Death to the RNG!

[Itek]

*sigh*

roll random 100 and get less than 1

or

roll random 100 and get less than 2

or

roll random 100 and get less than 3

-repeat-

The odds become FANTASTICLY high that you will roll randomly under an increasing "cap." (The math is annoying, call up a college math/statistics professor and ask them to show you the proof, if you ever have trouble falling asleep.)

And just setting the cap REDICULOUSLY HIGH to avoid this amounts to not having a cap at all.

Dunno if they have "Compass Bank" where you are, but here they run ads about their "free checking and free ATMs, and free money when you open an account" which explain how they can afford to do that...

"trust us, we're BANKERS.... we're seriously good at math."

-edit-

It is left as an exercise for the user what the odds of failing to get a skill up merely 50 times in a row are, given a counter of 350 (the max proposed in this thread so far).

[KyrosKrane]

Itek, it has been my experience that outside a casino, statistics have very little bearing on reality. =) People study statistics in an attempt to get a grasp on abstract or massive concepts they can't quite comprehend on an individual scale, but it is a gross mistake to accept the statistics as a substitute for that reality.

You're arguing that statistically, a cap would be a bad thing. However, I honestly don't see the difference between the current situation and a capped one for anyone except the few people who currently get dry runs big enough that they would hit the hypothetical cap. Unlike the mouse scenario, we're not saying we should take away the random element. We're just saying that working to death is a bad thing. Again unlike the mouse scenario, no one "gets a penny each time the mouse hits the bar" -- the only thing affected is the mean number of combines to get a skillup, since we're removing a few outliers at the extreme high end.

You seem to have boned up on statistics. Consider this. What happens to the mean if you remove all data points more than three standard deviations above the current mean? The mean goes down a bit, sure, but is the drop substantial? I'd argue no. The data points above three standard deviations comprise about 0.5% of the total sample size (or population size, depending on context). Even if we set the cap at two standard deviations, we'd only be excluding about 2.5% of the overall sample.

(Edit: Assuming a normal distribution, of course, which may not be the case with tradeskills.)

[Maevenniia]

Itek, as I envision it, the cap of 350 (whatever) would not have any efect on the skillup forumula itself.

skillups would still be calculated as usual.

take baking as an example...

then, if bakingskillup = no & non trivial combine = yes, bakingcounter = bakingcounter + 1. if bakingcounter = 350, then force skillup (ie: bakingskill = bakingskill +1) and set bakingcounter = 0.

if bakingskillup = yes, then set bakingcounter = 0.



Not sure how thats exploitable? All it does is allow the formula to continue working as usual, BUT, it institutes a "take pity on the poor slob being hated on by the RNG" mechanic.

[gggrant]

Itek, as I envision it, the cap of 350 (whatever) would not have any efect on the skillup forumula itself.

skillups would still be calculated as usual.

take baking as an example...

then, if bakingskillup = no & non trivial combine = yes, bakingcounter = bakingcounter + 1. if bakingcounter = 350, then force skillup (ie: bakingskill = bakingskill +1) and set bakingcounter = 0.

if bakingskillup = yes, then set bakingcounter = 0.



Not sure how thats exploitable? All it does is allow the formula to continue working as usual, BUT, it institutes a "take pity on the poor slob being hated on by the RNG" mechanic.

Actually the skillup formula would have to change.

Lets say the chance to skill up is 2% for an average of 2 skillups per 100 combines. Now we add a take pity counter that fires at 100. The chance to skill up percent would have to be lowered (to 1.75% as a wild guess) to give an average of 2 skillups per 100 combines or there would have to be a check that would prevent a three skillups per 100 combines.

[KyrosKrane]

That would only apply, though, if the cap were set so low that it reduces the average number of combines required to get a skillup. In my post above, I suggested a setting of ten times the average number, so if on average you'd only get one skillup per 50 combines, then the "take pity" counter would only kick in at 500 combines with no skillup. (Obviously, the 10x multiplier would have to be tweaked a bit, but the principle is there.)

[olhoss]

Which was my point to begin with...this idea wasn't to raise the skillup rate, it was to prevent the long skillup drouts that people sometimes encounter, particularly on the high end.

Lets say the skill up chance is 2% at 295. So the average is that it should take you about 250 combines to get to 300. Now give or take a few this is no big deal, but how about when it takes you 730? Chalking that up to bad luck blows, when you consider what it took to get the materials for those extra 480 combines.

Even with a very liberal counter cap like I was talking, it is saying the max number of combines to finish would be 500, and you wouldn't go more than 100 without a skillup. This does not in any way raise the skillup rate unless the unlucky streak is broken by the counter, then the other skillups come quickly, which is still not going to happen much.

[Gaell Stormracer]

Itek: you've misunderstood the proposal. The "pity counter" isn't a number to be rolled against; but if a player has had a certain number, when it hits a certain value it sends a skill-up. Either way, if you skill up, pity counter re-zeroes.

Pseudocode:

- do test for skill-up with current formula.
-- if skill-up check succeeds, increase skill, set pity counter to zero, and exit
- Increment pity counter
- Has pity counter hit threshold?
-- If so, increase skill, set pity counter to zero, and exit.

No exploit possible.

(From a coding perspective though, a countdown would be better, since it can be initialized to a value based on the skill level).

[Tanker]

Hi guys. You know how much I LOVE RNG threads.

Ok, first, to anyone who thinks that setting this "pity" counter high enough wouldn't add "free" skill-ups is kidding themselves. Look, either it's set so high that it never happens, or there WILL be a time when you get a skillup that you didn't "deserve" based on randomness. When this happens, since it in NO way affects how many attempts will be needed for the next skillup (say, making it MORE), then you will have "saved" every attempt you would have made before getting that skillup you just got "for free".

So, if it "would" have taken you 106 attempts, and you got it "free" at 100, then you just saved 6 attempts. If it would have taken you 512, and you got it free at 500, you saved 12. That's the only way to look at this.

Now, is that game-breaking? Not my call. It would seem that there would be a constant battle with some wanting the "countdown" to be lowered, and the designers not wanting to give too many "free" skillups. Look at it this way, would you want to spend time coding/designing something that you don't want to happen very often?!

Now, normally, I'd have a bunch of math up here, showing you the average reduction in attempt counts based on different formulae for the "pity counter", but I don't really need to. If you look at the graphs I gave you all earlier (or redo them, they weren't hard), you'll see that we weren't talking more than a few thousand combines on average to reach max skill in a tradeskill. The variation on the TOTAL # of combines is not a lot. It would frankly stun me to hear of more than a few of you who took more than 20% more attempts than the average. (now isn't that just asking for it? )

Balancing a pity counter could be simple, though. Simply don't allow a skillup within like 5 attempts of the previous one. Betcha wouldn't like that much though, would ya? (yes, 5 is arbitrary, it probably wouldn't take more than 1 or 2 to balance a sufficiently high "pity counter").

[olhoss]

Ok, first, to anyone who thinks that setting this "pity" counter high enough wouldn't add "free" skill-ups is kidding themselves.

It would. But that isn't the point. It would add them where it would reduce frustration, and it would not add enough to make any significant difference in the game.

It would frankly stun me to hear of more than a few of you who took more than 20% more attempts than the average. (now isn't that just asking for it? )

Yes it is. I kept track on all of my skills (after JC which I did before I knew about the calculator) from 280 to 300, where the reagents for the combine get rare/expensive. I got significantly hosed on Tailoring and Fletching.....more that 200% of the expected number of combines. I did better than expected on Research, and did worse than expected on Smithing, Brewing, Baking and Pottery, though not significantly. And why don't you here from people who get really lucky? Or don't get hosed too badly? Because they aren't suffering the terrible luck that causes enough frustration to propose coding changes, that's why. My suggestion was a way that with a simple formula (I can't say simple coding change because I don't want to presume that it is) these bad bouts of luck can be eliminated without throwing the game out of whack.

[Ngreth Thergn]

NOTE!!!!! this is just for conversation sake and in no way anything planned, or even really being considered at this time.

But what did people think of the "reverse" part of the pity counter Tanker mentioned.

Lets say we did this with the pity counter.

The pity counter = max(2*(average combines per skill-up at skill), 100)
"anti-pity" counter = 0.5 * (average combines per skill-up at skill)

Anti- pity would be the MINIMUM amount of combines needed to gain a skill after you just gained one.

If you want to remove some of the randomness, it needs to go BOTH ways.

Folks would no longer have horrid runs of bad luck, but there would also no longer be great runs of good luck as it would be disallowed.

NOTE!!!!! this is just for conversation sake and in no way anything planned, or even really being considered at this time.
The numbers could widely change from that. maybe we want pity to be much higher and let anti-pity be much lower... This is in no way currently being planned or considered any more than idle conversation.

[KyrosKrane]

Tanker! Great to see you on here (again). =)

I totally agree that the variation on the total number of combines isn't that huge. The discussion is more about the single-skill-point outliers that are significantly far away from the predicted average. The objective isn't to make skilling up easier overall. The intended objective is to mitigate only the worst-case disasters, the kind that get players so frustrated they quit tradeskills entirely.

I'd have to dive back into my stats software and graph it, but at a guess, I'd say we're talking about less than 1% of an entire skilling-up run, from 0-300 (that would be less than three points where you get horrible dry runs).

If we were to go with the 10x figure, then the change would be:

pity counter = max(10*(average combines per skill-up at skill), 400)
"anti-pity" counter = 0.1 * (average combines per skill-up at skill)

The 400 there is a figure I pulled out of my hat -- it takes about 40 attempts on average to get a skillup at 299 skill, assuming your prime stat is high enough to guaranteed a Pass 1 success.

In fact, I would implement the pity counter a little differently. I wouldn't make it an absolute number of combines you had to do until you could get a skillup again. Instead, I would do something like this. If you got a "pity" skillup, then for the next skillup you would get, roll a number and compare to the anti-pity stat. If the roll fails, void the skillup. If it passes, allow the skillup. The effect is that you got one freebie skillup after a huge dry run, so now you have a chance at negating the next skillup you would get.

That way, multiple skillups in a row would still be possible, even after a pity skillup -- but the overall number of combines to reach 300 would still be maintained.

(I've got a nagging feeling in the back of my mind that this somehow doesn't actually balance out, that it needs a little tweaking ... but I can't quite put my finger on it.)

[Xodar]

I recently did pottery from 288 to 294. Max stats and AA for a 75 ranger.

The number of combines per skill up were 2, 6, 55, 119, 21, 34.

237 combine for 6 skill ups, or 39.5 combines per skill up.

That has to be close to the baseline average. I would not trade the 119 for the 2,6. As mentioned before part of the appeal of tradeskills is the random aspect, as frustrating as it might be. 2 skill ups in 8 combines is so sweet because of the 119 combine point. I understand the consideration about terrible runs, but I have to assume in practically all cases it evens out across 300 skill points.

[Gaell Stormracer]

but I have to assume in practically all cases it evens out across 300 skill points.

True randomness has no guarantee of "evening out" in such a way...

[Tanker]

Olhoss (sorry, not going to quote much), I certainly see that people can have bands of skillups that are more than 20% outside, my point was that across all 300 points you won't be that far off the expected total # of attempts. I can see that you're trying to alleviate specific skillup instances, so my argument is almost apples and oranges, but it does set a context.

Ngreth and Kyros (HI!), it doesn't take an anti-pity counter of inverse proportion to "balance" the pity counter. Keep in mind that the pity counter will only do something pretty rarely, yet the anti-pity counter I/we suggested would ALWAYS affect the skillup rate. That's why I was thinking a pity counter of 100 versus an anti-pity counter of 1 or 2 would probably even out. i.e. you'd maybe fire the pity counter off 5 to 10 times during the 0-300 run, saving a few to a dozen attempts each time, but the anti-pity would add 1-2 attempts to EACH skillup (*300)... see?

As a side note (and it's been a while since I've been involved), I don't see this happening soon. For one thing, as hard as it is to get you guys all on the same page as far as the statistics goes, it's just that hard to explain this to the bosses over there. Also, it's one of those things that's bound to stir people up, whereas leaving it the same probably won't. Finally, it's a feature that the EQ team wouldn't want to work very often, so why spend the time?

[Xodar]

True randomness has no guarantee of "evening out" in such a way...

And what does true randomness have to do with the EQ RNG hehe?

But across 3,000+ combines, and in my firsthand experience, the runs of 153 are balanced by the single digit combine points, we just find the former more memorable.

As Tanker said, while not impossible, it is pretty unlikely someone would vary from the expected by a large percentage over that many points.