TH Research - Is TH cumulative? Does "PT killshot" matter?

[Shamaya]

I don't post often here at BG, so bear with me. Takes at least a lil guts posting this because I don't know what's "Bg-worthy" and not bg-worthy. But I think its due time I posted what I've slowly been working on for the last two months. Note to public--I post WoT's (sorry).

In any case, after seeing a lot of discussion following SE's recent TH commentaries in the fall, I decided to tackle a few of the remaining issues regarding TH, those that didn't seem to have a clear consensus or answer.

I wanted to come to a conclusion on the following topics:
-Is TH cumulative? ; That is, does 2+ thf in different parties in an alliance(s) = improved drop-rate.
-Does the party with the thief need to get the kill?
-If the thief performs an action on the mob which generates enmity, then leaves alliance, does TH remain?
-If the thief "stickies" the mob with TH2+2, can they safely melee in no TH+ gear for the remainder of the fight, with the TH2+2 effect still remaining upon mob death?

I bolded those first two, because those are the issues I chose to first address. I haven't yet tackled the other issues and I don't know if I ever will have time to. I imagine the answer to the last question might be confidently inferred if the answer to the 3rd is found. In any case--is TH cumulative? Does the PT with TH inside it need the killshot?

You can find a detailed explanation of how I conducted this test, etc, at the lolLivejournal I made just for this kind of stuff. For here, I'll keep it short:
*Target mob: Death Jacket in Yuhtunga
*Can drop 4 different items, but only one of each kind per kill
*There were three phases to the test. All phases consisted in an alliance of 2x 1-person parties.
Phase 1: TH2+TH0 (TH2 PT Killshot)
Phase 2: TH2+TH0 (TH0 PT Killshot)
Phase 3: TH2x2
*Both players hit the mob at least once.
*The samples all took place in a single moon phase.
*There was no control for Day or other potential voodoo-nonsense type stuff.
*Roughly 900 mobs served as the total sample-size for this test.

The following image shows total drop-rate difference between having one thief in the alliance and two thieves.
http://img.photobucket.com/albums/v7...1812-151-7.jpg

The next image shows the drop-rate differences among the three different test phases mentioned earlier.
http://img.photobucket.com/albums/v7...1812-151-7.jpg

This next image shows the "drop-event" differences. For this test, I defined a drop-event as the status of the number of drops in the treasure pool at death, ie. no drops, 1 drop, 2 drops, etc.
http://img.photobucket.com/albums/v7...1812-151-7.jpg


The following is an example of increased precision resulting from greater sample sizes.
This drop-rates in this graph were derived from a total of roughly 430 kills.
http://img.photobucket.com/albums/v7...pRatesx500.jpg
The drop-rates in this graph were derived from a total of roughly 900 kills.
http://img.photobucket.com/albums/v7...-151-7x500.jpg


Alright, so enough of graphs. I think they illustrate things quite nicely though. Is around 900 total mobs killed enough for this kind of test? I think so, but I'm no expert in statistics, so I don't know how to calculate my margin of error. It seems pretty clear to me though--if there is some sort of cumulative effect for TH--that is, if stacking more thieves in different parties within an alliance were to increase drops, then it's low enough to not be picked up in a controlled sample of this size. If such a cumulative effect did exist, it would be such a low bonus as to not be worth considering. Or, it could be the case, as I think, that there is no such cumulative effect. The same can be said for the results regarding which party got the killshot (the party with TH or the party without TH). Thus..

What I conclude -
1.Putting a thief in every party does not improve drop rates, so long as one thief performs an action on the target mob which produces enmity.
2.Further, the party with the thief in it does not need to get the killshot in order for TH to activate.

The drop-rates I experienced were lower when the party with TH got the kill. Even lower when a 2nd thief was added to the alliance. This obviously should not be taken to mean that "omg TH has a negative effect on drops!", but the result is simply the artifact of the sample size not being large enough to make the rates completely equal. It's been months since I started this, and I went out and performed multiple samples. I'm fairly confident that with a large enough sample sizes, the drop-rates would be completely equal.

I think that's about all I'll post on this in the OP. It's a mouthful anyway. There's more information in the livejournal, as well as further data, including raw data in excel documents that I posted. You can also find even stronger TH research than this, regarding answers to all the aforementioned questions, here.

Surely there are people here with enough knowledge in statistics to help me find margin of error. Given a certain hypothetical potency assigned for how cumulative TH might improve drop-rates, I'm interested in knowing my margin of error for said hypothetical potency.

Please, questions/comments/criticism. I'm all ears.

[kareface]

This doesn't answer the one question I have. Is TH specific to the killing alliance? (I.E. Dynamis areas) We already knew most of the info collected from soundbites from SE. But even SE mentions that TH is alliance wide. What if you have more then 1 alliance?

[Tythera]

I thought that recent interview had said it was like a debuff on the mob more than anything relating to parties or alliances, which would make it so it wouldn't matter in dynamis :bagel:

[Catrina]

I'd like to point out that you'd need a bigger sample for conclusive results (at least 10000, or 3000-4000 kills per test at least).

From what I can see from your graphics, the answer is no, the graphic differences come from the sample size, it doesn't make a difference, once TH is in, only one is needed.

[Skeetaru]

Early for me, but is that 5 charts just to say "Yeah we really don't know shit about this still"?

[nekoshoonen]

Wether SE said something or not 90% of my LS is still retarded about TH and keeps on using superstion when it comes to it.

Thank you for this I'm going to post a link on my LS website!

[Shamaya]

Thanks, Neko. I'm hoping, if I can get sufficient sample sizes, as Catrina says, that this can all be of some practical value. I know that my dynamis shell and a few friends already act in accordance with these results.

But getting the right sample size is crucial. That's why knowing my margin of error given a certain hypothetical potency is crucial. But that's something that I don't know how to do. 10,000, 3,000. You might be right. Virtually 1,000, as I have, might be right as well. If I'd done a sample of 200, people would have said 1,000 is needed. I have roughly a thousand kills. If I had 2,000, someone might go ahead and say I need 8,000. But really, those estimates are just spit-balling.

I thought that recent interview had said it was like a debuff on the mob more than anything relating to parties or alliances, which would make it so it wouldn't matter in dynamis

This was inferred, not stated. You can check wikia's wiki to see what, thus far, SE has actually commented on (much of this, I updated myself) - http://wiki.ffxiclopedia.org/wiki/Treasure_Hunter

This may be implicit to some that read the OP. And they'd be right. Part of the reason why I'm after knowing the things stated at the beginning of the OP is because I'd like to put that debuff theory (because that's what it is--a theory) to the test.

This doesn't answer the one question I have. Is TH specific to the killing alliance? (I.E. Dynamis areas)

The reason I haven't tested this is because, of course, I simply don't have the means to test it. I can't fathom doing a TH test in dynamis or einherjar. However, I leave it to the readers to infer the answer to that based on these results. All I can offer is this -

Q: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.

[Nyuneko]

Source: viewtopic.php?f=2&t=25584&start=120

Original text(JP):

???????????????????????
??????????
Kumasuke??

???????????????????????????
pei??

?????????????????????????????????????????????????? ?????????????????????????????????????????????????? ?????????????????????????????????????????????????? ?????????????????????????????????????????????????? ??????????????????????????????????????HP?????????? ???????????

And then, this is Elmer's translation.

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.

In any case,

-Is TH cumulative? ; That is, does 2+ thf in different parties in an alliance(s) = improved drop-rate.

This question was actually asked in the Q&A that day, I don't know why Elmer missed it in his translation, anyway, their answer was: When there are multiple players with the Treasure Hunter trait, only the strongest effect will be active, the additional Treasure Hunter will not be added to the effect.

-Does the party with the thief need to get the kill?

I would say from what Elmer translated and the original text that the effect is like a debuff that you put on the enemy, with that in mind, it probably was pretty obvious that THF only has to "put" the treasure hunter on the target to trigger its effect.

What I am more interested in is really the last point of yours: whether or not I can switch back to a better dagger and hands(when I get my armlets...) after the TH effect is placed. Is TH like a debuff(think "Slow" "Paralyze")? And does the THF need to be actively be on TH2+2 to keep the effect at TH2+2, if +2 equips are removed, will it become TH2 only?

Lastly, what is the general belief of "Getting on the hatelist"? I would say any action that causes an enmity increase would activate TH, however, according to the last few posts on the Treasure Hunter thread, only an offensive action that is performed onto the target will activate TH; the problem is, it wasn't fully tested, with lots of assumptions as well.

[Jast]

this proves nothing

[Plow]

Early for me, but is that 5 charts just to say "Yeah we really don't know shit about this still"?

More like 5 charts to say "th uses the strongest version which attacks the mob and nothing else matters."

[Shamaya]

The translation was by and large fantastic.

Translations are so difficult, though. So in a nutshell what you're saying is that SE commented on multiple instances of TH on a mob in general, and not just multiple instances of TH in a single party?

I've been participating in discussions on this in-game, on BG, on Allah, and on KI, and a lot of debate has gone around over what was meant regarding "the strongest effect takes place for the party."

I've heard about applying the gilfinder testing to TH--that's kinda cool. The post in there isn't bad, but then again I can see why there weren't a lot of responses. People like to see numbers before they feel comfortable.

As for the whole enmity thing, there was this following commentary,

For TH to be active THF must perform an action on the target. Does not have to be killing strike.

I believe this was a follow-up on TH. Someone correct me if I'm wrong. Original statements were made at Osaka. I believe this statement was made at Anaheim. Someone mihgt say "An action on the target eh? Well what about hitting for 0 on the target?" In a previous statement, SE also mentioned enmity. So in a nutshell I think what SE's trying to say is "an action on the target, which produces enmity." Not simply an action, not simply getting enmity--but both.

Edit: Wow, responses in BG happen a lot faster than on other forums.

[Catrina]

Shamaya, usually in engineering, when I want to estimate costs and schedule variance, mean, expected value, etc, we set an automatic program at 4000-10000 iterations.


However, being that this is a game, it should be done within reason, and accept the limitations that a statistically small sample brings in. That's why I say that for the sample size, the results appear to be what is expected from a size statistically small (I know that technically and for human capabilities, tallying all that data is actually a lot of work and how difficult it is to collect as much data - once I tried and took me about an hour to collect data for 100 kills). I'd ramble on more but I actually strongly dislike probability and statistics lol...

[Vejitta]

Jast as always providing us with tons of arguments

Nonetheless I have to agree with him. Sample is quite big I admit but either it's not enough or the difference between 1 TH or 2 THs is placed within error margin. Just look at particular drop drop rates. We could say that some drop rates are improved by the quantity of THs in the party, which denies overall results. That's why I suppose this results stay within error margin.

[Amok]

I think the OP has questions I'd like to know the answer to. However, the first two things you decided to focus on were somewhat cleared up by SE's comments and the results you found afaik confirm what SE has said. Things like whether you can tag with TH3 or 4 and then switch gear and still get TH3~4 are worth figuring out. I'd like to see the difference between TH0-1-2-3-4 expressed in percentages, but that may be too much to ask/expect.

[Wizerd]

I'd like to see the difference between TH0-1-2-3-4 expressed in percentages, but that may be too much to ask/expect.

This is what I'm after as well. Unfortunately, it seems to be a little more complicated than increasing drops by certain percentages. (No, I don't mean things like TH1 affects 1 type of drop, and TH2 affects another kind. That kind of thinking is what gets you the masses of morons that currently play this game)

I started testing this using the bees also, but quickly found myself busier in my free time than before, so I haven't been able to continue. I only have 778 kills under my belt, out of a projected need of ~60,000, so it's definitely something that would take a while (and would be easier to do with collective and accurate help).

My idea was to choose 10 different monsters, all with at least 2 drops of varying rates, and kill each one 1000 times per Treasure Hunter category (TH0, TH1, TH2, TH2+Knife, TH2+Arms, TH2+Knife+Arms). Even 1000 is a small sample size, but it would be sufficient given the fact that 10 different monsters would've been used.

The reason I decided to set up the testing that way was also to see if different drops on different monsters allowed TH to affect them differently, or if we'd find similar results across the board for different levels of TH. For example, when killing bees, granted only 258 and 200 kills respectively, TH0 -> TH1 jumped from a 16.28% drop rate on Beehive Chips to a 36.5% drop rate; over double. Some drops can't very well be doubled due to already being higher than or equal to 50%, and a static increase of ~20% certain isn't possible as evidenced by the many times we've all killed Qiqirn Treasure Hunter without drops.

[Cautery]

What I conclude -
- This game is fucking random!

Fixed

[Wizerd]

What I conclude -
- This game is fucking random!

Fixed

Random or not, being able to assess whether or not one of your members can come a job other than THF and output more damage or help with survival serves a very important purpose in this game.

Even the most cynical players like to increase their chances of something to the highest possible.

[Octavious]

I think the tests were pretty interesting. Perhaps if you select a mob with a lesser item loot pool we could determine it more accurately? What would be full drops on a Bee..like 3 Chips,Wing,Stinger? If there's a mob with a smaller pool of possibly 3 drops you could test THs influence on it and maybe get a better idea idk.

Other then that the graph & data seem pretty solid for the test size. A larger test size would be very interesting to say the least. Or possibly a showing of TH 0-4 on certain mobs as was previously mentioned. Overall props for the effort & work it seems well put together. ^o^

[Wizerd]

I think the tests were pretty interesting. Perhaps if you select a mob with a lesser item loot pool we could determine it more accurately? What would be full drops on a Bee..like 3 Chips,Wing,Stinger? If there's a mob with a smaller pool of possibly 3 drops you could test THs influence on it and maybe get a better idea idk.

Other then that the graph & data seem pretty solid for the test size. A larger test size would be very interesting to say the least. Or possibly a showing of TH 0-4 on certain mobs as was previously mentioned. Overall props for the effort & work it seems well put together. ^o^

A full drop is 1 chip, 1 honey, 1 wing, 1 stinger.

[MaachaQ]

The only TH question I still think needs to be answered is, does the TH4 Thief need to make his 1st hit on the mob while wearing full TH4, or can he, say, do a WS with higher damage dagger and gear, then swap to TH gloves and knife and do 1 hit to set TH4 active on the mob? Feels really gimpy having to TP for gods in THF Knife when there are several higher damage options available, just to be able to set TH4 on it properly...