TH testing: final results

[Byrthnoth]

Assuming this is how it works, it only happens once, right? Otherwise a mob always drops something XD Tbh, I don't believe this is true (but as I said earlier, all ideas are welcome). I just want to know why you think this is true, maybe I missed a post in here.

Oh, I don't believe all the options I listed are true, in fact some of them are exclusive (I tried to number them the same). However, if you use Miraun's theory on the drops from your previous page, some of the predicted values fall outside your 95% confidence interval. If you use this theory, some of them also fall outside your 95% confidence interval. This was the version I used above in response to Mojo's post. I doubt it's actually what happens, especially due to your Stirge data. If you try to work it out algebraically, you'll find that they're not the same thing I think.

Also, 4a and 3b are not exclusive, 4a is just a mechanism by which 3b could occur.
40% drop rate -> .784% drop rate using 1/2/3b/4a with TH2 (~100% increase)
20% drop rate -> .488% drop rate using 1/2/3b/4a with TH2 (~150% increase)

It's just a matter of how the increase is determined. From your data, it doesn't look like the drop rate increase granted by TH2 is linear, which means it's either static (3a) or higher order (4a/b).


By the way, if you simplify the equation, it turns into (1 - No Drop Rate^3) = Drop rate with TH2.

Edit: There were some tests done on Dhalmels where they actually recorded what the drops were, right?
Also, does anyone have an example of a drop that is possible with TH2 but not without? Like, Lizards in CN can drop 4 skins at once with TH2, but their max is 3 skins with TH0.

[Enedin]

However, if you use Miraun's theory on the drops from your previous page, some of the predicted values fall outside your 95% confidence interval. If you use this theory, some of them also fall outside your 95% confidence interval.

This made my brain explode.

Also, does anyone have an example of a drop that is possible with TH2 but not without? Like, Lizards in CN can drop 4 skins at once with TH2, but their max is 3 skins with TH0.

Gah! If I hadn't terminated my account I would SO test that right now. I'm fairly sure it's not true - I like to believe that mob drop pools are static and not influenced by TH, disregarding potency off course - and I'm more than willing to prove that. Just that I can't XD

I wanted to test single-drop mobs before moving on to multi-drop mobs (sheep in beadeaux, leeches in bubu) and those lizards were pretty much next on my list.

Yours is a good idea, and while I'm pretty sure it's false, I haven't got any evidence to support either side.

[Byrthnoth]

Well, actually I thought about the statistical significance and I am not sure I calculated it right. I think I would need to do propagation of error on your TH0 value, which would enlarge the interval hugely probably.

The formula for Mir's idea, 3a as I understand it, would be:
n=odds of no drop
(1-n^3)=TH2 odds of drop.

Going by that formula, the error of the predicted value should be three times the error of the "no drop rate?"

[Amok]

If you use this theory, some of them also fall outside your 95% confidence interval.

The base drop rate is also inside a confidence interval so the range where the true value could be in widens as you multiply it though, doesn't it? If you take the min and max of the confidence interval for the base drop rate and apply the the multiple rolls to them, then I expect there to be overlap with the confidence intervals of the measured TH droprates. If that's the case then that would mean the "TH level = number of extra rolls" theory hasn't been disproven yet.

Also, a 95% confidence interval still means there's a 1 in 20 chance the true value lies outside that interval. With enough measurements, one of them is bound not to fit.

[Miraun]

Well, actually I thought about the statistical significance and I am not sure I calculated it right. I think I would need to do propagation of error on your TH0 value, which would enlarge the interval hugely probably.

The formula for Mir's idea, 3a as I understand it, would be:
n=odds of no drop
(1-n^3)=TH2 odds of drop.

Going by that formula, the error of the predicted value should be three times the error of the "no drop rate?"

Err... don't think that formula is quite right... I think you meant 1-((1-n)^3).
Where N = drop rate

Kindof... I outlined the potential for multiple base rolls, rather than a single roll, which would accommodate for the 50% and 100% variations observed in the data. So it would more be 1-((1-n)^(x+TH)). Where X can be more than 1, therefore controlling the effects of the increasing levels of TH.

So if you want to be brutal on drop rates (ie: Salvage), use 10 rolls at .01%, which would be an observed drop rate of .1%, however, TH4 would only increase that to approximately .14% rather than .4%

Going by your earlier outline of potential effects by TH, this model could accommodate for either 3a or 3b.

[Aryden]

There's no way to tell that SE coded all jobs to have the same drop rates. All I'm saying is, the simplest way to get the result you describe is to do it with one loot slot as I said.

In order to test this, you would need a high sample size list (years of doing dynamis and meticulously recording even unwanted drops) of, like,
Red Mage : XX drops
Warrior : XX drops
Summoner : XX drops
etc.

which I'm almost completely sure is out there. If a histogram of the drop rates is normal about the mean, then it seems likely it would be my way. If it seems frequency is fairly constant (perhaps in a way you'd predict based on your model), then it may be your way. Overall, it'd be somewhat hard to tell though. I think parsimony is on my side.

I've been trying for hours, but there was a post a couple of years ago on Alla where a bloke listed out every drop his ls had gotten from all dynamis runs over the course of something like 3 years. I cannot recall whether he specified that there was/was not th4 present. I am still trying to find it, but someone with premium access should be able to find it faster than i.

[Eldelphia]

I've been trying for hours, but there was a post a couple of years ago on Alla where a bloke listed out every drop his ls had gotten from all dynamis runs over the course of something like 3 years. I cannot recall whether he specified that there was/was not th4 present. I am still trying to find it, but someone with premium access should be able to find it faster than i.

Pretty sure that was in the endgame forum

[Byrthnoth]

Err... don't think that formula is quite right... I think you meant 1-((1-n)^3).
Where N = drop rate

Kindof... I outlined the potential for multiple base rolls, rather than a single roll, which would accommodate for the 50% and 100% variations observed in the data. So it would more be 1-((1-n)^(x+TH)). Where X can be more than 1, therefore controlling the effects of the increasing levels of TH.

So if you want to be brutal on drop rates (ie: Salvage), use 10 rolls at .01%, which would be an observed drop rate of .1%, however, TH4 would only increase that to approximately .14% rather than .4%

Going by your earlier outline of potential effects by TH, this model could accommodate for either 3a or 3b.

n being the odds of no drop:
For TH2: (1-n) + (1-n)*n + (1-n)*n^2 = 1 - n + n - n^2 + n^2 - n^3 = (1 - n^3)
For TH1: (1-n) + (1-n)*n = 1 - n + n - n^2 = (1-n^2)

So the general formula would be (1-n^(1+TH level)) I think, which may or may not be the same as what you said.

Also, the uncertainty of the prediction would be (1+TH level)*(uncertainty of n), which is quite large really. I'm trying to think of a way to test this without killing 10000 mobs.

[Thorny]

For what it's worth, I've observed the same (likely) placebo as Masamune. While SE has stated that 'TH is active' if THF touches the mob, I don't see how it's unreasonable for it to increase as THF's enmity hits certain levels. I'm not saying it's done this way, but TH2's proposed doubling droprate could easily be something like +5% droprate per 1000 CE or VE. Also, extremely limitted sample size, but I got pretty unholy drops trioing zhayolm with THF BRD RDM and tanking on THF. When I changed to bringing a naked THF and going MNK I stopped getting anything at all.

Onto something useful(sorry for delay, 4th of july and been distracted):

Colibri Feather(Lesser Colibri, TH2): 6382/11040 - 57.81%
Colibri Beak(Lesser Colibri, TH2): 5084/11040 - 46.05%

All kills were done with solely THF damage(THF chars left at the camp farming beastmens seals), occasionally being hasted/cured by a RDM char. I have a similar sample of kills done with MNK under same conditions that I'll post during maintenance sometime(once I get through all the files). Assassin's Armlets and Thief's Knife were NOT used. I feel this is an adequate sample to use as a control for an Armlets/TK test though, so in the future I will use them and establish a proper sample for comparison over the next month or so.

[Byrthnoth]

Colibri Feather(Lesser Colibri, TH2): 6382/11040 - 57.81% (+/-0.92%)
Colibri Beak(Lesser Colibri, TH2): 5084/11040 - 46.05% (+/- 0.93%)

.5781 = (1-n^3)
n = .75, so Feather drop rate would be predicted to be 25% with TH0

.4605 = (1-n^3)
n = .81, so Beak drop rate would be predicted to be 19% with TH0

[NynJa]

I'm not saying it's done this way, but TH2's proposed doubling droprate could easily be something like +5% droprate per 1000 CE or VE.

Such coding would shit all over Trick Attack mechanics.....and knowing SE this puts it right up their alley.

In all seriousness, it could check how 'active' the thf is though.

[Thorny]

Seems a large portion of my MNK logs are unusable due to also having a THF character in party. They did not perform any action, they were 500+' away and killing seperate mobs, but it still makes me unable to do a simple count script. Maybe if I care more I'll make one that skips any drops after a 'too far away to recieve exp' until a 'gains x xp' message to salvage them. I did not include these in this sample, which is why the size is so much smaller than expected.

Colibri Feather(Lesser Colibri, TH0): 792/2976 - 26.61%
Colibri Beak(Lesser Colibri, TH0): 578/2976 - 19.42%

[Byrthnoth]

Colibri Feather(Lesser Colibri, TH0): 792/2976 - 26.61% (+/-1.59%)
Colibri Beak(Lesser Colibri, TH0): 578/2976 - 19.42% (+/-1.42%)

So, that fits better with the prediction and pretty well excludes TH being a static increase (like +30%). One more thing Thorny, what were the maximum drops per kill? 2 Feathers and 1 Beak, or 1 Feather and 1 Beak?

Also, if two feathers are possible, it possible to figure out how many feathers dropped single vs. doubled?

[Thorny]

Maximum is one feather and one beak.

[Byrthnoth]

So TH being a static increase can be discounted. (+31.2% feather drop rate, +26.63 beak drop rate, outside each other's intervals)

TH being a direct drop rate multiplier can probably also be discounted. (+237% feather drop rate, +217% beak drop rate), but I am less sure how to run stats on this one.

Assuming drops are independent, we can also discount the "only rerolling if nothing drops" option.

So, now we have two very well defined points. It alone does not validate any specific theory, and I am sure there are plenty of theories that could fit the two points. Still, leaning somewhat heavily on the crux of parsimony, I prefer the rerolling drop rate explanation.

[Thorny]

Hate to bring this up again since I'm sure it'll result in clueless idiots trying to post one-ups(Viewpoint), behavior similar to my own directed at me(Mojo) and more anecdotes passed off as fact, but I'm thinking of how best to design a test for thief participation vs droprates. Obviously the bare bones are simple, compare kills where a THF has capped enmity to kills where a THF has minimal enmity but still acted upon the mob. A few guidelines need to be approved by the community for testing to have meaning though, primarily:

1. What is the minimum needed to apply TH? Will casting one of the 1 CE nerfed RDM spells suffice?
2. Do I need to account for day or moon phase? This type of testing would require active play, so it's not feasible to get a sample size of 10,000 and negate it through law of averages. Does the community believe day or moon phase have an effect or not?
3. Do -ga killshots in any way effect droprate?
There isn't going to be a straight yes or no to these, I want opinions because I want my test to have meaning to the people who matter.

Furthermore, I'd like help identifying a proper test mob that meets the following criteria:
A. Has at least two drops, one of which is in the 5-15% range and one of which is in the 40-60% range. This would make any effect more easily noticable.
B. Is at least Decent Challenge to 75. I want to be able to cap CE and VE before it dies.
C. Is plentiful and accessible. I don't want anything in an instanced area, nor do I want anything that spawns less frequently than 20 times per hour. If indoor repops, there should be quite a few of them.
I can pull off 3x 75 BLM + a THF with up to TH4, or RDM BLM BLM THF, etc, so difficulty is not a major issue. I just want something that can be done in a timely manner.


Any information/help would be greatly appreciated, even if you think I'm being silly.

[Byrthnoth]

Well, first step is to examine what you already have and see if SE is full of shit about Thf needing to do something.

I have a monster in mind, but there might be some problems with it, I will pm.

I'd go:
1) Yes
2) No, screw day/moon
3) I have heard it does matter

[Shamaya]

I have also heard that there was a patch several years ago which nerfed TH's effect in the event of -ga spells. I believe the targeted mob gets TH on it, but the other mobs do not. I have no source at all to back this up though.

Just going to link some more SE commentaries,

Developer Interview References

I scoured the internets tonight for the developer interview in which they said that TH is only dependent on the Thief being in the party, and not having to contribute to the final blow and such. I got into an argument with someone tonight camping stupid ValkEmp regarding TH, and I wanted to make sure that what I heard was correct.
An old school interview...

• FFXI.crgaming Interview from Nov 2003

• By having a thief in the party, how much increase is there in the Gil/Item drop after defeating a monster?

We are unable to give too many details with regards to the value of points. However, if the thief is skilled with the Treasure Hunter job ability, it may not be a bad idea to adventure with a thief. Also, with regard to gil, when a thief joins a party, there are no changes to the amount of gil obtained. Does this mean that Gilfinder only works in solo, or something? Tahngarthor 20:37, 17 January 2007 (EST)

• Does having a thief increase the possibility of a drop even in key/quest items (aprons, skull, wadi grass, etc.)?

Of course! Only the thief has the ability to acquire the Treasure Hunter job ability. There is an increased chance in obtaining dropped items. More recently...

• FFXI.Stratics Interview from May 2005

• I have to ask on behalf of all thieves out there, are there any conditions to activating the Treasure Hunter ability or is just being in the party enough? There are a whole bunch of theories such as thief having the last hit.

If there is a thief in the party then you will have the treasure hunter ability, if you want to have some conspiracy theories well we can't tell you. If you want to think that it's the last hit then you can go ahead and think that. And the kicker...

• Allakhazam Boards Post from Mar 2006 (this is someone's liveblog deal thing... I couldn't find a more "official" transcript of the interview. All other references I found regarding treasure hunter pointed back to this post.)
• I also found this on a more reputable site, but it does not have the full transcript here: 1up Festival Summary

• How does Treasure Hunter work?

Can't say too much, but it has nothing to do with damage or the last blow. You simply must have it in the party. I suppose this was more for my own personal reference and to reconfirm my belief in ye olde Treasure Hunter. I have a feeling that I read the last information on an actual transcript, perhaps even on the Fan Festival website... but I don't know, it seems to have been changed from what I remembered. --Waluigi 07:33, 20 August 2006 (EDT)

• Fan Festival 2007:

How does Treasure Hunter effect drops in battlefields, dynamis, chest contents. No effect on chests. Same effect in dynamis etc. as elsewhere. For TH to be active THF must perform an action on the target. Does not have to be killing strike.

The way they talk about TH, especially the middle of the three bolded portions, makes me think, again, that TH does not differ in potency whatsoever, so long as it is active. If people want to go ahead and test it though, I myself am very skeptical in nature and can fully understand.

There's more commentary from Fan Fest 2007'. I believe the one linked is from November of that year, in California. There was more though, translated by Elmer from the Osaka Vanafest about a month earlier (iirc). But the link I've got (an old BG thread) is broken, it appears ;/. // Actually, after just searching, I found the comments in question in my blog, but I still do not have the original source. Either way:

"Q: Do multiple instances of Treasure Hunter in a single party increase the chances of a drop? When exactly does the effect of Treasure Hunter take place?

A: When more than one member has Treasure Hunter, the player with the strongest version will have theirs take effect for the party. The effect comes into play as soon as the Thief generates enmity from the target enemy. Even if the Thief dies, the effect remains."

"When there are multiple instances of Treasure Hunter in a party, the strongest effect is used. Treasure Hunter goes into effect the moment the Thief actively puts their self on the target's hate list. For the most part, once the Thief attacks a monster, they remember the effect of Treasure Hunter. If a Thief dies, they do not have to be Raised to strike the monster once again. However, if you lose possession of the monster and they regenerate back to full health, this information is lost."

I'm not sure if the second part is Elmer's added commentary. If it was SE saying that again, it'd sound a bit redundant.

Again, if you wanna test for potency, be my guest ^^. The part I underlined makes it seem as if there is also no potency difference, but if you look at the commentary prior, you can see that they've basically spread open falsehoods (or at least been very misleading) in the past.

[Thorny]

Aside from vague wording, doesn't even seem to be an official transcript of that interview around.. while not claiming it's made up it certainly isn't the most reliable source I've ever seen.

[Byrthnoth]

Plus, SE's interview panels have never really shown themselves to understand much about how their own game works. I would be unsurprised to discover their primary reason for not answering game mechanism questions was a lack of knowledge about how they work.