Tradeskill Trophy and Successes

[PinyonTreedotter]

I finally got my baking above 230. Working with Misty Thicket Picnic, I found the success rate dismal (expected, but still disheartening). So, I decide to make the Grandmaster Baker's Spoon. Equipping this in my range slot, I hoped to get a decent success rate.

I tried 6-8 X 20 combines and didn't notice any increase in success rate. So, I ran an experiment.

Trophy in Primary (237 skill): 14 successes out of 20
Trophy in Range (237 skill): 7 successes out of 20
Trophy in Secondary (237 skill, 238 skill for last 5 combines): 9 successes out of 20.

The 7 out of 20 seems about right from my experience. I did the statistics, and the Primary vs Range difference is significant (Chi-square = 4.9). I didn't check the Primary vs Secondary, but suspect it is not significant.

Is anyone else willing to do some tests to confirm or refute this? And to add more data to the Secondary slot. Baking is a good one to use, since the costs are relatively minor, although tailoring may also be a decent choice if you are using Solstice Robes. Having a skill of 200-230 would be the best, since the difference in success rates should be largest. Also, choosing a hard item (the Solstice Earring parts are good choices) will give the best chance to see a difference, if it exists.

For now, I'll act as if the mod for tradeskill does not work except in the primary slot.

[Verdandi]

20 attempts is not remotely enough to draw a conclusion, I'm afraid. 200 attempts would make me think you may be on to something, but it'd take more like 2000 for me to put any trust in it. It's been several years since my last statistics class, but just from experience in EQ, some weird things can and usually do happen over 60 combines.

Randomness is a tricky business.

[PinyonTreedotter]

Verdandi, that's why I asked for help from people to redo the experiment. It is not an unreasonable concern. The item can be put in the ammo slot, but the effects don't work. It seems plausible that the tradeskill mod might not work unless it is in the Primary slot. Hence, the call for some people to test this.

I know there are severe problems with autocorrelation and the random number generator. As such, I probably shouldn't have given a Chi-square value. Even so, the results are dramatic.

It sounds like no matter how many we do, you will be unconvinced. I did the test since I had noticed a trend in 120+ combines over several days (I normally don't do more than 20 per day). In those 6+ sessions, I never got 50% success. Most have been between 25 and 30% success. The first time I do the combines with the trophy in the primary spot I get over 50%.

If you require 2000 combines for you to "trust" the conclusion, you clearly missed one of the most important lessons in whatever statistics classes you took.

Randomness is, indeed, a tricky business. Which is why some of us spend a lot of time learning the tricks.

[tlag]

Physicians have a very simple rule of thumb to determ if they have enough data to make some prediction:
take the squareroot of number of observations (here combines)

if you see effects from the average that are more far away then the above calculated number you are likely to have seen something (but in a third of case you are still wrong to make a prediction). Only if it is twice that away you can be upset to have found something.

So lets take your numbers:
14+7+9 / 3= 10

In 20 subcombines you are supposed to encounter sqrt(20) ~= 4.5 fluctuation from the average. Yep your numbers do fit into that range.

[Itek]

Hmm... might be misunderstanding...

you are asking / saying "does the tradeskill trophy make a difference in successes based on where it's placed?"

If so....

Tradeskill trophies ONLY give the +skill modifier when equiped in the PRIMARY slot. (this is a known, I thought it was in the FAQ, it's also true with Geerlok, so you can't stack +skill modifiers or have more than one, like say a brewing and a smithing)

14 in 20 for the primary slot
7 in 20 for the ranged slot
9 in 20 for the secondary slot
(which also fails to take into account an additional point of skill, and therefore an estimated 1 percent higher success rate, for the last 25 percent of the test, which is almost 5 percent, which is 1 in 20)

[7 in 20] and [8 - 9 in 20] are VERY VERY close to [8 in 20].

[14 in 20] is VERY VERY far from [8 in 20].

Hope that clears up the confusion... unless it's me that's confused and you are talking about something totally different.

[PinyonTreedotter]

Another set of data (Picnics for baking):

Spoon in Primary, Skill 238: 12 successes, 8 failures
Spoon in Range, Skill 238-239: 9 successes, 11 failures

My skill is now 241, so I am beyond what can be tested.

Note: Even if we are suspicious of the RNG, we can frame this question in a different manner. If there is no difference between the Range and Primary, then half the time we should have more successes with the range, and half the time we should have more successes with the primary. This provides a very simple nonparametric test (albeit with extremely poor power).

Because of skillups during the experiment, you should always do the primary one first, then the range. This will be the most conservative. Or, if you wish, you can alternate or randomly decide which is first (less conservative, but still valid). Do the experiment only when your unmodified skill is between 220 and 240 at the start of each experiment.

Please, your help will be invaluable.