on other sites
10 minutes ago, slackhideo said:

It's very interesting. I'm a computer engineer, but I don't know much about this area. It's nice to know what are possible to achieve.
But yeah, I can imagine how expensive is to do such a thing.

I only read about it, didn't actually bother trying it, the access to the required drivers and platform dependence that comes with it can be a pretty notable pain.  Still, it's there for applications that need them, most definitely not your typical game, however.  If you're playing an online D&D session on the other hand... might be worth considering :P

  • Upvote 2

Share this post


Link to post
Share on other sites
1 hour ago, luponius said:

I only read about it, didn't actually bother trying it, the access to the required drivers and platform dependence that comes with it can be a pretty notable pain.  Still, it's there for applications that need them, most definitely not your typical game, however.  If you're playing an online D&D session on the other hand... might be worth considering :P

It's good to have options. In this case (The Long Dark game), I completely agree with you on the user experience over trying to appear as random as possible. I guess most people hate (me included) to waste 5 matches to get a fire going.

  • Upvote 2

Share this post


Link to post
Share on other sites

@luponius  First off, Please accept my genuine thanks for your input here. It's the first time this has been explained in a way that makes sense.

I have always accepted that in a true random system, a failure is technically possible hundreds of times in a row.  But theres a huge difference in something being possible, and being likely. My primary point is that, while I accept the possibility, I refuse to accept the commonality of it. If it happened once in a while, no complaints. bad luck.  It happens to me all the time in this particular game. The odds of it happening are slim. The odds against it happening repeatedly are astronomical. 

I do really appreciate the light you've shone on this though, and maybe, just maybe, somebody (I'm looking at YOU Raph), might build in a tiny amount of that bad luck protection. Its far easier to accept a couple failures, even if its routine. 

  • Upvote 1

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now