Talk:Drop rate/Gift of the Traveler

This page doesn't have the 10 Hunter's Ales (not aged) drop. I'd add it myself, but I have no idea how. --Totally Imbalanced 00:28, 27 April 2009 (UTC)
 * I believe it's not there because none of the people listing their drops have gotten it yet. Technically there should also be Vampiric Dragon Sword, etc. -Phazor 00:32, 27 April 2009 (UTC)
 * Well, I got the Ales, and someone will eventually get the VDS. The Mini Gwen Doll, Mini Professor, Mini Brown Rabbit, 5 Crates of Fireworks, Everlasting Cottontail, Traveler's Bo Staff, Ascalon Razor, and Icy Dragon Sword also need to be added.  Again, I'd do it but have no idea how. --Totally Imbalanced 00:36, 27 April 2009 (UTC)
 * I'll see if I can figure it out when I have more time. -Phazor 02:24, 27 April 2009 (UTC)
 * Alright, added those that you listed. If you think of or see more, just tell me.  Basically to add the columns I just followed the template that had already been set. -Phazor 03:52, 27 April 2009 (UTC)
 * Oh, Totally Imbalanced. You forgot one of the gifts that you got (you put 10 as the total, but you only put 9 gifts).  I don't know if you remember what that last gift was so that we can get it back to a multiple of 5. -Phazor 01:25, 28 April 2009 (UTC)
 * My bad. Fix'd and updated totals. --Totally Imbalanced 01:44, 28 April 2009 (UTC)

Just post or pre also?

 * is this just post data or do we want pre also here? NCA 23:05, 26 April 2009 (EST)
 * It's just post. You can start one for Pre, though. [[Image:Sunsmoonsig.jpg]] talk & cont  07:04, 27 April 2009 (UTC)
 * Have done so, at Gift of the Huntsman NCA 17:42, 27 April 2009 (EST)

Tempted...
I'm tempted to make a note at the bottom saying to please update the totals and percentages whenever you add your drops/gifts to the table. I've had to correct it three times so far, and it will probably continue. A note might reduce some of that. I won't do anything until given permission, though. -Phazor 03:41, 27 April 2009 (UTC)
 * I personally wasn't certain as to who is doing the updating and such, so left it alone. A note would help a lot. NCA 1:51, 27 April 2009 (EST)
 * Okay, I added the note at the top. If anyone finds the wording to be confusing, please change it so that it is not. -Phazor 23:02, 27 April 2009 (UTC)

My data
I suck at editing tables and really don't know how to go about doing it. Here's what I got from five packages: 5 mysterious summoning stones 5 mysterious tonics 5 cottontail tonics 10 champagne poppers 5 green rock candies
 * Can't do anything with that if you don't sign it. -Phazor 04:02, 28 April 2009 (UTC)

I looked at entering the data and it looked pretty tricky. Would someone be willing to enter? Here's my data: 5 blue rock candies, 10 bottle rockets, 5 crates of fireworks, 5 crates again, 10 jars of honey Truemyths 04:43, 28 April 2009 (UTC)

I added my data, but I'm not 100% sure how to calculate the percentages at the bottom. -Zyanya386 09:16, 30 April 2009 (UTC)
 * do you have a spreadsheet program? I use OpenOffice, but Microsoft Excel should work just as fine. You can just copy and paste the whole table (without sum line) in there, then calculate the new totals and percentages with relative ease. You'll have to copy&paste those values back one by one, but it's still faster than counting it all yourself :) Ailar 16:23, 7 May 2009 (UTC)

Sorry, Also no good at wiki code. 20 Gifts as follows:


 * 4 Mysterious Tonics (20)
 * 3 Mysterious Summoning Stones (15)
 * 2 Champagne Poppers (20)
 * 2 Keg of Aged Hunter's Ale (2)
 * 1 Bottle Rockets (10)
 * 1 Hunter's Ales (10)
 * 1 Jars of Honey (10)
 * 1 Sparklers (10)
 * 1 Aged Hunter's Ale (10)
 * 1 Blue Rock Candies (5)
 * 1 Red Rock Candies (5)
 * 1 Mercantile Summoning Stones (5)
 * 1 Cottontail Tonics (5)

Also planning on buying 200 gifts soon (1 million gold worth @ 5k each) and will post findings once that is done. Thizzzguy 6:30AM PT, 13 July 2009

219 gifts


 * 34 Mysterious Tonics (170)
 * 20 Mysterious Summoning Stones (100)
 * 17 Sparklers (170)
 * 14 Red Rock Candies (70)
 * 14 Green Rock Candies(70)
 * 13 Bottle Rockets (130)
 * 12 Champagne Poppers (120)
 * 12 Cottontail Tonics (60)
 * 11 Honeycombs (110)
 * 11 Mercantile Summoning Stones (55)
 * 11 Jars of Honey (110)
 * 9 Blue Rock Candies (45)
 * 9 Aged Hunters Ale (90)
 * 9 Keg of Aged Hunters Ale (9)
 * 8 Hunters Ale (80)
 * 8 Disco Balls (24)
 * 6 Crates of Fireworks (30)
 * 1 Everlasting Crate of Fireworks (1)

Thizzzguy 14 July 2009 23:36 UTC

Vertical text
The text in top row that say what item each colomn is for, can be vertically written using style="writing-mode:tb-rl; white-space: nowrap;" in the row or in each cell. I'm going to try this on w:c:fr.guildwars:Cadeau du Voyageur/taux de drop... — TulipVorlax


 * Ok, after testing, it need to be putted on each cell :


 * — TulipVorlax 02:23, 28 April 2009 (UTC)

Odds/expecations based drop data
(As of about 15 minutes before this post), there were 1006 entries (on the official site§). And the percentages of drops based on type are as follows:


 * {| class="stdt"

! Type !! %
 * Drunk || 15.71%
 * Mini || 0.40%
 * Other || 0.20%
 * Party || 20.08%
 * Summon Stone || 34.79%
 * Sweet || 23.56%
 * Tonic || 3.58%
 * Weapon || 1.69%
 * }
 * §I used the official wiki's date for three reasons: (a) they had more data, (b) I computed the avgs on my own, (c) I was too lazy to recalculate using the data here.
 * Sweet || 23.56%
 * Tonic || 3.58%
 * Weapon || 1.69%
 * }
 * §I used the official wiki's date for three reasons: (a) they had more data, (b) I computed the avgs on my own, (c) I was too lazy to recalculate using the data here.
 * Weapon || 1.69%
 * }
 * §I used the official wiki's date for three reasons: (a) they had more data, (b) I computed the avgs on my own, (c) I was too lazy to recalculate using the data here.

&mdash; Tennessee Ernie Ford ( TEF ) 18:55, 2 May 2009 (UTC)

(Reset indent) (As of about 15 minutes before this post), there 3,424 entries (on the official site§).


 * §I used the official wiki's date for three reasons: (a) they had more data, (b) I computed the avgs on my own, (c) I was too lazy to recalculate using the data here.
 * §§The expected value is based on each gift drop's merch value, gwauctions typical price, or average WTB/WTS prices as seen in the game.
 * &mdash; Tennessee Ernie Ford ( TEF ) 16:30, 2 July 2009 (UTC)

mid September 2009 Odds/Expectations
Based on 5051 drops from the official wiki and auction-site/in-game valuations, the expected value of the gifts seems to have dropped to around 4063 (the actual number is probably lower, since it's hard to price the rare items, as very few of them are traded). My opinion of their value: if like to buy lottery tickets, you might consider paying as much as 8k for a GotT; if you are buying for title points and/or to resell contents, consider paying no more than 3k/gift. (And, clearly, at the moment, you can make considerable profit buying at 4k and reselling at 5k, the low/high of the current trade prices)  &mdash; Tennessee Ernie Ford ( TEF ) 17:13, September 12, 2009 (UTC)

Invalid rows
Currently there are 340 Gifts entered, but the sums of the column totals add up to 346. Originally, there were 6 invalid rows in there, where the sum of the entries was higher than the value in the first column (one of them is off by 2). Somewhat embarassingly, one of the entries was mine :S I've corrected that one, but as I don't know if we'll be able to correct the others, should those rows just be deleted? They're the rows which are signed as follows:
 * --NCA On behalf of a guildie 17:57, 27 April 2009 (EST)
 * -- Zyanya386 09:10, 30 April 2009 (UTC)
 * --GW-Topinambour 15:57, 1 May 2009 (UTC)
 * Yujiko Ineluki 16:20, 2 May 2009 (UTC)
 * Viruzzz 08:39, 3 May 2009 (UTC)


 * I would think that we should separate them from the grand total (either new table or separate section) until/unless the users re-confirm their counts. Doing so will have low impact on total numbers but high impact on the credibility. What do other people think?  &mdash; Tennessee Ernie Ford ( TEF ) 18:38, 7 May 2009 (UTC)


 * Good idea (I'm somewhat late to reply, because I really couldn't be bothered to farm Mandragor Roots last week...). So I started the new section now - currently there's only one faulty row though, looks like someone cleaned up the rest -- Ailar 19:01, 18 May 2009 (UTC)

More confusion
I've been visiting the wiki for going on three years now, yet I am still a complete wikinoob and have no idea how to add to this chart without completely screwing up the data and making it all look like gobbledygook. (On a side note, the Firefox spellcheck apparently liked that word... but not the word "spellcheck"). Anyways, last week, on Sunday June 07, I got: 5 merch stones, 10 bottle rockets, 10 champagne poppers, 5 mysterious summoning stones, and... *drumroll please* an Inscribable IDS! That would be the first to be added to the chart, so I thought it was worth trying to figure out how to put it in. Thanks to anyone that can do this for me. :) 72.68.45.67 22:30, 9 June 2009 (UTC)

Standard Deviation
I've added a standard deviation row to the end of the table, which (if my math is correct) should help to give us a fair measure of how accurate the numbers at the bottom are. If my understating of statistics is correct, then one can expect that ~95% of the time, the actual drop rate will be plus or minus 2 standard deviations from the observed average rate. My understanding of statistics however may be way off base. How I derived the standard deviation: I produce 3 figures for each item, and use them to then derive the fourth:
 * S0 --Sum of packages opened (733 in all cases)
 * S1 -- Sum of instances item was encountered (for merc stone: 42) (Technically, this is rate encountered * packages opened, but bits cancel out leaving this as simply the count of item encountered
 * S2 -- Sum of squares of encoutner rate Sum of: %encounter rate squared for row times number of packages opened for the row.
 * Standard Deviation = (S0*S2-S1*S1)/(S0*(S0-1))
 * Derive square of deviation from average for each entry, multiply bythe weight for the entry (weight = (number of packages^0.5*total entries/total packages) --Eg, if a row had 2 kegs from 10 packages (20%), and the average is 4%, this results in ((4%-20%)^2) * 79/773 -> (-16%^2)*(10^0.5*79/773) -> 2.56% * 3.16.../10.58... ->32.31%
 * Derive the average of those squares for each column (remembering to divide by number of rows minus 1) This is the sample variance.
 * Take the square root of the variance - This is the standard deviation of the sample. (edit -- provided corrected calculation) Yamagawa 18:35, 26 July 2009 (UTC)
 * (One thing this does is weight a given row in the table based on the number of packages that row represents,

I'd appreciate it if someone could vet my math, and if I'm misunderstanding the significance of the deviation, if they could correct me there too. This is fairly shaky ground for me, but heh, I gotta start learning this stuff somewheres.

ALso worth noting, if I recall rightly, that an ideal sample size is in the 15-30 range. Too small and it doesn't give a measure that can be expected to reflect reality. Any larger and the measure's accuracy won't change much by growing it. Yamagawa 05:40, 26 July 2009 (UTC)


 * Bugger, looks like I left out a square-root on my stddev calc... which has the approximate effect of multiplying the values by about 10...  Corrected figures to go in, and it's time I read more on estimating a population from a set of samples... Yamagawa 15:42, 26 July 2009 (UTC)


 * Ok, I've got a 'corrected' standard deviation in place. From reading up on confidence intervals, I find the calculation to be:
 * Confidence interval = Average +/- Standard Deviation/(n^0.5)*(z value for percentile)
 * After some fussing with how I weight my standard deviation calculation for the size of the sample, I am getting some confidence intervals that seem fairly reasonable. Please, someone kick me if I have the math wrong. Yamagawa 18:35, 26 July 2009 (UTC)


 * Updated intervals put into a new table. No corrections to math used, rather I've found that if I pool the results in r rows (where r = n^0.5), I can shrink the confidence intervals.  I'm using the drop rate of mr yakkington (0.14%) plus his confidence interval (0.27) to get a max possible drop(0.41%) rate for items with 0 occurrences reported.  Alternate math suggests that 0.47% may be the correct high-ball estimate for those values (95% confidence), but the two numbers are close enough that I won't worry over a whopping 0.06%  Yamagawa 23:17, 27 July 2009 (UTC)
 * Updated the summary table & error rates. Yamagawa 22:13, October 17, 2010 (UTC)

Excel Code
I put some Excel VBA code on my userpage - First procedure can parse the text from the edit page for Drop Rate Raw Data into Excel where new rows can be added. Second procedure updates totals. Third procedure creates text than can be pasted into (overwriting) the existing text on the Raw Data page. Once I figure out the confidence interval stuff, I will add that. Please send comments/questions/suggestions to me Separ 23:20, 15 August 2009 (UTC)


 * Wow! Fancy stuff :-) Thanks for taking the time to do this (and publish). (For my own purposes, I just copy/paste the data into a spreadsheet that uses formulas to calculate everything. If you are interested, you can take prices from the table above and use them to calculate expected value of the gifts, which might or might not translate into market value).


 * I'm wondering about how you calculate confidence intervals here. Since each gift only produces a single drop, it seems to me that there's only a single function that applies to all entries. Consider the extreme of two possible drops, heads or tails: out of 100 gifts, we see 55 heads and 45 tails. The formula that tells us the 1-sigma interval for heads (55% +/-X) necessarily has to lead to determining the 1-sigma interval for tails (45% +/-Y), since heads can't be 55% unless tails is 45%. The same holds true for the more complex GotTs and GotHs: the chance that we are seeing a reasonable drop rate for Kegs depends on whether we are seeing one for Crates. So, I suspect you should be using binomial distribution functions and (therefore) Chi-squared confidence intervals.  &mdash; Tennessee Ernie Ford ( TEF ) 03:15, 16 August 2009 (UTC)
 * I assume you went to college. Docta Jenkins 08:56, March 24, 2010 (UTC)
 * Apologies for my late reply here... I've been occupied...elsehwere... Yes, I went to college. Never took stat, but I find myself using it more and more in work...  (Gosh golly darn... we tweaked the code here... did the error rate change any? THat's right... F-Test will tell me that...Hmm.. Odds are 85% we made it better)  Mostly did it with the occasional look at a stat book, and lots of digging on Wikipedia .  I've put notes for my actual method here.  The basic idea is as follows:
 * Treat each drop-type independently. This makes the math easier on my poor head.
 * Group the counts int N buckts of approximately N drops each. (There's an average of averages method that stat uses when collections of samples are available, which is what we have here.  I find grouping them into N buckets of N narrows the error margins, as compared to N buckets of M).
 * Weight each bucket according to the total number of drops in it. Weighting used is square root of # of drops.
 * Take an average of the averages (use weighting). Call this the average rate.
 * Take a standard deviation of the values (use weighting). Use this to calculate error margin.
 * Use alternate methods to figure the max likely drop rate for items with 0 drops.
 * Yamagawa 21:59, October 17, 2010 (UTC)

No everlasting fireworks?
it seems there is no fireworks on this list whats up with that? 01:10, March 21, 2010 (UTC)
 * Probably because nobody who records here got one yet, and they didn't know it was a drop when the table was first made. It's probably on the order of the minis or even less. RoseOfKali [[Image:RoseOfKaliSIG.png]] 12:23, March 21, 2010 (UTC)
 * It occurs, that by omitting fireworks from the table, people who get them are biased against posting the finds. (consider the work of inserting a column in the table).  Could we find someone sufficiently bored, or skilled, to add them?  Yamagawa 01:44, October 21, 2010 (UTC)

Observed Rates - Summary
What are all those x6 x7 x11 next to the items? They're obviously not the number of items dropped from a single gift, or most of them would be wrong. Explanation, please? RoseOfKali 10:49, May 6, 2010 (UTC)


 * Thought those were used in the calculation for observed rate, like sample sizes used for each item. The only person to really answer that is Yamagawa tho. Bottle 22:02, May 6, 2010 (UTC)
 * My bad :p the numbers were changed in the Sept 2009 edit for some reason by IP. Changing it back now. Bottle 22:29, May 6, 2010 (UTC)

Weird Gift Opening Trend
Alright. Recently, as in like the last 20 gifts I have opened 4 of the 5 prizes I got were party points. This could happen easily enough, but what makes it even stranger is that I currently have 6.2k party points. This makes me wonder if the gifts give you stuff that need for titles. Thoughts? WGreg 14:50, September 26, 2010 (UTC)


 * Just lucky. If you open, say, 200 in a row, you'll see similar trends. Party points account for 40% of the drops (according to GWW/GWiki stats) → it's not that odd that you have seen 80% for 20 drops. &mdash;Tennessee Ernie Ford ( TEF ) 22:55, October 17, 2010 (UTC)

Late 2010 Opening Trend Observations
Looks like the rough breakdown is:
 * 40% party
 * 25% sweets
 * 20% Summons
 * 13.5% Alcohol
 * 1% Weapons
 * 0.5% Other

(These figures are all +/1 about 1%)

Further (cheating some by looking at external data sources to help fill in obvious blanks): Sweets:
 * When a sweet is encountered, each occurs roughly 20% of the time (+/- 1.5%)
 * I'll sign of on the rates for this as plausibly accurate.

Party: When a party item is encountered:
 * Rockets, Poppers, and Sparklers all occur roughly 12.5% of the time
 * Cottontail Tonics also occur roughly 12.5% of the time. Combined this constitutes
 * 20% of the total drops, or 50% of the party item drops.
 * That leaves Disco Balls, Crates of Fireworks and Mysterious Tonics, which share a roughly 4:1:1 relationship. I would call it 34%/8%/8% as an approximation, which gives tonics a nearly 14% overall drop rate.  Assuming the rate to be AT 14% (which it is near) and working backwards, I get 35% to the tonics, leaving 15% for the disco balls and crates -- 35%/7.5%/7.5%.  Assuming that Disco Balls and Crates share the same rate, this works backwards and forwards nicely, but it's something I'd rather see more data on.

When a Summon item is encountered:
 * This one doesn't fit a neat bucket... looks like 27.5% merchant, +/- 2% (so, not quite 1 in 4, not quite 3 in 10), the remainder going to Mysterious Stones (72.5% +/-2%). Looking at them as a fraction of the whole tends to encourage them landing exactly there: 5.5% and 14.5% -- if a-net took 100.0% and divvied it up...
 * Given the odd percentages thing going on here, I'm not going to sign off on this.

When Alcohol is encountered:
 * The keg is encountered 25% of the time
 * The Aged and Regular Ale split the remainder just about evenly (38.3% vs 36.8% -> they lie within each others margin of error, which is +/- 2%).

When a weapon is encountered:
 * More data needed. They roughly fit the 20% each, but the error margin is at about +/- 8%, so there is plenty of room for something in the fine print here.

When Other is encountered:
 * For anything else(mini, or moss egg, or perma-use items)
 * More data is needed.
 * I lump the perma use items here because 1: They are very rare, and 2: They don't count for titles.
 * Moss Spider Eggs occur roughly half the time (+/- 15%)
 * The remainder all occur in the range of 2%-25%, but each appear to be near 10% (and there are 5 items...)

So: A best guess at true drop rates: Yamagawa 21:18, November 21, 2010 (UTC)

Would there be any objection to posting all but the uncertain rates on the main gift page as 'approximate drop rates'? Yamagawa 16:25, 1 December 2010 (UTC)