neuropsychosocial


quality posts: 171 Private Messages neuropsychosocial

Have anyone else's voting patterns/habits changed since the rule changes were announced?

I hesitate to say this, but... One change that followed from the changes to the Derby rules is the timing of submissions. There used to be significant pressure on the artists to submit by Friday noon or later on Friday at the latest. When the rule changes were announced, several artists immediately noted that this meant they could submit later. We've even seen a Wednesday-at-noon submission print.

However, as a voter, I used to stop by on Friday evenings to peruse the entries. It was part of the ritual of the start of my weekend. Now that entries trickle in over the weekend... If I stop by on Friday to vote, I see fewer designs. Plus, I feel guilty if I vote on Friday because that means I'm disadvantaging the artists who haven't submitted yet, and overall, I think it's a good thing that the artists have more than 24 hours to create their designs. Thursday noon-Friday noon is tough for people with jobs, families, and a desire to sleep.

So now I basically ignore the Derbies. I could have tried to replace my Friday ritual with a Sunday ritual, but it's not the same: Friday Derby voting was a way to unwind at the end of the week. Over the weekend, I'm not thinking about shirts; by Monday, the rat race is in full swing again.

RIP A.A. Blanks (Obituary)

neuropsychosocial


quality posts: 171 Private Messages neuropsychosocial

j5, your chart is excellent (as always). I can't recall seeing the Double-Take Derbies noted on your chart before; I may have missed them in the past, but they helped "anchor" woot-time for me, so thank you for including them.

kwilder wrote:Going back the the OP's seeming purpose of the thread, I remain unconvinced that the change in derby rules is the direct cause of decreased derby votes.

Technically, I think the OP's invective (I hesitate to call it a "point") was that the new Derby rules allow favoritism to play a greater role in selections than under the old rules. I don't need to look at the data to state that's false: if there's favoritism, perhaps one type has been swapped for another, but I suspect the same people are printing in about the same proportion (except Ramy) with about the same proportion of newbie artists.

If we really want to get down into the OP's goal, I suspect he was using a tactic favored by teenagers: yell and scream at the people who aren't giving him what he wants, convinced that will yield the results that he wants.

Dear artists,
If you want to guarantee that you're not picked as an EC, I'd suggest the following three tactics:
- get caught plagiarizing or using clipart/photos/disallowed brushes/filters; bonus points if you're caught more than once by reputable woot artists who post flashing gifs demonstrating how you cheated
- yell at woot
- create sock puppet accounts to yell at woot on your behalf and to yell at other artists on your behalf; bonus points if your yelling creates extra work for the forum moderators
That is the perfect way to guarantee that woot will avoid choosing your designs. Also, if you manage to tick off members of the community who have a policy to deliberately remain neutral towards all artists not named Sylvia - you've probably managed to tick off most of the Derby voters. This isn't going to help you win the Derby.

Love, neuro

no1 wrote:that would imply individuals have more than one account each! shocking! *cough*

How else are the artists' evil twin sisters supposed to get Random Shirts?!?! (My evil twin sister has ordered three sets of Random Shirts and - inexplicably - a set of Lock&Lock. However, she has never voted in the Derby.)

RIP A.A. Blanks (Obituary)

Narfcake


quality posts: 280 Private Messages Narfcake
neuropsychosocial wrote:... Also, if you manage to tick off members of the community who have a policy to deliberately remain neutral towards all artists not named Sylvia - you've probably managed to tick off most of the Derby voters. This isn't going to help you win the Derby.


Thanks for the laugh, dear. Now I have to clean pieces of dinner off my keyboard and screen here ...

j5


quality posts: 63 Private Messages j5
mkdr wrote:The first thing I noticed on this excellent graph was a very obvious change in the rate of drop off in votes right around the beginning of October. It appears to be the derby of Oct 13, which is the one directly after the Shirt as a Costume Redux Derby. The Sunday shirt was highly controversial, (I happened to love it and bought one for all of my family members) and was a low vote getter. Looking back again at the comments I can see how it may have led to a dip in voting the next derby. I certainly don't have the data to do a true analysis, but just thought I'd throw my 2 cents in as well. That is one discussion I vividly remember.

Interesting point. I did notice the dip in October, but didn't immediately correlate any "woot event" with it. So we have that in combination with the full catalog for a pretty significant (apparent) voter apathy effect.

neuropsychosocial wrote:j5, your chart is excellent (as always). I can't recall seeing the Double-Take Derbies noted on your chart before; I may have missed them in the past, but they helped "anchor" woot-time for me, so thank you for including them.

Thanks! DT labels are new. I added them because a DT removes some of the time-based voting uncertainties (as you mentioned above), but there weren't enough of them to matter statistically, other than DTs seem to have lower vote counts than non-DTs.

move along

tjschaeffer


quality posts: 7 Private Messages tjschaeffer

I compiled some data regarding who has print before the rule change and who has printed after, since the charge of favoritism was the OP's original complaint. I've looked only at 2nd and 3rd place votes, since those are the ones that would be a result of "favoritism" and I've removed prints due to double takes.

I also looked at print rates from before the rule change starting with derby 205 (because I didn't have the time or patience to go back any further).

The "leading" artists for post rule change prints are:

(name - #prints - %total)
radscoolian - 4 - 8.7%
walmazan/fishbiscuit5/helgram/ochopika - 3 - 6.5%
dijay/fuacka/Naolito/qetza - 2 - 4.4%

And many of the "preferred" or "favored" artists have only 1 EC print under the new rules:

patrickspens/apelad/kevlar51/odysseyroc/ramyb/rasabi/rglee129 - 1 - 2.2%

I apologize to other artists for pointing out your names, I'm not actually implying that there is favoritism, just using you folks as examples.

And how did these same artists print before the rule change?

ramyb - 26 - 15%
patrickspens - 19 - 11%
walmazan - 13 - 7.5%
radscoolian - 8 4.6%
kevlar51 - 4 - 2.3%
apelad/ochopika/rglee129 - 3 - 1.7%
fishbiscuit5/dijay/rasabi - 1 - 0.6%

Ramyb isn't printing via the EC, but he's also not really submitting a lot. Patrickspens hasn't been printing quite as much either, and everyone else seems to be fairly normalized. So I'm not sure there are any conclusions to draw here other than the fact that, well, maybe there is no favoritism and the editors choose what they feel are the best/most wearable designs.

no1


quality posts: 7 Private Messages no1
tjschaeffer wrote:I compiled some data regarding who has print before the rule change and who has printed after, since the charge of favoritism was the OP's original complaint. I've looked only at 2nd and 3rd place votes, since those are the ones that would be a result of "favoritism" and I've removed prints due to double takes.

I also looked at print rates from before the rule change starting with derby 205 (because I didn't have the time or patience to go back any further).

The "leading" artists for post rule change prints are:

(name - #prints - %total)
radscoolian - 4 - 8.7%
walmazan/fishbiscuit5/helgram/ochopika - 3 - 6.5%
dijay/fuacka/Naolito/qetza - 2 - 4.4%

And many of the "preferred" or "favored" artists have only 1 EC print under the new rules:

patrickspens/apelad/kevlar51/odysseyroc/ramyb/rasabi/rglee129 - 1 - 2.2%

I apologize to other artists for pointing out your names, I'm not actually implying that there is favoritism, just using you folks as examples.

And how did these same artists print before the rule change?

ramyb - 26 - 15%
patrickspens - 19 - 11%
walmazan - 13 - 7.5%
radscoolian - 8 4.6%
kevlar51 - 4 - 2.3%
apelad/ochopika/rglee129 - 3 - 1.7%
fishbiscuit5/dijay/rasabi - 1 - 0.6%

Ramyb isn't printing via the EC, but he's also not really submitting a lot. Patrickspens hasn't been printing quite as much either, and everyone else seems to be fairly normalized. So I'm not sure there are any conclusions to draw here other than the fact that, well, maybe there is no favoritism and the editors choose what they feel are the best/most wearable designs.



I WAS TOLD THERE WOULD BE NO MATH.

youse guyz are way too smart, mine augen are glazing over alla the numbers. but two things bother me:

1. you say there is no favouritism, so imma struggling to reconcile that with the numbers you posted. specifically, if i am reading this rite, you say that radscoolian's % of prints went up from 4.6% to 8.7% - that's nearly double! ochocinco went from 1.7% to 6.5% - that's nearly quadruple! fishbiskit went from 0.6% to 6.5% - that's nearly, uh, eleventuple! those increases sound a lot like woot likes those artists better than the voters do! can you explain, hopefully in words of one syllable or less, why that is not favouritism?

2. the whole idear of comparing print rates before and after the rule change bothers me. so what if fishbiskit's print rate went up 10x, mabe her entries are stronger or the competition is weaker now or whatever. there are too many variables that aren't being controlled in a before/after comparison. imho a fairer comparison would be to compare their actual print rates with the print rates they would have had for the same derbies but applying the old voter-determines-1-2-3 rules. as wotansday noted, many times woot just picks 2nd place voterages anyway, so mabe (for example) fishbiskit would have printed the same designs anyway. if fishbiskit's print rate under the old rules would have been the same as they actually ended up being, then THAT says there is no favouritism.

man, i hope that made sense. i usually don't. //crosses fingers

Mavyn


quality posts: 22 Private Messages Mavyn
no1 wrote:I WAS TOLD THERE WOULD BE NO MATH.

...

man, i hope that made sense. i usually don't. //crosses fingers



That's what it looked like to me as well. Rather than showing NO favoritism, it shows a shift...but still 'more than would be expected' prints to some artists. I don't have a problem with anyone who has printed, so no finger pointing, just trying to understand how the numbers given are being used to say something different.

My speech is not parsing. I am speaking in ellipsis.

no1


quality posts: 7 Private Messages no1
Mavyn wrote:I don't have a problem with anyone who has printed, so no finger pointing, just trying to understand how the numbers given are being used to say something different.



i should have said this too!

tjschaeffer


quality posts: 7 Private Messages tjschaeffer
Mavyn wrote:That's what it looked like to me as well. Rather than showing NO favoritism, it shows a shift...but still 'more than would be expected' prints to some artists. I don't have a problem with anyone who has printed, so no finger pointing, just trying to understand how the numbers given are being used to say something different.



I REFUSE TO EXPLAIN MYSELF!!!!

Actually, there was an error in what I wrote. I forgot to remove the latest double-take info. Radscoolian has 3 EC prints, not 4. Dijay has 1, not 2.

But anyway, I guess my point was that I don't look at the numbers and say "oh man, look at how many prints so and so has, there must be favoritism." I think the percentages are rather misleading since we're looking at a smaller sample size.

Looking at the pre-change numbers shows that there was already a bias by the voters towards certain artists (1/3 of all prints were by 3 people) and the change was put forward to try to even things out a little bit.

The easiest way to put it is this: by unbiasing the derby with the change in rules, the print opportunities were spread to other artists and therefore creates the appearance of "favoritism" because their number of prints has gone up. If you take prints away from a few people, they have to go to somewhere. And the most likely place is to other artists who have proven themselves as talented and in touch with the average woot buyer.

Me no explain meself so goooooood.

grrlmarvel


quality posts: 12 Private Messages grrlmarvel

I think that in some ways the quality of the derby entries has increased... or at least some artists have matured. I think I've seen more new artists of late than I used to, which is exciting to me.

I know the numbers show some people have jumped in their printing and I think it REALLY IS because some of them have really bloomed or have continued to press and gotten better.

But I agree, I would still like to see Woot print some different stuff occasionally. I have always wanted that. I love pop culture, heck my degree is in comics... but I love some of the original art that Woot has had too. In fact some of my fav Woot shirts are just cool art or tell a great story.

no1


quality posts: 7 Private Messages no1
grrlmarvel wrote:I love pop culture, heck my degree is in comics...



this piques my curiosity. history of comics or makin' 'em?

neuropsychosocial


quality posts: 171 Private Messages neuropsychosocial
tjschaeffer wrote:I compiled some data regarding who has print before the rule change and who has printed after, since the charge of favoritism was the OP's original complaint. I've looked only at 2nd and 3rd place votes, since those are the ones that would be a result of "favoritism" and I've removed prints due to double takes.

Thank you! That was really interesting.

One thought: the OP's hypothetical agenda (I think we've identified the underlying agenda...) was along the lines of "the shirt.woot editors and guest editors are displaying favoritism that the voters aren't." Therefore, I think it's worth considering who has won Derbies since the rule change. For example, fishbiscuit has earned three 2nd/3rd place prints since the rule changes; this is disproportionate to her Derby prints before the rule changes. This could look like favoritism. However, over that same time period, fishbiscuit won three Derbies - chosen by the voters - and also won a double-take Derby - chosen by the voters.

23 completed Derbies: four wins, two ECs, one GE. One of those wins has set a new record for consecutive weeks at #1 and is well over 10,000 sales. The GE choice was one of the best selling (maybe the best selling?) GE choice. It was from an extremely strong Derby; the 1st and 2nd choice shirts from that Derby have been in the top ten for 21 weeks!

When I look at fishbiscuit, I see an artist who has improved dramatically; may be devoting more time/effort to entries; has been an extremely pleasant and supportive member of the community; and has been picking up fans via submitting excellent designs and being kind to fellow artists and community members.

Except for Patrick, most of the artists who have been printing from the Derbies tend to post nice comments on other artists' submissions and tend to post "way to go!" messages when a shirt prints. (Patrick is not unpleasant; he's just absent.)

Am I being too subtle? Sometimes I can be too subtle.

(Psychology: not rocket science. Among other reasons, rockets are predictable and people... well, anyway, if you want to win votes and influence people, being nice tends to sway voters more than the opposite. Exhibit G: Howard Dean.)

One other thing: someone pointed to the Costume Derby as the Derby after which votes fell. I believe "blaming" Russian Nesting Doll is inaccurate: Hurricane Sandy hit, like, ten(?) days later and we saw a huge drop in sales of Reckoning shirts that lasted for several weeks. By the time sales of Reckoning shirts picked back up, it was getting close to Thanksgiving. I think the Thanksgiving-Christmas corridor might lend itself to high sales-low votes because people are preoccupied with holidays. The votes for this past Double-Take were higher than the other December Derbies, right? Sorry, I don't have time to check right now... must - catch - plane!

RIP A.A. Blanks (Obituary)

fishbiscuit5


quality posts: 32 Private Messages fishbiscuit5
neuropsychosocial wrote: When I look at fishbiscuit, I see an artist who has improved dramatically; may be devoting more time/effort to entries; has been an extremely pleasant and supportive member of the community; and has been picking up fans via submitting excellent designs and being kind to fellow artists and community members.


Thank you, Neuro, for validating the time and effort I've put in to learn this craft and to become a contributing member of the woot community. I would hope the reason I'm getting printed more is because I'm finally getting better at this because let me tell you, my designs stunk in the beginning! I think further evidence that there's no favoritism here is that I have also been printing on other sites, not exclusively on woot. You should see my spreadsheet filled with rejections! :P

As far as the statistics and numbers go, I'm not sure they can definitively prove anything because there are too many factors to consider not just on the woot side, but also on the customer side. Sometimes life gets in the way of our wooting. It's likely a combination of things that can't all be identified at any given time. I'm also guessing the sales are being split across the different woot sites and woot plus sales.

Now, just for fun and to remind you how truly awful I was, here is my very first t-shirt design and woot submission. Don't laugh too much. I didn't even know how to use Photoshop, but I was convinced I would win!
http://shirt.woot.com/derby/entry/39221/soulmates

ochopika


quality posts: 25 Private Messages ochopika
fishbiscuit5 wrote:Thank you, Neuro, for validating the time and effort I've put in to learn this craft and to become a contributing member of the woot community. I would hope the reason I'm getting printed more is because I'm finally getting better at this because let me tell you, my designs stunk in the beginning!
http://shirt.woot.com/derby/entry/39221/soulmates



Teehee! I mean, that's not funny at all. Thanks for sharing that, it was very encouraging! I'm really happy to see you printing more.
Can I show one of my first designs too? It was made on GIMP without a tablet. I was afraid of technology at the time:
http://shirt.woot.com/derby/entry/51972/simply-anatomy]

The woot community really helped me and other artists improve. I agree with grrlmarvel that a lot of artists are printing because they've gotten better. Also, grrlmarvel needs to make moar derby entries.

orabbit


quality posts: 30 Private Messages orabbit
neuropsychosocial wrote:
Exhibit G: Howard Dean.)



Ha! The scream heard round the world. I forgot about that guy.


neuropsychosocial wrote:
I believe "blaming" Russian Nesting Doll is inaccurate.



Thank goodness. I was starting to think I broke the derby.

orabbit


quality posts: 30 Private Messages orabbit

Not to derail this thread, but thanks for sharing your B-sides FB and Ocho. Here's the first shirt I ever printed, back at the end of '05. I didn't really know about placement, so my copy of the shirt has the "art" centered on my belly. Egad.

fishbiscuit5


quality posts: 32 Private Messages fishbiscuit5
orabbit wrote:Not to derail this thread, but thanks for sharing your B-sides FB and Ocho. Here's the first shirt I ever printed, back at the end of '05. I didn't really know about placement, so my copy of the shirt has the "art" centered on my belly. Egad.



Derail away! I'm not sure this post resolved anything for the OP anyway. I want to see more B-sides from artists!

mkdr


quality posts: 35 Private Messages mkdr
neuropsychosocial wrote:

One other thing: someone pointed to the Costume Derby as the Derby after which votes fell. I believe "blaming" Russian Nesting Doll is inaccurate: Hurricane Sandy hit, like, ten(?) days later and we saw a huge drop in sales of Reckoning shirts that lasted for several weeks. By the time sales of Reckoning shirts picked back up, it was getting close to Thanksgiving. I think the Thanksgiving-Christmas corridor might lend itself to high sales-low votes because people are preoccupied with holidays. The votes for this past Double-Take were higher than the other December Derbies, right? Sorry, I don't have time to check right now... must - catch - plane!



Aaannd...this is why I shouldn't attempt to analyze without the actual data. Correlation does not equal causation, mkdr! Excellent point about the weather and holidays leading to an overall dip in participation.

forever fighting the urge to buy something just because it is cheap and cool...wait, it's cheap and cool, let me click that big yellow button!

j5


quality posts: 63 Private Messages j5
orabbit wrote:Thank goodness. I was starting to think I broke the derby.

some people claim that there's orabbit to blame, but I know...it's nobody's fault.

move along

grrlmarvel


quality posts: 12 Private Messages grrlmarvel
no1 wrote:this piques my curiosity. history of comics or makin' 'em?



Both! But it's the full title is Comic Book Illustration with another degree in Design.

grrlmarvel


quality posts: 12 Private Messages grrlmarvel
fishbiscuit5 wrote:Thank you, Neuro, for validating the time and effort I've put in to learn this craft and to become a contributing member of the woot community. I would hope the reason I'm getting printed more is because I'm finally getting better at this because let me tell you, my designs stunk in the beginning! I think further evidence that there's no favoritism here is that I have also been printing on other sites, not exclusively on woot. You should see my spreadsheet filled with rejections! :P

As far as the statistics and numbers go, I'm not sure they can definitively prove anything because there are too many factors to consider not just on the woot side, but also on the customer side. Sometimes life gets in the way of our wooting. It's likely a combination of things that can't all be identified at any given time. I'm also guessing the sales are being split across the different woot sites and woot plus sales.

Now, just for fun and to remind you how truly awful I was, here is my very first t-shirt design and woot submission. Don't laugh too much. I didn't even know how to use Photoshop, but I was convinced I would win!
http://shirt.woot.com/derby/entry/39221/soulmates



I will admit Fishie is one of the folks I think has grown and that is why her sales have gone up. But she gets mad when I call her out. ;)

She told me never to admit that I know her or she would kill me.

grrlmarvel


quality posts: 12 Private Messages grrlmarvel
ochopika wrote:Teehee! I mean, that's not funny at all. Thanks for sharing that, it was very encouraging! I'm really happy to see you printing more.
Can I show one of my first designs too? It was made on GIMP without a tablet. I was afraid of technology at the time:
http://shirt.woot.com/derby/entry/51972/simply-anatomy]

The woot community really helped me and other artists improve. I agree with grrlmarvel that a lot of artists are printing because they've gotten better. Also, grrlmarvel needs to make moar derby entries.



I need to do MOAR!

Gimp without a tablet? You should get an award for not just quitting after that!! I also think you are one of those artists who have pushed, bloomed and are a great member of the community. I love to see people who succeed by being supportive!

lyonscc


quality posts: 5 Private Messages lyonscc

Well, lloyd finally got me the data over the weekend, but I couldn't analyze it until tonight. It is only data from 2012 and 2013, but that should be enough for an initial analysis.

I'm not sure what he did with his Excel template, but I (like the earlier commenters) couldn't find any real correlations related to vote totals.

However, I did find some interesting things when looking at the sales data (instead of vote data) that would support some of his comments.

In thinking about the data, it seemed to make the most sense to look at first-day sales data, rather than overall sales data (to cut out having to deal with the vagaries of the reckoning, prices, etc.), which was supported by the (somewhat obvious) strong liner correlation between first-day and overall sales data.

Looking at the time-series sales data, there is significantly less erosion of sales than there is with derby votes in 2012-2013. Even so, there is a downward trend in sales during this period.




So first, I did a direct comparison of the different sales types (where "Derby 1" = first place in the derby; and Old/New is the old vs. new determination of derby winners) across the entire year.



From this, we can see that 1st Place in the Derby, both pre- and post- rules change (and 1st Place in the Double-Take) are not significantly different from one another, and result in significantly higher sales than all of shirt sale types.

We can also see that the only Level containing the Guest Editor's Choice ("Derby 3: New") - Level D - results in significantly lower sales than other Levels of sales.

So, there is evidence that the GEC is significantly costing Woot shirt sales.

Now, we know that overall sales decreased over the time period we're looking at, so I ran the Oneway analysis both pre- and post- rule change.


Prior to the rule change, sales of the Derby winner were significantly higher than all other types of shirt sales. All other types of shirt sales were not significantly different from one another. (I'm ignoring the double-take data, which has a high standard deviation because of the small samples size.)


After the rule change, the top level remained similar to pre-rule change (as we would expect), but another level emerged, in which the GEC was only present in the lowest Level.

So, curious to try and compare the Old Derby 2nd and 3rd place to the New Derby EC and GEC, while trying to account for the loss in sales over the year, I decided to look at the percent "loss" in sales between the First place entry, and the other two derby winners.



Interestingly, the 3rd place Derby winner in the old system had a significantly lower drop in sales (relative to the derby winner) than the 2nd place Derby winner in the old system. This is reversed in the new system.

And finally, I used the overall sales numbers (not just the first day sales) to compare the different types of shirt sales, just to see if it made the picture any clearer.



So - with all of the evidence in Lloyd's sample - it appears that sales for the 2nd place winner (the EC) in the new rules are less than in the old system, but not significantly. However, the evidence does seem to show that the Guest Editor Choice is, indeed, costing Woot shirt sales.

Additionally, one of the stated aims of the new voting/selection system was to help newer artists print more often.

One way to look at this (since my data set is incomplete) is to see how many first-time artist prints (by percentage) pre- and post- rule changes.

In 2012-2013, 31 of 194 daily entries (16%) were first-time artists. Prior to the rule changes, 11 of 84 derby winners (13%) were first-time artists. After the rule changes, 5 of 66 (8%) derby winners were first-time artists. (I can't really do a full analysis of the "spreading" of wins without full data, so I can't go much farther here).

As to lloyd's other comment about 2nd and 3rd place finishers (in the votes) in the new system not receiving an HM, this has happened some (which was why he initially asked me about my feelings on it, since he knew that I'd had a design in that category). As it turns out, I guess I'm the unluckiest artist in this regard.

In the new system, 44% of the 2nd and 3rd place finishers (in the voting) still printed as an EC or GEC. 41% of the 2nd and 3rd place finishers received an Honorable Mention. 15% received nothing at all (no print, no HM).

(Personally, I had four 2nd or 3rd place vote finishes during the time period, one of which printed, and the other three were not HM's.)

I can't really say how many of the 2nd/3rd place finisher HM's will print, since not all of them have gone through the double-take yet.

So - while there is some anecdotal evidence for lloyd's comment/suggestion, I'd say there's not enough data to conclude some sort of favoritism (or, in my case, anti-favoritism).

lyonscc


quality posts: 5 Private Messages lyonscc

Since I'm dealing with the data anyway, I decided to do some artist comparisons, just for grins. So please don't hate me (or love me) if the data swings for/against you, just see it for what it is - a statistical analysis of the past 13 months.

First, I ran a sales comparison of all of the "top" artists over the past year (defining "top" artists as ones with 5 or more prints on shirt.woot)





Sales-wise, Ramyb and Patrickspens are only in the top statistical level, suggesting that their sales over-perform when compared to normal sales. Randyotter, on the other hand, is only in the bottom statistical level, suggesting that his sales under-perform.

Unexpectedly, Walmazan is the only other artist not in the top level, suggesting a statistically significant difference between his sales and those of Ramyb and Patrickspens.

So, just to see what other differences there might be, I compared artist sales (for all artists with 2 or more designs) within a set of daily shirts, and then again for derby shirts (for all artists with 2 or more derby winning designs).



For daily entries, the top sales Level includes Gabe Pyle, Vincent Carrozza, Blair Sayer, Robbie Lee, Shirt.Woot, Walmazan, DoOomcat and WinterArtwork. Nathan W Pyle isn't in the top tier groups because he printed enough designs that his standard deviation is tighter, whereas some of the artists with similar average sales have higher standard deviations (most likely due to fewer prints).

Conversely, RandyOtter, Spacesick, Letter-Q and Fishbiscuit5 underperform with daily shirt designs.



With Derby entries, the story is a little different. In the top statistical sales tier, we've got firefrogh, Pixelpants and Rasabi.

Patrickspens has enough prints to drive down his standard deviation, which shows him in the middle sales tiers and not in the lower or upper tiers.

Fishbiscuit5, who didn't do well, sales-wise, in the Daily designs, is near the top of the average sales for Derby winners. (Rasabi is in similar circumstances, sales-wise, with mediocre daily sales and really good derby sales).

Conversely, Walmazan, whose daily prints were in the top sales tier, is the only artist in the lowest Derby sales tier, suggesting that his derby designs consistently underperform in sales.

What does all this mean?

Don't know that yet - just interesting stats.

So one last thing before bed (probably more useful than the prior graphs):

no1


quality posts: 7 Private Messages no1
lyonscc wrote:[massive statistical analysis snipped]

What does all this mean?

Don't know that yet - just interesting stats.





seriously, though, thanks for all the hard work!

lyonscc


quality posts: 5 Private Messages lyonscc
no1 wrote:
seriously, though, thanks for all the hard work!



Thanks :-)

Just as an FYI - my comment "What does all this mean? Don't know that yet - just interesting stats." was in reference to the statistics about the different artists and their sales numbers - NOT about the stats regarding the new voting system resulting in lower sales (which does seem like it may be the case, especially for the GEC).

One hypothesis about the poor sales for Walmazan derby wins, but good sales for Dailies (that could be better supported/rejected if I went back in time and picked up Ramyb and similar artists from the first 220+ derbies):

(EDIT: Reworded to avoid misunderstanding)

Hypothesis: "Walmazan is an incredibly good artist, and is also one of the only artists still posting in derbies with a really big fan base. As such, he has a built-in number of votes, regardless of what he submits. He even has enough of a "fan following" that could influence the GEC's, depending on who the Guest Editor is.

As such, designs that wouldn't fly for other artists could easily make it through the derby and underperform in sales. Meanwhile, the designs he submits as dailies get enough scrutiny from the full Woot Editorial Staff, so his dailies sell better than average."

(And no - this isn't about bashing on Walmazan, it's just trying to explain the discrepancy in what the data shows. He is a great artist and sells lots of shirts, which is good for Woot and good for the community.)

Perhaps there's a better hypothesis out there - it fits the data for the past 13 months, but it could be flawed logic...

fuzzzydunlop


quality posts: 0 Private Messages fuzzzydunlop

1. Open multiple sock puppet accounts
2. Start a thread in the forums to support yourself and trash successful woot artists.
3. ??????
4. Profit

lyonscc


quality posts: 5 Private Messages lyonscc
fuzzzydunlop wrote:1. Open multiple sock puppet accounts
2. Start a thread in the forums to support yourself and trash successful woot artists.
3. ??????
4. Profit


Well there's a logical fallacy for you. You don't like the message, so you blame the messenger.

I'm not sure if this is addressed to me, but I only posted and analyzed data (which I did spot-verify). You can do the exact same thing yourself, and you ought to get the same results.

When I look back through the thread, I don't see any commenters I've not seen before, or newly-created accounts, so your "sock puppet" comment (if I understand it right) makes no sense. I don't think I ever said I agreed with the OP, and actually *disagreed* with its premise about voting, etc.

As for "trashing successful woot artists", I don't see that I "trashed" anyone - I analyzed sales figures and noted the outliers on both ends of the scale. In the case of Walmazan, it was statistically interesting because he's (statistically speaking) one of the consistently high-selling Daily artists, but (statistically speaking) one of the poorer-selling Derby artists. I also don't see where I did any lobbying for myself.

(EDIT: Upon reflection, I edited my post above to make it more clear that I wasn't criticizing the artist, but just trying to explain the data. Sorry if you took offense - none was intended.)

So, if you've got more data I've not considered, send it to me and I'll gladly add it to the set (I'm just too lazy to go get it all manually right now). If you've got a problem with the OP, PM him.

no1


quality posts: 7 Private Messages no1
fuzzzydunlop wrote:1. Open multiple sock puppet accounts



I AM SPARTACUS!

lyonscc


quality posts: 5 Private Messages lyonscc
no1 wrote:I AM SPARTACUS!



Rofl! He had, what? 10,000?

mrwednesday


quality posts: 12 Private Messages mrwednesday
lyonscc wrote:stats and stuff



Interesting stuff. I'm curious how JMP fit the first day sales ~ date you showed at the top. It's showing an ANOVA table leading me to believe data was categorical, but then it also spits out an estimate as if it date was continuous. I'm curious because looking at the scatter plot I can't for the life of me imagine there is a significant trend there. The confidence bands should be way too huge for any kind of significant result (which it is showing). It's very strange.

Anyway even if that trend is real, I don't think there's surprising here. Side sales and the full catalog are going to cannibalize the dailies. Woot doesn't much care, I would think, because the other offerings are more profitable.

From the rest of it the only thing that looks convincing to me (can you confirm if JMP does or does not account for multiple comparisons?) is that the guest editor pick sells worse than expected. I don't think anyone is shocked by this. It's the least likely to have wide appeal due to the quirks of the editor and most likely to be chosen with design quality as a major factor. Neither of those things play well with the average woot crowd.

As for the top artists stuff, I can't say I read it too closely. Outside of a few artists like ramy/patrick/robbie/spirit/ocho, I'm not sure how many would have enough shirts printed to get any kind of real estimate. My worry is that many of mean estimates are based off of a few shirts or simply the one shirt they printed and there are far too many other things to control for (again I can't say, but a number of shirts printed per artist would be extremely helpful to see).

Thanks a bunch for the work! It was really cool to see it all and get an idea of what things kind of look like.

lyonscc


quality posts: 5 Private Messages lyonscc
mrwednesday wrote:Interesting stuff. I'm curious how JMP fit the first day sales ~ date you showed at the top. It's showing an ANOVA table leading me to believe data was categorical, but then it also spits out an estimate as if it date was continuous. I'm curious because looking at the scatter plot I can't for the life of me imagine there is a significant trend there. The confidence bands should be way too huge for any kind of significant result (which it is showing). It's very strange.


I treated the date and sales data as continuous, and worked in JMP to find the best fit model. (Believe me, having done all my stats by hand in school and in Excel during the early 90's, JMP allows SO much more flexibility and speed, it's sick). I didn't think I would find a significant trend, either, when I started.

Anyway even if that trend is real, I don't think there's surprising here. Side sales and the full catalog are going to cannibalize the dailies. Woot doesn't much care, I would think, because the other offerings are more profitable.



I would agree there, as well. My real reason for finding the trend was just to see how much a "Pre-Rules" and "Post-Rules" comparison might be affected by a significant overall trend.

(Also, as you probably know, but most folks reading this might not, the trend-line is a little bit deceptive, because it can only show the trend between the date endpoints. Trying to predict (or fit) past or future data beyond those bounds isn't Hoyle.)

From the rest of it the only thing that looks convincing to me (can you confirm if JMP does or does not account for multiple comparisons?) is that the guest editor pick sells worse than expected. I don't think anyone is shocked by this. It's the least likely to have wide appeal due to the quirks of the editor and most likely to be chosen with design quality as a major factor. Neither of those things play well with the average woot crowd.


JMP *does* do multiple comparisons. I didn't show that table in my screen-caps because it would have made the graphic too large (and I didn't think most folks would appreciate all of the internals).

As for the top artists stuff, I can't say I read it too closely. Outside of a few artists like ramy/patrick/robbie/spirit/ocho, I'm not sure how many would have enough shirts printed to get any kind of real estimate. My worry is that many of mean estimates are based off of a few shirts or simply the one shirt they printed and there are far too many other things to control for (again I can't say, but a number of shirts printed per artist would be extremely helpful to see).


I'll see if I can get the internals to show up with the graph. I agree with the means being "squishy" (which I think does get reflected in some of the artists, like me, being listed in levels B, C, D, E and F - which means I could be in any of those because I don't have enough printed shirts yet).

There are some daily artists (like Nathan W Pyle, Anna-Maria Jung and Randyotter) who have enough prints to get a lower standard deviation, but, as you point out, most artists don't have enough prints to get a tighter SD.

The Walmazan thing (which is only visible because of his high number of prints) I found interesting only because it flipped ends of the scale, depending on whether it was a Derby print or a Daily print.

It could be some of the other "squishy middle" artists follow the exact same pattern, but just don't have enough prints yet to reveal that.

Thanks a bunch for the work! It was really cool to see it all and get an idea of what things kind of look like.


I found it rather fun, as I recently switched positions in my company, and my new position (project management) uses JMP far less than my previous one (six sigma black belt), and I like seeing processes with enough data from which to mine something interesting.

Shalom

taternuggets


quality posts: 21 Private Messages taternuggets

I have a headache after reading this thread. Interesting stuff though.


Nothing follows.

lyonscc


quality posts: 5 Private Messages lyonscc
taternuggets wrote:I have a headache after reading this thread. Interesting stuff though.



Rofl...

Just to follow-up

Mrwednesday:

Here is one of the comparison tables (they're hard to fit into a screen-cap)

(Click for a larger view)

Also, just to help clear up any confustion - the sales Means (averages) are not of most importance in the graphs/tables I included above. What you should look at is demonstrated in the area I've circled in the "top artists' graph" below.



The circled area shows the statistical "levels", or groupings, where members of the group are statistically similar to one another, but "levels" are statistically different from each other.

So, if an artist is present in one level, but not another, one could infer that he/she is statistically significantly higher/lower in sales than the group in which they're not included.

mrwednesday


quality posts: 12 Private Messages mrwednesday

No I got that part. What I was saying is that if those significance levels are reached by just doing paired t-tests at 0.05, then the alpha level is far too generous. Because each test will have a false positive once in twenty and you've got around 20 tests. I'm not saying it needs a bonferroni correction since that might be too conservative but the alpha should be adjusted in some way. I just wanted to know if it was because it doesn't look like it is. It just lists the alpha as 0.05 for each test.

Edit:

Scratch that. You've got 120 comparisons of 16 designers. That means that on average you'll get 6 significant differences that aren't real. That's kind of a lot.

thegambler311


quality posts: 1 Private Messages thegambler311

This kind of worked out perfectly for the "OP" didn't it? Woot announced a (partial) return to pure vote-based winners, and in the same week there's a nice cheap random sale.

paintednightsky


quality posts: 6 Private Messages paintednightsky

Staying simple, the reason I haven't bought a shirt in months is because I haven't liked any of the designs. When I first found woot, I was coming on to the site everyday sometimes staying up till midnight to see what the new design was. But as of late, I find the designs lackluster and I dislike having words as the main focus of the shirt. Before, I was buying at least once a week, but now you will probably be lucky to get me to buy once a month. I am still waiting out for a design worthy of buying. I feel like even the sales in the other categories have gotten less interesting.

lobsterbutt


quality posts: 0 Private Messages lobsterbutt

Hello massive walls of statistics and plethora of theories, I am bumping to give my own two cents like many others have:

I stopped buying from Woot when I realized they were still printing sekiyoku designs lol, like seriously guys??

lobsterbutt


quality posts: 0 Private Messages lobsterbutt

... also because i apparently lost the password to my old account??