is success rate formula still valid for high end combines?

[Xapp]

One does not always need large sample sizes to trigger concern, chrisian. This is a classic Bernoulli Trial case and the likelihood of your reality can be easily calculated.

According to basic probability and statistics, a Bernoulli Trial is defined as a true or false event (you either did or did not succeed on the combine) with a consistent probability of success over each trial, and the trials are independent.

For your case we have a sample of 13 combines and a probability of success = .91;

Statistical question: What is the probability that chrisian would fail 6 or more times?

Probability he fails greater than or equal to 6 times:

13
S U M (13 choose k) * (.09)^k * (.91)^(13-k)
k = 6

I've calculated this sum for you, and the probability that you will fail 6 or more times under these conditions is approximately .00052, which is far better than the standard level of statistical significance. You were right to express concern.

Now, this result makes me believe that at least one of the assumptions was wrong:

(1) The probability of a success is .91
(2) The trials were independent
(3) The probability did not change over time.

I can't see their code, so any of the three could be possible.

[miakiss]

I took a quick look at Mousi's skill up charts in the various forums. In most cases the skill up items were the 335 trivials such as MTP, champagne, robes etc.

I don't pretend to have analysed the data, but the success rates look pretty consistent with the old formula - and are certainly closer to 95% than 70%.

However, this data is some weeks old - perhaps they updated the formula as part of the "unplanned" changes in the last patch. It seems a bit odd though - why not change the trivial levels if they want to make things more difficult?

I suppose it could be a bug, but then surely that would only have been introduced if they were changing the code associated with calculating the chance of success.

Maybe I'll do a batch of MTPs and see how it goes (300 skill, no AAs).

[Bobaten]

For my skill ups with brut champagne and gem studded chains I did not have a higher than expected failure rate. It has only been on the 351 trivial augments that my failure rate has been high. 4 more failures on the 16 combines I have done since I last posted.

[Diani]

What did they change after the big tradeskill patch??
Before the patch I had about 80% succesrate on cutting augs as a wizard with skill 300 (and geerlock). Since the patch I have tried 25 cuts and only had 1 succes which is depressing me.

[Luvimt]

Ouch, 1/25 is incredibly bad.
I am 3/5 on radiant cuts after patch (315 skill, no jcm) sorry I dont cut augs often enough to help much with the stats.

Failed another 2 of 3 so its 50% success rate for my incredibly small set of data (8).

[Tsaos]

Better than 95%?

Kyros,

You said with the mastery AA's you can have a 97.5% chance of success on an item... has that been validated anywhere? (If so I have been misadvertising my 300 JC skill being the same as a 300 chanter with JC AA's...) I was under the impression that 5% chance of failure no matter what until the item was 200 points below your raw skill level (regardless of AA's).

Any info would be appreciated.

[Sukrasisx]

One does not always need large sample sizes to trigger concern, chrisian. This is a classic Bernoulli Trial case and the likelihood of your reality can be easily calculated.
...
I've calculated this sum for you, and the probability that you will fail 6 or more times under these conditions is approximately .00052, which is far better than the standard level of statistical significance. You were right to express concern.

Well, maybe. I agree with your calculation. So that means there's about a 1 in 1920 chance of it happening. So if 1920 people did 13 combines each, we'd expect on average that 1 of those people would see 6 or more failures. I have no idea how many people are actually pounding away on this stuff, but it could just be that lots of people are doing combines, and only the people who get really bad results are reporting it.

Of course, that's why other people were urged to submit their results, good or bad. I don't know if a pattern is emerging or not.

The other way to approach the problem is via a controlled experiment. Instead of asking people to report failure rates and hoping that you get a representative sample, pick a set of combines, do them, and record the results. Do that several times. That way you're sure you're getting all the results, good or bad. I'd recommend picking a relatively cheap combine, e.g. Kaladim Constitutionals. You probably need a few hundred combines to get statistically significant numbers.

As a side note, I'll remark that my success rate for combines in the 120-180 range has been pretty much spot on the formula, even after the patch. My skillup rate has sucked bigtime, so much so that I've given up on trying to improve my tradeskills at all until I have some reason to think it's gotten fixed. (That 1 in 1920 figure? Try 1 in a million, or worse, for the likelihood of seeing the results I'm getting, across multiple sessions.)

[Unmei]

I've had some really bad luck -marking- augs on my rogue. He's Poison Making 300, with Trophy and Poison Mastery 1, and the aug marking combines are trivial at 327 - so by rights he should have a 95% success rate, but so far his successes seem closer to 2/3rds or so. Very frustrating.

[Phantron]

Didn't a dev posted at one point that the minimum chance to fail any recipe can be adjusted by the recipe guy as he sees fit? It seems to me that having some recipes with >5% min chance to fail is a pretty good explanation for what's observed here. If certain recipes can never fail, it seems like it's just as possible that certain recipes fail more than others.

[Liwsa]

Went 6-10 cutting augs last night have a 315 skill(300+ eyepiece) with JM3 and salvage 3.

On a good note, the other day went 4-4 adding first level effect with rogue 315 skill(300 + geerlock). Then went 4-4 adding the black inlays. Person I did them for was very happy and gave me a slugworm as payment which I prompted failed with trying to make myself an aug.

[Bobaten]

2 months later, doing around 20-30 augment cuts a week, and I am still holding at around an overall 75% chance on augs. (last 21 attempts 3 of 7, 10 of 10, 2 of 4). Now that I know to expect that success rate, it doesn't bother me at all. But it can be suprising to those who expect 95% success at a modified 315 skill.

[Nolrog]

Went 6-10 cutting augs last night have a 315 skill(300+ eyepiece) with JM3 and salvage 3.

On a good note, the other day went 4-4 adding first level effect with rogue 315 skill(300 + geerlock). Then went 4-4 adding the black inlays. Person I did them for was very happy and gave me a slugworm as payment which I prompted failed with trying to make myself an aug.

I don't mean to laugh at your misfortune, but I got a good laugh out of this.

[Flehmen]

G'day,

Can I suggest an experiment, Sukrasisx? Try to skill-up without a geerlock. My skill-ups are right on the old formula(or better) without the geerlock. Perhaps this is even intended by the devs. The geerlock is about saving money and they didn't want it to, as a side effect, help skillups? <shrug> I just know that on 100 combines I'm much much happier without the geerlock at least with respect to skills... it is costing me a fair bit more though. :/

Take care.

As a side note, I'll remark that my success rate for combines in the 120-180 range has been pretty much spot on the formula, even after the patch. My skillup rate has sucked bigtime, so much so that I've given up on trying to improve my tradeskills at all until I have some reason to think it's gotten fixed. (That 1 in 1920 figure? Try 1 in a million, or worse, for the likelihood of seeing the results I'm getting, across multiple sessions.)

[KyrosKrane]

You said with the mastery AA's you can have a 97.5% chance of success on an item... has that been validated anywhere?

This hasn't been validated; it's merely a best guess of how the system works. I'll try to find out more at the Fan Faire.

[Sukrasisx]

G'day,

Can I suggest an experiment, Sukrasisx? Try to skill-up without a geerlock. My skill-ups are right on the old formula(or better) without the geerlock. Perhaps this is even intended by the devs. The geerlock is about saving money and they didn't want it to, as a side effect, help skillups? <shrug> I just know that on 100 combines I'm much much happier without the geerlock at least with respect to skills... it is costing me a fair bit more though. :/

I saw someone (maybe it was you) suggest that maybe you can't get skillups if the geerlok puts your effective skill over the triv. I know I've gotten such skillups before the recent changes, but I figured, what the heck, it's only more plat, so I actually tried this experiment about a week ago. As I posted in the "Problems skilling up" thread, I did another 100 pottery combines with triv 188, at skill 174, without my geerlok, and got ZERO skillups. This isn't 250+, folks, this is *174*. If it were an isolated case I'd chalk it up to bad luck, but I've had nothing but this kind of skillup rate (well okay, not zero, but pretty darned bad) all the way from skill 122 to my current 174. All of this was post-patch and post-correction-to-patch.

I do see various people citing reasonable skillup rates at low levels in smithing, tailoring, and brewing. Maybe the problem is specific to certain tradeskills? Has anyone seen reasonable skillup rates in pottery in the 120-180 range lately? How about jewelcraft, which is the other that I keep seeing rants about?