Courting controversy?

[Hateborn]

I just wanted to celebrate getting to 201 through casual smithing - I didnt think it would be such a relief to see this number arrive - but by heck its a bit of effort required to do this as a casual unguilded smith. (By casual I mean, not dedicated to farming - just occasional runs of combines when hunting and opportunistic bargain hunting has built a stack of components).

Thanks to the forum regulars for posting so many useful thoughts - and occasional stories of woe with the RNG that ensure we never feel alone. There is always someone out there who has had it worse.

Now the potentially controversial bit - when I first came here to learn about smithing a few months back one topic often under discussion then was the relative merits of geerlok vs stat increasing item.

The consensus was that you should use your geerlok because the chances of a skill up were significantly greater on a success than on a fail - therefore increasing chance of success is a good idea.

I decided there and then to not only track every skill point for number of attempts (starting at 116 - which is where I joined the forum) - but also to track the number of skill ups on success vs fails.

Now - despite the capricious nature of the RNG most of us succeed on combines far more often than we fail (when skilling up and following the recognised path of making non trivial items that arent too far out of reach - not when considering attempts at high end never trivial combines for profit or use) - so skill ups on fails seem very uncommon - and they are. However when considered statistically over this medium sized sample (skill 116 - 201) and bearing in mind this covers the nice easy skill up pre 188 stuff - I got the following results - 1 skill up every 5 success vs 1 skill up every 10 fails.

So I propose that you are exactly twice as likely to get a skill increase on a success as a fail. (I would definitely use my geerlok over a stat increasing item).

[Xenephex]

most of us succeed on combines far more often than we fail

You don't post actual numbers of successes vs. failures, but extrapolating from your post and the above quote, I would hazard to guess that you failed somewhere in the area of 100 to 200 times over your skill-up run.

That's not enough to be statistically significant. The data is worth noting but additional numbers would be required.

[Footsore]

I'd also like to point out that a statistical correlation does not necessarily indicate a cause and effect.

Even if you are able to show that you receive more skill ups on successes than failures, it doesn't mean that the success contributed to the skill-up.

It could simply mean that the success and the skill-up are generated from the same randomly generated number. (Instead of running the RNG twice - once for success and once for skill-up, they may run it once and compare each to the same number.)

[Boleslav]

This has been discussed for quite a while. In the past, I did some trials and took data from other people's trials. In general it showed a bias toward getting skillups on successful combines. The likelihood of success (based on skill and trivial) was taken into account to isolate success vs. failure as much as possible. At that time I estimated a 50% increase in chance to get a skillup on a success versus a failure.

The tests were done by tracking all attempts, and noting whether each skillup came from a success or failure. Each observation was weighted by the chance of a successful combine at that skill and trivial level. There were multiple observations from Jewelcraft, Pottery, Baking, Brewing, and Fletching.

The data showed a higher skillup rate on successful combines than failed combines. I understand the previous poster's argument that association does not imply causation. However, association does imply association! So regardless of what the actual cause is, if you can create more successful combines than failed combines you are more likely to get skillups.

In a separate experiment, I took two toons with equal stats from 0 to 100 in Jewelcraft. In one run, the character made the next trivial item to keep the chance for success as high as possible. In the second run, the character made Gold-Amber rings (trivial 102). The first character took 20% fewer combines to reach a skill of 100.

I am very convinced that there is a significantly higher chance to get a skillup from a successful combine than a failed one.

Boleslav Forgehammer
Paladin of Brell in his 65th Campaign
E'ci – Sacred Destiy

[Zeralenn]

As am I. I did an experiment six times with eight identical brewer toons (level 1 dark elf agnostic mage with 135 INT), four doing brewing to 122 on fetid essence alone (triv 122) and four doing a path that kept them closer to their trivials and thus less failure. On average, those who stuck close to their trivials got to 122 in 20% fewer combines than those who did it to 122 with fetid essence alone. Again, statistically insignificant (we'd have to have a mass trial of this with hundreds of people participating and I believe there are people here who are just bent on not believing this to be true anyway, so there really is no point in doing the experiment).

And of course there are also people who say that that is also no proof that these would apply to upper levels -- I say there is no proof that they would NOT apply to upper levels, so again, stalemate.

I too am a "success yields faster skillups" theorist, based on my own tradeskill experience.

...Zera

[Reflan]

Some people lose an awful lot of stats dropping their main(often an epic with lots of wisdom or INT, as well as str/dex for skills that use those.

So an argument that using a geerlok is the best thing to do is a bit different than arguing that making things close to your trivial is the best thing to do.

It's arguing that you using a geerlok can make up for a sometimes very significant loss in what may be your primary skill-up stat.

Personally, my skill up rate was terrible for very long stretches. so I can't in any way point to my personal experience as proof that I did things the best way.

But I did seem to notice much better success with high stats, as did everyone I spoke to in the game. I got lots of tradeskill AA to push my wisdom as high as possible and found the results were positive over the course of JC, smithing, and brewing and pottery.

I always made the thing that was closest to my trivial.

But the results of greater wisdom for my shaman tradeskiller, I found, were notable enough that I really didn't want to give up my epic for a geerlok.

Six of one, half a dozen of the other. More wisdom/int is great, and a geerlok is great.

Is it better to dump an epic for a geerlok and dump your stats when you do it? I'm not so sure it's arguable that it is. You give up one method of improving your skill gains more quickly in hopes of using a different one. I'm a little more confident in the benefit of having high stats than of sacrificing them for a geerlok.

[kiztent]

Originally posted by Reflan
Is it better to dump an epic for a geerlok and dump your stats when you do it? I'm not so sure it's arguable that it is. You give up one method of improving your skill gains more quickly in hopes of using a different one. I'm a little more confident in the benefit of having high stats than of sacrificing them for a geerlok.

Depends on your stats without it. No one has done a conclusive study for how much 20 int/wis increases your chance to skill up, but we do know that at 100 skill, a geerlok increases success rate by 5 percent, and 10 percent at 200 skill (assuming, of course, success rate is not low capped) - absolute rate of success not relative.

Of course this assumes you are not capped at 95 percent success rate (which happens a lot on close to trivial items in the 200 range).

Even if the correlation (success yields skill ups, is causal or just a measurable correlation - which is the same thing for people trying to skill up) until you could gather the effect of increased int/wis there's no way to tell if a geerlok helps.

Of course, I cap STR with a geerlok when I smith. Shrug.

[Astroshak]

Originally posted by Reflan Is it better to dump an epic for a geerlok and dump your stats when you do it? I'm not so sure it's arguable that it is. You give up one method of improving your skill gains more quickly in hopes of using a different one. I'm a little more confident in the benefit of having high stats than of sacrificing them for a geerlok.

My level 52 enc has a 255 int (the cap) without any buffs on at all. She can EASILY swap out her main hand item (a 15 int book) and get an int buff to get her back to 255. She could almost as easily replace a slot or two for items with more int (such as her ODS, which has no int).

In my opinion, if you intend to seriously persue a tradeskill, you should get both high (one of : int/wis/tertiary stat) and get a geerlok. This is especially true if you've got the plat to try to get a skill up quickly. Obviously, someone who intends to persue a skill over the course of a character's career, as that character obtains the funds for a skill run, would have a much harder time going about this; but those people would be better off with a geerlok instead of the best stat mainhander they could afford.

[Kronepsis]

Or you can compromise like I do and get a trophy early.

Hehehe.

[Kongming]

I'm not sure i do alot of mine combines with much higher trivial then my skill and fail alot on other skills. Mainly misty for baking, imbued idols for pottery, arctic wyvern crap for tailoring, and skill up happy quite often on failures. heck on arctic wyvern im failing 90% of the time, or i was but still getting skill up at a good rate.

[OberonMiM]

my experience is that the closer you are to the trivial of an item the more likely the skillup. But in turn that is also more likely to succeed

So it might not be the chance of success that deems a skillup. It might be the closeness to the trivial of that item (which just so means that you will get more successes on average during a skillrun which makes it seem like your more likely to skill up on a success)

[scarbrowed]

The problem with most of the tests mentioned here is that they do not distinguish between 2 similar but different possibilities:

1. Skillup more likely on success.

2. Skillup more likely when trivial close to current skill.

Zeralenn's test is actually a test of the second. But since close trivials lead to more successes, it's hard to tell the difference. I personally suspect number 2 above is closer to the truth. But it would take a lot of data to prove either way.

[Vilantra]

I saw two people that did tests skilling from 0-120 or so. With their tests, they used items that trivialized at 120. Going in a big jump took 20% more combines than going in small trivial steps. The same should be done with a +skill modifier that doesn't boost stats. +5% would probably make a significant change in success rate. If the skillup rate is faster with the +5%, success affects skillups. Else, it is based on closeness to trivial.

Of course, this all relies on the chance of skillup and the chance of success being global, instead of being based on some hidden stat each tradeskill might have. It all goes out the window if two items with the same trivial in the same tradeskill have different success or skillup rates. I'm guessing that isn't true, but with SOE, you never know.

[Iumiena]

the only way to test the first hyphothesis (skillup is more likely on success) is to take a single char and try to get 100 or so points using items much higher than your current skill level (100 or more point difference). Keep track of how many skillups you get on success and how many on failures. Since the skill is the same there won't be the ambiguity in whether closeness to trivial affects skillups more than success, because the skill will remain constant (or as constant as possible) Would probably have to redo this several times to get a suffienctly enough data pool (rules at 100 skill might be different than 200+ skill, so keep the range very tight). Would also have to try to make the distance between your skill and the trivial relatively stable by slowly making harder items as you go up.

I'd say this would be best tested on either brewing or JC (seeing how they have a fairly well distributed trivials on their items).

However, I am not a massocist to such a degree, so someone else could test it. The problem would be to get a success sample sufficiently large (20% success rate on items with high triv, so 100 combines, only 20 sucesses, would have to be at least 500 combines)