User talk:Yiab

From TheKolWiki
Jump to: navigation, search

You're doing a great job there on an area of KoL that I've never messed with much. Keep up the good work! --jin 23:33, 9 May 2005 (Central Daylight Time

Thanks very much. I'll likely put in a few more hours on stuff again tomorrow. I just love completing databases for some reason. -Yiab

You've got entirely too much time on your hands! Not that I'm complaining.  :) Over half of what was done on the Wiki today has YOUR name by it. Keep it up! --jin 23:17, 15 May 2005 (Central Daylight Time)

It's "Fettucini Inconnu", not "Fettuccini Inconnu".. Unless you did that on purpose and I'm a n00b :/ --Woemwoem 04:42, 4 September 2006 (CDT)

template question

Do you know how to make a template? I was trying to make a new page, but not sure how to create a template to use... --jin 11:39, 10 Jun 2005 (Central Daylight Time)

  • The way that I figured out how to make a template was to look at the Template:adventure page and construct it from there. it's fairly simple to figure out, once you can actually get to one of the basic template pages.--Yiab 11:45, 10 Jun 2005 (Central Daylight Time)

Excellent work

Would you be interested in having admin privs for the wiki? You've put insane hours of work into it with lots of improvements overall. I'm also losing steam on it with RL picking up again. I'm in need of extra assistance in just keeping an eye on things. Let me know if you're interested in helping out in the context of admin. --jin 23:26, 20 Jun 2005 (Central Daylight Time)

  • Okay, I've given you privs. It's not much different from 'normal', but you have deletion privs and can edit the protected pages (like the main page). I just need some back-ups, because RL tends to keep me too busy lately to pay as much attention to this wiki as it's now demanding. Just keep an eye on things, maintain 'conduct' standards, etc... And be a name on the admin list.  :) Welcome to the growing admin 'team'... --jin 20:02, 21 Jun 2005 (Central Daylight Time)

Patrolled edits

Thought you might want to check this out. --Alpaca 20:16, 27 November 2005 (Central Standard Time)

Conditional Templates

Good luck. ;-) You're not the first person to try and dabble in them. Here are some, but they don't really show any examples of them in use. --JRSiebz (|§|) 21:57, 12 December 2005 (Central Standard Time)

Item Statistics

Yiab, your statistics info is awesome. I feel bad for causing you so much extra work!--Snickles 18:58, 21 January 2006 (Central Standard Time)

  • Thank you. The extra work isn't a problem, it only looks like a lot because I'm putting up a huge batch of it at once right now. This is exactly the sort of reason for putting these things up incrementally; now I can modify my record-keeping slightly to incorporate the date the record was created, ensuring better long-term reliability.--Yiab 19:51, 21 January 2006 (Central Standard Time)

Collecting Item Statistics

Yiab, I just wanted to know what you use to help you collect statistics. I've been (slowly) collecting some by hand, but I get the impression that you have some nice way of automating things. I use KoLMafia from time to time, but as far as I can tell, there's no way to know which encounters you're getting. Let me know; I'd love to help! -- TheArchivist 19:53, 18 February 2006 (Central Standard Time)

  • I do indeed use KoLmafia, but I had to modify the program myself to get it to record things. I coded it to record details of each adventure into a Microsoft Access database, then I wrote a separate little program to analyze the data I recorded. If you like, I could give you a little help in figuring out how to automate the recording, but basically only if you've already got some java programming familiarity.--Yiab 16:39, 19 February 2006 (Central Standard Time)

Area Statistics

The Parseable Area Statistics values for the degrassi knoll are inconsistent -- they add to 109.3%. Is one of the encounters double-counted?--DirkDiggler 20:38, 10 October 2006 (CDT)

  • Thank you very much for pointing this out. There are 9 combat encounters, 8 of which are equally likely and the 9th is 1.5 times as likely as one of the others. Apparently, I used 2/21 and 3/21 as the probabilities when i should have been using 2/19 and 3/19. I have recalculated the chances and they add up correctly now.--Yiab 21:51, 12 October 2006 (CDT)


Hi, the Haunted Patry drops in Parseable Area Statistics look corrupted; see Talk:Parseable Area Statistics. Any thoughts? Thanks! --DirkDiggler 13:49, 31 January 2007 (CST)

  • Yup, they were in error. They should be fixed now. I'm really glad that people like you are checking my numbers or we'd have a lot of errors wandering around.--Yiab 15:02, 31 January 2007 (CST)

Haunted Bathroom combat frequency

The combat frequency on The Haunted Bathroom is clearly wrong, but I don't want to make the obvious fix without checking with you. Can you look at your notes and see if you missed a (100% - X) step? --Jonrock 12:05, 13 October 2006 (CDT)

  • You're right, I did accidentally put in the noncombat frequency instead of the combat frequency. Thankfully, the error did not make its way into the Parseable Area Statistics page, which has correct numbers. Thank you very much for pointing this out.--Yiab 14:49, 13 October 2006 (CDT)

The Road to the White Citadel

I was wondering if your work so far on The Road to the White Citadel allows you to to determine whether or not the massive bag of catnip and the hang glider are one time drops. Thanks! --Gymnosophist 16:25, 21 October 2006 (CDT)

New Trophies

Should I be at all bothered that you apparently have 5 level 30 characters? I guess that would explain how you determine item drop data...--SomeStranger (t|c) 19:19, 26 October 2006 (CDT)

  • I would have thought that it was obvious I'm using a small array of multis to determine item drop data. I assure you, none of them are abused unless you count deciding where they adventure to be abuse. Also, the sixth new trophy is up now, too.--Yiab 19:24, 26 October 2006 (CDT)
    • It was more of a "OMG Wow that is a lot!" than anything else...I guess I figured you kept ascending them or something....--SomeStranger (t|c) 19:28, 26 October 2006 (CDT)
      • I do ascend them, but if I tried to ascend them all at once there wouldn't be enough hours in a day to play all their adventures unless they were all oxycore (in which case stat days would be impossible). I do a bunch of consecutive ascensions on one character before moving to the next, which is why I'm so slow on collecting the Doc Galaktik's quest item drops. Sometimes human limitations just have to win out.--Yiab 19:39, 26 October 2006 (CDT)

7-Foot Dwarf Drops

Yiab, after reading the notes (yours and Gelthrexians) on the 7-Foot Dwarf drops, I'm also confused. I'm not sure what to make of Gelthrexians comment of "I encountered an elusive triple-drop from a 7- foot Dwarf, which, after talking with Yiab (who has done far more research in this area than I have), suggests that there is likely only one 7-foot Dwarf - especially because its combat messages appear to be consistent." Based on ingame fight text, there are definitely two different 7-Foot Dwarves, each with their own separate and distinct combat messages. This whole thing caught my attention because I got a mattock from the 7-Foot Dwarf (Royale), who is only supposed to have a helmet and pants. Is it possible that there are two dwarves, each of which drop all three outfit items? --Gymnosophist 04:27, 28 October 2006 (CDT)

  • It certainly is possible that there are two dwarves with identical drops. My research would not be able to distinguish this from a single 7-foot dwarf since KoLmafia can't tell the difference between identically named monsters. In fact, given that the 7-foot dwarves were likely part of the game before Jick developed the ability to give different encounter chances for different monsters in the same area, he may have created two simply in order to make the dwarf twice as likely to encounter as the dwarf foreman (which it is).--Yiab 12:38, 28 October 2006 (CDT)
  • Thanks for the feedback. I guess what I'll do is add the mattock to the 7-Foot Dwarf (Royale) with an assumed 3.6% chance. I'll also put up a needs work notice on the mines asking for dwarf drops confirmation. Thanks again. --Gymnosophist 03:45, 29 October 2006 (CST)

Haunted Ballroom

I have a request of you. I've been poking at the Haunted Ballroom, specifically the Strung-Up Quartet adventure. I believe it may affect adventure rates OUTSIDE of the Ballroom, perhaps even outside of the Manor. It might just be the RNG screwing with me, but I lack the resources to get a decent sampling. You seem to be good at this kind of thing - can you help out? --xcorvis 11:24, 13 November 2006 (CST)

  • That is certainly an interesting possibility and it is something I can check out. However, since I would like to get the item drops for new stuff first, it will likely be a week or two before I can put significant work into it.--Yiab 12:50, 13 November 2006 (CST)
    • Whenever you get a chance. Thanks! --xcorvis 13:37, 13 November 2006 (CST)

Airship and Adventure Queue interactions

  • In this HCO forum thread there is a lot of spading on the airship -- done by Hellion and Swann_88 and some other very careful peeps -- which seems to argue strongly that the ISORE encounter happens more frequently than the other airship combats. This differs from the info in [Parseable Area Statistics]. Would you mind checking your data to see if there's a reason for this discrepancy?

Also, I'd like to ask whether you attempt to control for the adventure queue in zones having more than 5 encounters and unequal encounter probabilities. Since the more common adventure will more commonly be in the queue, it will be suppressed more often and appear at a lower long-term rate than its actual immediate frequency. I'm doing calculations on the total future rate for encounters given the queue state and so I'm interested in whether the area calculations are aggregate or if they factor in the queue in any way. --DirkDiggler 01:45, 4 January 2007 (CST)

  • Rechecking my old data, it appears that the Irritating Series of Random Encounters is approximately 1.5x as likely as each other combat encounter. I would not be surprised if my data on this was innacurate or out of date, though, since the major recording I did was almost a year ago and my more recent run there was only to get the utensil drop rate, so I may have deleted some old data (to save space). Chances are, though, the ISORE is 1.5x as likely and I wasn't aware that this was a possibility when I first wrote up Parseable Area Statistics (this was one of the first areas to go in there). Does the 1.5x encounter rate agree with HCO's data better?

Regarding your second query, I do not take the adventure queue into account in my recordings, for three reasons. 1) My original purpose was to collect item drop rates. Encounter rates were an incidental bonus, so I haven't put much thought into their calculations. 2) Over the course of many thousand adventures spent, I assume that the queue's effect will be relatively close to zero. This assumption is based on nothing more than my intuition, though; I could easily be wrong as I haven't done the calculations. 3) There are two ways I can think of to include the queue in my calculations: First is to negate its effects, which would essentially require adventuring in six different areas in a cycle so that the queue would never matter, second is to factor in its effects by calculation which would involve counting each sublist of six encounters as an "encounter with queue history" and counting each variety of those (which would effectively turn an area with 5 "encounters" into an area with 15,625 "encounters with histories"), which is a problem I do not think I would enjoy tackling. Since I began these recordings before the queue's existence was generally known (at least, before it was known to me) I haven't been adventuring cyclically and I have nowhere near the size of data set necessary to factor the queue completely into my calculations. If, however, you can suggest a relatively simple way to account for the queue's effects, I would be happy to attempt recalculations.--Yiab 02:54, 4 January 2007 (CST)

  • I think that point #2 is true if all combats in a zone have equal probability. However, if one adventure has higher immediate odds, it will more often be in the queue and will more often be suppressed. It will still appear more frequently but its apparent rate will revert to the average rate in that zone. This will, I think, be a small effect that will be worse in a zone with many different encounters: so, negligible in kitchen, worst in the Hole in the Sky.

I can't think of a good way to easily extract this effect post-facto without doing "encounters with histories," a problem I agree nobody would enjoy tackling, especially not someone who is already doing as much good work as you are :-). I'll see if I can figure out how to estimate the effect on the Hole, to see how bad it actually is. Cycling among 6 areas would certainly fix the problem, how hard would that be to add? --DirkDiggler 21:38, 4 January 2007 (CST)

  • This problem is not too dissimilar to what Johnthescavenger did for NPZR farming...just backwards. Combinatorics is what he called it. So, in theory, we could calculate that an adventure's apparent frequency (affected by queue) to be X% of an adventure's base frequency (not affected by queue); it would just require some math. It may not be X%, it could be something as crazy as (apparent)/log(apparent)=(base), but what I'm saying is that it's figure-out-able (solvable, methinks...maybe).--Dehstil (t|c) 22:21, 4 January 2007 (CST)
  • It seems to me that using combinatorial techniques would be rather complicated for this, but it could certainly get to the answer. Really, if someone does the calculations, it can be useful for about half of the existing data and any future data, but a lot of my old data wasn't dated so we wouldn't know what data was taken during the period of time when the queue wasn't working, or back when the queue was 3 adventures instead of 5. I might even do the calculations myself if I'm bored enough to at some point, but if someone else does them I'd be happy to apply them to my data.--Yiab 15:46, 9 January 2007 (CST)

empty suit of armor

I believe that Jick mentioned that he had upped the drop rates of these items after he implemented the breaking feature of these items. Therefore the percentages that you currently have listed (6%) for all would be incorrect and should therefore be respaded when you get a chance. Thanks.--SomeStranger (t|c) 08:54, 14 January 2007 (CST)

  • Thank you for pointing this out, I may not have noticed this without it being pointed out to me. I should be on it in two or three days.--Yiab 11:28, 14 January 2007 (CST)

Palindome

I see the palindome is on your to-do list... I'd wait a bit on this, as there are rumblings of a revamp in the works.--DirkDiggler 12:56, 22 January 2007 (CST)

  • Indeed, thanks for pointing this out. I was simply going to check if the combat encounter rate changed with Torso Awaregness and the new shirt drop, but the apparent upcoming changes give good reason to wait.--Yiab 09:30, 23 January 2007 (CST)

Multiple Dropper Rates

Hi, I was wondering if you could briefly describe how you're breaking out drop rates for things like the Hole in the Sky monsters. I suspect they may have the same problem as the pixel monsters -- the first star drops at a much higher rate than the second and higher again than the third. You mentioned in Talk:Statistics Notes that you were reworking your methods here -- did you ever get anywhere? I'm willing to help on the math with this if you like.--DirkDiggler 06:27, 13 February 2007 (CST)

  • What I did to disentangle the pixel monster drop rates (as discussed at Talk:The Inexplicable Door) was to separate the data provided (which is organized as "1 or more, 2 or more, 3") into "exactly 1, exactly 2, exactly 3" and then fiddle with the base rates until I found something that matched the observed values reasonably well. Fortunately for me, my fiddling hit upon a good match very quickly. In doing the same with the stars for the The Family Jewels, it is quite evident that the distribution of 1/2/3 stars dropped does not correspond to 30%/30%/30%, but I haven't hit upon any other drop combination that matches up at all decently. (I'll post the processed data that I used on that Family Jewels talk page....) --Hellion 16:45, 14 February 2007 (CST)
  • I've anecdotally suspected the Hole drop rates for some time, and after I found your comments on the pixel data the other week I checked this, and also came to doubt the 30%/30%/30% among others. I suspect that I can set this up as a matrix optimization equation but don't want to go nuts with the math if the problem has been solved.

After playing with the numbers a little bit, the problem is that we need to threshold one of the drop rates to get good separation within uncertainty in the outcomes (That is, unless you force one rate to cap out at 100% it's hard to distinguish 50%/30%/10% from 30%/30%/30% within uncertainty). I think that to find good values we need a few thousand adventures in the hole at like +150% or +200% drops. I'm a few days from being able to take this data myself, but I'd be glad to help in the analysis if either of you can get some super-high-drop-rate figures, and will be able to toss in some of my own once my current project concludes. --DirkDiggler 20:11, 14 February 2007 (CST)

  • This (forcing one of the drop rates to 100%) is exactly what I've been working towards in the last little while, unfortunately the death of my motherboard has kept me away from paying much attention to KoL for the last little while. I'm mostly back in the game now, though, and once I'm done with the new Palindome I intend on trying to work the hole in the sky and 8-bit realm. The main problem I can see right now in my approach is consistency - I use Jekyllin hide belts to get high drop rates and those vary with moonlight, so it might be difficult to orchestrate a uniform search method, but I'll come up with something. Of course, if work is already being done I'd be happy to integrate into an existing project to help out.--Yiab 10:09, 26 February 2007 (CST)
  • I've been gathering data at drop rates between +132 and +187 percent. I have 1500ish adventures worth so far, I haven't done any analysis on it -- we should coordinate our data though. It will be hard for me to go above 220ish% as my spade multi has few skills, so I'd be grateful if you could provide data using Jekyllins.

My idea is to plot drop rates at different buffs and look for a "knee" in the drop rate, showing where one of the items has capped out. (The items dropped vs drop rate will change slope as each item caps out). I have done a little bit of thinking on a direct method for extracting raw rates from observed results: its a matrix optimization problem, I think, and I need to go learn a little before I can actually fit to the numbers. You can kmail me in-game and I'll give you my email and a landing place to transfer data if needed. --DirkDiggler 19:26, 4 March 2007 (CST)

  • When I can get to this (probably in the next day or two) I'll be recording the drops adventure by adventure, rather than looking at the overall number of items dropped in a day. Looking for a cutoff in drop rate velocity is a good idea, but I don't really see how it'd get accurate numbers and I don't see it working at all except on a very long timescale. Chances are, one of the drop rates is over 30% though, so you should be able to find that "knee" within the data you've already got (it shouldn't take more than +200% to max out one of the drops).--Yiab 01:20, 7 March 2007 (CST)

Tiki Necklaces

  • forgive my presumption, but i've been doing a bit of spading and i'd like the opinion of someone who knows something about statistics. would you be prepared to look it over and answer some questions? thanks anyway. --Evilkolbot 15:14, 13 February 2007 (CST)
  • I'm not exactly an expert in statistics, but I have some familiarity and I'd be happy to look over what you've got and answer some questions. I've got AIM, ICQ, MSN, IRC access and of course KoL Chat.--Yiab 10:09, 26 February 2007 (CST)

Item drop data

I've noticed that your item data takes the form of several different measurements. (At various levels of item drop.) What method do you use for combining the measurements? --Starwed 16:39, 29 May 2007 (CDT)

  • I use what I consider to be the obvious method of combining data:

weighted average together for each item drop bonus: [number of item dropped]/([number of monster encountered]*(1+[item drop bonus]))

ignoring anything that dropped 100% of the time (i.e. anything that hit the cap) and using a more complicated formula for items that drop twice from the same monster (and a less effective formula for items that drop 3 times from the same monster). If you have any suggestions for how this could be improved I'd be happy to look at them.--Yiab 01:02, 30 May 2007 (CDT)

Just to be clear, by weighted average you mean weighted by the inverse variance of each measurement? (As described here.) --Starwed 02:39, 30 May 2007 (CDT)
  • No, I did not use inverse variance thingies because we are not dealing with a normal distribution, we are dealing with yes/no selections with a fixed probability. Here's the formula I use:

For a given item drop bonus c

let nc be the number of encounters of the monster at item drop bonus c
and ac be the number of encounters which dropped the item in question

then the approximate chance of the item dropping is: (sum over c)(nc*ac/nc) divided by (sum over c)(nc)

which is equal to (sum over c)(ac) divided by (sum over c)(nc)

If this wiki is capable of displaying LaTeX, I would love it if someone would tell me so I could make this prettier. Also, if this weighted average is wrong, please tell me.--Yiab 23:45, 1 June 2007 (CDT)

I think you must have missed a step somewhere? You don't seem to account for the actual item drop bonus you're running with anywhere in the formulas you state above, which is clearly necessary. Since the results of your calculations make sense, you must account for it somewhere.  :) (Or perhaps I'm just not understanding your notation.)
We are not dealing with a normal distribution
This isn't quite true; although each individual sample is categorical (the item drops or doesn't), we calculate from each set of samples both a measurement of the real drop rate and a measure of how confident we are in our result. See, for instance, wikipedia:Margin of error. If we treat each run at item drop c as a different measurement, then each c will have a different variance associated with it, and to find the average of these separate measurements we could weight by the inverse variance as I proposed above. It's entirely possible the way you're doing it is equivalent or superior, which is why I'd like to see it.
If this wiki is capable of displaying LaTeX, I would love it if someone would tell me so I could make this prettier.
Unfortunately, I'm pretty sure the wiki doesn't have this capability right now.  :( There was discussion about remedying this at Proposed_Standards#Mathematical_Formulae. --Starwed 04:00, 2 June 2007 (CDT)
  • Yes, damn, you're right. The formula I use is actually
(sum over c)(ac/(1+c)) divided by (sum over c)(nc)
While you're correct about the results if we treat the variables that way, you'll notice that the distinction in techniques only affects the calculations of error range, which is something I am not doing. If you or someone else wishes to calculate error bars for these statistics, the necessary data is in the talk pages of each monster.--Yiab 11:23, 2 June 2007 (CDT)
Ok then: your method is exactly equivalent to treating each run as a separate measurement, and then weighting by the sample size of the measurement. But it actually works out that 1000 turns at +200% item drop is less error prone than 1000 turns at +100% item drop. So ideally, when finding the average of all measurements, you give more weight to the +200% run. Using the inverse variance is one way to go about this. (Note that the inverse variance is proportional to the sample size, so your method is pretty much equivalent if your combining multiple runs with the same drop bonus.)
Now, this does effect the calculation of the drop rate, but by a tiny amount. I was mostly curious in how small the errors are on your item data; it looks like they can be made pretty small. Thanks for going through through your methods, and of course for collecting the data in the first place.--Starwed 19:36, 2 June 2007 (CDT)


Okay, excuse me. About the Wossname. Directly under where I wrote, "NO. THE PIRATE GEAR DOES NOTHING." you added another suggestion about the pirate gear. Now, I sincerely hope you were trying to be funny, because the alternative is that you can't read. I'm removing your edit. --Unnatural20 23:15, 15 July 2007 (CDT)

Proposed Changes to Recipe article

I noticed you have been a contributor to Recipe. I've proposed some fundamental changes on how it's connected in the wiki and would like to invite you to discuss them on Talk:Recipe.
Thanks! ColtsScore 16:59, 16 November 2007 (CST)

Balloon multidrops

I'm not following how you're arriving at the drop rate on multidrops. For example, the Green Ops Soldier can drop up to 2 gas balloons. The talk page stats indicate that he drops 1 balloon 20.4% of the time, and 2 1.9% of the time, but the main page data you entered was 16.7% and 2.9% for the two drops. This would give an expected single/double drop rates of 18%/0.5%, in conflict with the data you gathered.

It's a little tricky to infer what the real drop rates. Doing some plots of what is statistically likely, combined with what typical drop rates are, I came up with 3 models:

  1. Both balloons have a drop rate of ~12-13%
  2. One balloon has a drop rate of ~10%, the other ~15%.
  3. There are three possible drops, each at a rate of <=10%. (All 8%, for instance.)

It's actually quite hard to fit a double drop rate to the statistics (the exact observed rates don't even admit a solution; you have to go more than a standard deviation away from both measurements to obtain one.) That means the 3rd scenario is the most likely: a triple drop at ~8% would occur about once every 2000 adventures, and it could easily take about 5000 to see one. --Starwed 07:03, 30 December 2007 (CST)

  • Chances are, you're being slightly confused by one of the weird things in the way I've entered the numbers - "gas balloon x1" means "at least 1 gas balloon dropped", rather than "exactly 1 gas balloon dropped". I realize that this is incredibly confusing, but when I first started doing double-drops I did it that way and to change now would mean going back through every statistics page I've worked on and correcting it (a rather large task).
I'm guessing that this helps, but does not actually bring my numbers close to what they should be. I can see two contributing factors to this
  1. I suck at statistics (anything more complicated than the basics, that is)
  2. I just realized that the spreadsheet I made to calculate double-drop rates had a plus where it should have a minus.
Here's what my spreadsheet has had to do this calculation.
  1. For each item drop bonus, calculate the modified number of double-drops as (number of double-drops)/(drop multiplier)^2
  2. For each item drop bonus, calculate the modified number of single-drops as ((number of single+double-drops)+(number of double-drops))/(drop multiplier) (I now realize that this is incorrect)
  3. Add up the modified double-drops across multipliers and divide by the total number of encounters to get c=(chance of getting exactly 2 drops)
  4. Add up the modified single-drops across multipliers and divide by the total number of encounters to get b=(chance of getting exactly 1 drop) (incorrect since step 2 was incorrect)
  5. Solve the quadratic formula x^2+bx+c to get the chances of each drop independently.
Keeping in mind that step 2 should have a minus instead of a plus, how does this look to you?--Yiab 09:18, 30 December 2007 (CST)
  • Okay, I'm getting myself confused about the quadratic again, so I'll just work it out again here.
Say there are two independent drop rates, p and q. The chance of getting a double-drop will be c=p*q and the chance of getting at least 1 drop will be 1-(1-p)*(1-q) = p+q-pq, which tells us that the chance of getting exactly one drop will be b=p+q=(number of double-drops)+(number of single+double-drops), oddly. Both of these numbers we have, so how can we get p and q back?
p and q will be the roots of the quadratic (x-p)(x-q) = x^2+(-p-q)x+pq, but this is not x^2+bx+c, rather this is x^2-bx+c, the solutions of which will be (b+-sqrt(b^2-4c))/2
So yeah, the addition seems right, correcting for the math errors I made in the first writing of this post. Does this look right?--Yiab 09:59, 30 December 2007 (CST)
  • Time to start beating on an old horse here, but your probability of "exactly one drop" is incorrect. You will see either 0, exactly 1, or exactly two drops. The probability of two drops is c=p*q, as you state. The probability of 0 drops is (1-p)*(1-q). The probability of exactly 1 drop is thus 1 - P(exactly two) - P(0) = p+q-2pq. You can still use that quadratic to solve for p and q, though, since the distribution of the drops is entirely determined by the quantities p*q and p+q. But if that's the formula you're using to find p+q experimentally, then you're going to get incorrect results. Just think about it for a second. If both p and q are, say, .6, then you're claiming the probability of exactly one drop is 1.2! --Flargen 04:16, 22 January 2009 (UTC)

The eq. I worked out is that the two drop rates will be p,q= ((2R2+R1) ± sqrt( [2R2+R1]2-4R2))/2, where Rn is the observed rate of seeing n duplicate drops. AFAICT this is the same as the equation you got to above. ^_^ --Starwed 10:18, 30 December 2007 (CST)

Lab distribution

Someone pointed out at Talk:Parseable Area Statistics that looking at the data on scientists in the lab indicated that the distribution had been changed. (40%/60%, err of 0.5%) This was inferred from the item drop data you collected (Era 3), so I wanted to check that this was a safe inference. --Starwed 09:00, 1 March 2008 (CST)

Sea spading

I noticed that for the item drop data provided for the sea, that you specified your equipment in case it turns out things work differently in the sea. Well, it seems they do for meat drops at least. I haven't gotten around to figuring out if the same sort of thing happens with item drops or not, but I'm suspecting that it does. Specifically: items, effects, and skills that give +meat% bonuses are halved, and those that give penalties are doubled. Apply these modifiers before combining penalties and bonuses together (+0% on the land is not always +0% in the sea, if there's a penalty in your modifiers). Familiars don't seem to be affected. I haven't tried parsing any of my item drop data for the sea zones I have logs for. And I'm not sure how one would go about detecting these things for item drops. For meat it is much easier, since there's a very visible set of integer values. Could you drop yourself to -100% drops under my hypothesis, and see if you get any items, and then slowly increase it maybe? See Talk:Underwater adventuring#Meat drop modifiers or this forum thread. The thread has some actual data in it right now. The talk page right now is just me summarizing the current findings. --Flargen 00:11, 11 January 2009 (UTC)

  • Well, with a slight bit more thought, I'm guessing that the exact same mechanic isn't in effect for item drops. You'd need a LOT of +items to overpower the scuba gear, or even the diving helmet, and people have been using below that and still gotten items. But it does mean it probably needs to be investigated to see if it is partially true: effects only, items other than diving helmet/scuba gear, etc., only applies to item bonuses, not item penalties? --Flargen 01:37, 11 January 2009 (UTC)
    • Okay, turns out I made a silly error in a couple of my spreadsheet formulas. There is still a penalty, but it has been established to be a zone-level penalty that depends upon how "deep" you are in the sea. Essentially, -25% for each "level", with the "level" currently being obvious from just looking at the picture of The Sea. So your item drop rates should be fine. --Flargen 22:57, 11 January 2009 (UTC)
      • So the penalty currently listed in Underwater adventuring for meat drops is accurate? How have you verified the same penalty for item drops?--Yiab 22:44, 12 January 2009 (UTC)
        • The meat drop penalty is correct. This is no known hidden item drop penalty. I believe Jick has often promised he would never do that sort of thing. Apparently it didn't apply to meat drops, though. My concern about hidden item drop penalties was from when I was originally thinking that meat buffs and effects were getting halved or doubled. But that idea was discarded once I corrected my formulas, and Eleron chimed in with his own data and observations. The meat penalties are just a flat penalty. So your numbers for item drops should be fine. It's just meat drop spading that has to take this into account. Sorry for something of a false alarm. Me and my inability to add correctly... --Flargen 22:50, 12 January 2009 (UTC)
          • Jick has stated that having ++meat or ++item will never cause your actual meat or item rate to go down instead of up. Establishing a "base penalty rate" which must be overcome in order to get any drops clearly does not violate that promise. I think that looking into the possibility of a "base penalty rate" for item drops as well as meat drops is a very valid line of inquiry. --Hellion 22:57, 12 January 2009 (UTC)
            • Looks like you're correct on both counts. I now support the theory that there is a hidden item drop penalty in each zone of the sea, and that it is equal to the meat penalty of the zone. See this forum thread, for example. Just skip past the parts where people are discussing the semantics of "negative penalties". --Flargen 10:01, 16 January 2009 (UTC)

Alright, it should be definitively proven that there's a hidden item drop penalty in the sea; well, "definitively" up to some small alpha values of course. See this forum thread. The Briny Deeps has a -25% it seems, and presumably all zones have an item drop penalty equal to the meat drop penalty. --Flargen 08:37, 17 January 2009 (UTC)


Spanish Fly

Have you collected stats on the spanish fly drops? It's not 100% even with the equipped item, but the hippy pages don't include the drop rates. --Club (#66669) (Talk) 21:30, 19 June 2009 (UTC)