Proving 'Success have a greater chance of giving a skillup'

[kiztent]

Assuming that all of this end up proving that skill ups are more likely on successes than on failures, this begs another question:

How much INT/WIS can one give up to balance the benefit of a Geerlok?

When I wield a Geerlok, I am losing 25 INT (wizard epic). I have always followed the formula of using my epic when I wanted skill points, and the Geerlok when I wanted successes. The main reason that I got my pottery trophy early was so that I could have the +% modifier and still have max INT.

Test: Run a skill set with a toon with 100 INT + Geerlok.
Run a skill set with a toon with 100 INT and no Geerlok.
Compare
Run a skill set with a toon with 110 INT and no Geerlok.
Run a skill set with a toon with 120 INT and no Geerlok.
Run a skill set with a toon with 130 INT and no Geerlok.
Compare

There is a problem with the tests. They don't test what you want tested. In more ways than one.

First:
The only report I've seen is that skillups vary as 1/(int/wis). Without knowing the other factors involved, we can't extrapolate the test results. It's not a linear relationship. Assuming it's a simple inverse relation ship we could calculate the difference 25 int makes.

Second:
At skill 20, a geerlok adds a skill point. At 100, it adds 5. At 200 it adds 10. Assuming the combines are within the 5-94 percent success range, each point a geerlok adds will add one percent chance to success. The effect of a geerlok depends on your raw skill.

Third:
Success rates on combines where people care about skilling up are very high. At skill 174 (I believe), there is a 95 percent chance to make a just trivial item. Check your success rate before skilling up, you may be suprised.

So, I don't think your question has a simple answer. It depends on the int change swapping in a geerlok, your int in both scenarios, your skill and your success rate without geerlok.

edit: typos.

[Craftah]

If I think of it, I'll let you know when I have JCM 3, and someone can tell me what the conditions should be when I practice jewelcraft.

I'd say that 255 INT and geerlok would be the ideal conditions, the 255 INT to help with skillups and the geerlok to help with success. Once you are 59, 60 is just a day or two away (yeah they would be long days). 60 can get you KEI, KEI can get you the needed cash to get just about anything you want, depending on how long you can take the boredom of KEIing.

[Troodon]

OK, while it seems that several of you have a good understanding of statistical analysis, all of your experiments have a serious logical flaw.

You're doing the experiment with the assumption that skillups and succeses are related events, then using the experiment to prove your assumption. This is a bad case of circular reasoning.

Look at it this way: Say you're brewing short beer, trivial at 31, with a skill of 0. Obviously, since it trivs lower, you will succed more often on combines.

Then another character, with identical wis or int, brews minotaur hero brew, trivial at 248, with a skill of 0. He will fail extremely often.

The first will gain more skillups on succeses than failures.

The second will gain more skillups on failures than succeses.

Why? The first is simply succeding more often than the second. If you succed more often, than of course just by simple fact you have more succeses means that more of your skillups will happen on successes. It in no way proves (or disproves, for that matter) that the two events are related. Simple probability will cause one event to coincide with another if one of the events happens more frequently.

The experiments can't prove what you're using them to try to prove. Becuase the experiments haven't fixed a variable that can affect the experiment: chance of success.

Want an illustration of this, say you're trying to prove whether or not it's easier to catch a ball when it rains. Two people with equal skill are trying to catch balls thrown with equal skill. One is doing it in Hawaii, one is doing it in Arizona. The one in Hawaii will catch more balls while it's raining than the one in Arizona, for the simple fact, that in Hawaii, it's raining more often. It doesn't prove that the rain had anything at all to do with him catching the ball.

In order to prove whether successes and skillups are related events, you have to fix all other variables, including chance of success. You could do this though. Instead of having characters trying DIFFERENT recipes, have them all try the SAME recipe, with the SAME skill, and try to do a combine. Spread this out over several people, each with identical wis or int, each trying a recipe with the same trivial and possesing the same skill. This would be areal pain, but this experiment would truly prove or disprove whether skillups and successes had any correlation or not. The way all the posters above did the experiments proves no such thing, because one of the variables (chance of success) was not controlled.

[Faymar]

I'd be interested in seeing what happens if you take one toon doing a very high level recipe vs one toon doing the mixed path.

So instead of them catching up with each other at 46 (so the "kiola" exp effectively has a time at the same sort of trivial gap), have one >starting on< a recipe that trivials at 150 or higher.

I'm trying to decide if I can face creating picnic precombines for this, actually.

Toon 1 puts one point into baking and from there only makes misty thicket picnics.

Toon 2 puts one point into baking and does a mixed path aiming for the next trivial (inedible goo, steaks, etc).

Assuming you do this to skill level 46, it widens the gap between toon 1 and toon 2 so that toon 1 is never doing anything near his trivial level.

Or was this something you intentionally wanted to happen?

[Raamil]

Also has it been proven or not that the trivial level isn't the deciding factor in skill-ups. In the tests done so far it's been a high trivial combine against several lower trivial combines. It is possible with these tests that the trivial level is the deciding factor and not the success rate. Of course the results will show that more skillups happen as a result of successful combines, but if you're working with a lower trivial item, you will by default have more successes.

[Boleslav]

Zararazu,

You are wrong. Zera did a good job of accounting for the fact that the chances of success and failure were different for different recipies at different skills. The most telling way she did this was when she compared the rate of skillups on the combines that were successful compared to the rate of skillups on the combines that were failures.

Your point would be valid if she were only counting total numbers of skillups. However, she is comparing rate of skillups using recipes that yielded a pretty good mixture of failures and successes. When she says that on average successful combines yielded skillups around 40% of the time while unsuccessful combines yielded skillups only 19% of the time she is not simply counting successes.

If you were zealous about wanting to remove 'chance to succeed' as a variable you could repeat the experiment using Mino Hero Brews, which would yield a 5% success rate for the entire experiment, but that would only yield a marginal improvement.

Boleslav Forgehammer
Paladin of Brell in his 60th Campaign
E'ci – Destiny Awaits

[kiztent]

Also from a historical perspective, the conclusion proved was not the one I wanted to see.

People has done correlations showing that chance to skill up is greater on success for skill ranges and recipies with 95 percent chance success. I felt this was bogus and went to do a recipe range that was (1) easy to get to so it could be reproduced and (2) had a mixture of success and failure.

Unfortunately, these results also showed the success had a greater chance to skill up.

While it may not be the most rigorous methodology, it does allow for ease of testing, and I felt that a large body of reasonable results was more significant than a small body of rigorous results (and it acutally becomes statistically significant more quickly that a larger number of combines in the 95 percent success range).

Why don't you follow my example, and since you don't like my methodology, describe and EXECUTE your methodology and show your conclusions.

[Aandaie]

Recently, on making patchwork and spider silk items as a newb on stromm, I again saw that success in making the item has zero to do with skill ups.

MOST of my skill ups were on things I failed to make.

[Morani]

Unfortunately, these results also showed the success had a greater chance to skill up.

Your data shows that there is likely to be a CORILATION between success and skill-up. I have seen nothing (from anyone) that has tested anything but a correlation between the two. Honestly, I can't think of a way we CAN test more than a correlation given the restrictions of the game. It is important however to note/remember the difference between correlation and direct relation. As Zararazu pointed out, two things that seem to follow do not have to be directly related. If skill-ups and success are calculated on SIMILAR formula (and they DO seem to be) that uses the SAME number out of the RNG (as in one role results in one, same, number that is plugged into each formula) then you will have a parallel correlation. Like two parallel lines, they don't ever touch but run pretty much the same course. The only way to test if they are correlated rather than directly related it to try to affect one without affecting the other. If you can then you have proof of correlation. If you can not than you have a theory that they are directly related. A lot of people forget that in science, we have no FACTS, just things that we haven't managed to disprove yet. Again though, we come back to being stuck at the question of "Can we even TRY to affect one without affecting the other within the limitations of the game?" In real life we'd have a LOT of options to play with but here we have a small handful.

I will say that I've found what is being presented interesting. I did not expect even a correlation between the two so that is a bit of a surprise. Also, the data indicating that fewer attempts are needed per skill-up when closer to the trivial point is nice to see. It's something that I've always believed I've observed but seeing actually data is helpful. I can handle being wrong about something better if I'm right about something else.

[kiztent]

Two thoughts.

First: it's a tested hypothesis. Until disproved, it's fact :P

Second: Since the two events, as you point out, are linked it doesn't matter if it's just a correlation or a true cause and effect. The end result (better to stay with high chance success items to skill up), is the same.

It may be interesting from a scientific perspective to show if they are correlated or cause and effect, but for all prictical purposes (i.e. skilling up) it may as well be cause and effect.

Let me reiterate my challenge. Devise a better testing scheme, and test it. Or show how my results can be better interpreted.

[Maldisdain Atheist]

I'd like to see the actual data runs... logs would be great, in fact.

(My suspicion is that we have a case of Simpson's Paradox here... though I could be mistaken. We'll see.)

Anyway, after doing a Bayesian causal analysis of 1042 data points of fletching skill ups from 201-250, I'll say with ~94% certainty that Successes have no causal influence over Skill Ups.

I've got about 5000 more data points to dump in, but since my log files make me cry (so I'm doing it by hand), I don't really have time for that at the moment. I'll put them in, and see if that changes anything.

Yes, I'll get the detailed information posted as soon as I can. Since I suspect people want an abstract: 953 Successes / 89 Failures. 45 SoS, 4 SoF.

Anyway, for the record, I think that the results that you are getting on these tests is related to the way that the data is being divided... we know that the early skill up rate tends to be more favorable than the late-skill rate (i.e. 100 combines starting at skill 1 is going to get you more than 100 combines at skill 240, for example)... so what I suspect is happening is that the average skill up rate is not evenly distributed across the low skill levels.

Until I see the logs, though, I won't know.

Furthermore, we're dealing with the sub-44 (or sub 66, depending on whom you ask) success formula, which is markedly different from the higher-skill success formula. So even if it is shown that there is a strong correlation between success and skill up (which doesn't imply causality), that correlation itself would not necessarily carry over to the higher skill ranges... which is where the interest in the problem arises, I think.

[kiztent]

post or PM an e-mail address you want the logs sent to (if I still have them).

[andyhre]

(I had accidentally posted this somewhere else before - sorry)

Okay, as I am pretty new to this I may be saying obvious or stupid things. I apologize in advance.

Do we have any knowledge of whether the skill up probability compares the Int/Wis (or other in special cases like Str for smithing) with the difficulty or trivial of the item? If that's true, that would seem to explain why it's relatively easy to get 10 skill ups early in the process and much harder when going (in my case) from 140 to 150. Besides the stuff costing more, the skill up rate has dropped a lot.

But if that does happen, then it's going to take fewer combines in the Kiola experiments for the folks who do easier projects, not necessarily b/c they skill up more on successful combines, but because they simply have a better chance for skill ups per combine and so it takes fewer.

Admittedly, I haven't sat and thought hard about the data, so I may be missing something, and separately, as a practical matter, either way the advice is the same (as others have mentioned), which is keep your trivial as low as possible if you want to skill up. However, if the causal reason is the relationship between wisdom (can you tell I play a ranger?) and trivial, then you'd be inclined to equip a wisdom booster in your primary hand, while if it is really success, you might want a geerlock. So there is a reason to try to break it apart.

And so my first question, then, is the one I started with: Do we know if the system for skill ups relates the skill-ip attribute (int/wis, etc.) to the trivial or not.

[Morani]

First: it's a tested hypothesis. Until disproved, it's fact. :P

Science doesn't work that way but your own opinion is you prerogative.

Second: Since the two events, as you point out, are linked it doesn't matter if it's just a correlation or a true cause and effect. The end result (better to stay with high chance success items to skill up), is the same.

If you are simply looking at it as a "play it safe" thing, than you are completely right.

If you are looking at if from the point of what limitations we have given by the game, it might NOT matter. If there isn't anyway to separate them then it's truly a moot point. The end results offer information but there is no new application of the information. We already know that stats and skill mods have an affect and how to maximize that affect. Showing a correlation only adds to the pile of "this is why it's good" and proving only does more of the exact same thing.

If you are looking at it from an informational or understanding point of view, there is a BIG difference and I don't agree that the end results are the same. Granted, I'm a BIG stickler for scientific process but I feel there is good reason for it. First off, not delineating between correlation and proof leads to bad judgment calls (things like declaring proof of cold fusion…). It also robs the observers and testers of insight that might have been gleaned towards other things. The reason the scientific process exists as it does is that we seldom ever find exactly what we are looking for. Buy making sure observations are interrupted correctly often leads to clues about other things. Really, the scientific process is NOT about trying to find an answer to one specific thing at a time but a process about exploring. Secondly, I prefer to call things what they are. Too much "abuse" of corollary data happens and I see good reason to put an end to it (or at the least, educate people on the difference because it can sometimes mean the difference of life and death). When a company can legally advertise that their diet pills have been "proven to work in clinical studies" and the study involved people who lost weight while eating more nutritiously, exercising more, AND taking the company's pills there is a problem with it. True, the studies have shown a correlation, but they are NOT PROOF the pills work. People then spend thousands of dollars because they believe that science is always right. Science didn't even enter into that study unless you count BAD science. The BEST thing that can be stated from that is; "scientific studies have shown a correlation". But, that isn't as strong of a selling point.

For practical application, there may NOT be a difference but I think it's as much bad form to call something cause and effect that you do not (and maybe can not) have proof for as it is to say the sun rises in the morning because the call of thousands of roosters summon it into existence.

Let me reiterate my challenge. Devise a better testing scheme, and test it. Or show how my results can be better interpreted.

Fair enough (especially as turn-about ) I rebut:

Devise better - As stated earlier, I do not know if there IS a way to test it given the constraints of the game. This means I have though about it and come up empty. Not everything is possible (unfortunately), however, I will continue to think about it and if I figure something out I WILL post it.

Test it - I will not be testing it. I only get MAYBE 10 hours in the week to play and neither have the desire to spend all of my playtime that way nor do I have the resources IN game to run the data pool necessary. (Though if I DO come up with a way to test this and I can afford it, I'll probably run a testing set or two). I will leave the testing to others with more resources than I.

Show a different interpretation - That's what I was intending my by above post. I agree with 95% of what you have concluded. The only real sticky point I had was proof VS correlation. I was also trying to show how the same results could happen via an entirely different process. In science, it's is ALWAYS easier to disprove something than to prove it. Technically, all one needs to do is to show a different way to account for the same results. About the ONLY impossibility in science is Ultimate Proof.


Side notes:
Why do I stick so soundly to scientific process? Well, because it's the best system we have for testing this kind of stuff and it's what I know the best.

In my experience, discussion like this can sometimes ruffle feathers so I want to make it clear that I see this purely as an intellectual debate (and am LOVING the fact that it's staid intellectual). If something comes from me that sounds harsh or feels like a personal attack, it's not meant that way at ALL and I apologize. Feel free to bring it to my attention and I'll try to explain myself better. Yes; that was a CYA maneuver, but I don't enjoy ticking people off by accident. It's only fun when I mean to.

[Raiyaeq]

I just finished fletching last night. It's ironic, but when I have no fails, I get no skill up. I had a run of 84 combines with all success and no skill up at skill 226. After that, I got 2 fails within the next 5 combines and got a skill up. 2 combines later another fail and follow by another skill up. Maybe it's coincident, but I've gotten more skill up while getting fails in fletching.