So for hypothesis 1, in order to really do this right from a statistical point of view, I'd have to do as much work as I did on my qualifying exams. (Assume that the slates have R and S respectively, while each candidate also gets a bonus amount of nonslate voters, drawn from a negative binomial distribution with parameters α and β, find the MLE of... then use empirical Bayes...)
However, a quick and dirty estimate is: each type of candidate is getting an extra bonus which averages to a fixed number. So the average double-puppy has 170.5; the sad has 132; and the rabid has 66. 132 + 66 - 170.5 =27.5 would be the estimate for the nonslate votes each pubby candidate was getting. Eyballing things, this would probably have placed them in about 10th-15th places if the slates had not existed. This seems not unreasonable.
So using those numbers, the number of to-the-letter slate voters was probably only about 143; with (in this category, at least) more sads (over 100) than rabids (under 40).
So for hypothesis 1, in order to really do this right from a statistical point of view, I'd have to do as much work as I did on my qualifying exams. (Assume that the slates have R and S respectively, while each candidate also gets a bonus amount of nonslate voters, drawn from a negative binomial distribution with parameters α and β, find the MLE of... then use empirical Bayes...)
However, a quick and dirty estimate is: each type of candidate is getting an extra bonus which averages to a fixed number. So the average double-puppy has 170.5; the sad has 132; and the rabid has 66. 132 + 66 - 170.5 =27.5 would be the estimate for the nonslate votes each pubby candidate was getting. Eyballing things, this would probably have placed them in about 10th-15th places if the slates had not existed. This seems not unreasonable.
So using those numbers, the number of to-the-letter slate voters was probably only about 143; with (in this category, at least) more sads (over 100) than rabids (under 40).
The problem with hypothesis 2 for explaining the fan writer category is that Matthew Surridge, the weakest double-puppy, was actually at the top of the sad list and second-to-top on the rabid list. True, he was one of the two double-puppies who was not linked by either list; but on the other hand, the other unlinked double-puppy was Amanda Green, who got the second-to-top votes (175).
So if the answer is something like hypothesis 2, at least 1/3 the puppies would have to have been exercising some "statistically biased" process such as actual aesthetic judgement, when they decided whom to leave off of the slate. (It doesn't have to be judgement; it could be, for instance, that some preferred the woman on the slate lists, while others preferred the people who were linked).
So, why isn't R+S≤Rð•ŒS?
There could be two reasons:
1. Extra non-slate votes for the candidates. You'd expect all slate candidates to be pulling from the same distribution (ie, rolling the same number of dice) for those extra votes; so at a first approximation, that would mean subtracting the same number from each total to get the true puppy numbers.
2. Incomplete slate voters. This would be people who say something like "I do have one favorite fan writer who is not on the slates, so I'll vote for them; but then I'll fill out the rest of my ballot with candidates from the slate." You would imagine that such people would be more likely to take from the top of the slate list than from the bottom; though it's hard to know how strong this tendency would be.
Further analysis to follow...
The category of "Best Fanwriter" is particularly interesting. This is a category where both the sad and rabid slates had 5 items, and yet did not overlap completely; and also one which seemed to get relatively low voting (both puppy and otherwise) relative to other categories.
Enness, the puppy-only candidate, got 66; Dave Freer, the sad-only one, got 132; and Matthew Surridge, the weakest of the both-slates candidates, got 150.
This is a bit of a head-scratcher: why is the overlap candidate getting significantly less than the sum of the two separate slates? I have some ideas which I'll post in the next comment.
Now for rabid but not sad. 18 such things.
Min totals:
66 Daniel Enness
86 Supernatural: Dog Dean Afternoon
100 Coherence James Ward Byrkit
100 Andromeda Spaceways InFlight
Max totals:
196 The Chaplain’s War Brad Torgersen
187 The Revenge of Hump Day edited by Tim Bolgeo
172 The Plural of Helen of Troy John C. Wright
166 Vox Day
Glitch again. The second sad but not rabid thing with 132 votes was
132 Dave Freer.
Oops. Grr.
So, for things that were sad but not rabid: that is 6 things. 2 of them are episodes on Cartoon Network which got dramatically lower votes than the others. Perhaps most puppies feel themselves to be too dignified to vote for Adventure Time?
Vote totals were:
35 regular show, saving time 0 0 0 0 0 0
41 Adventure Time: The Prince Who Wanted Everything
132 A Single Samurai Steve Diamond 0
132 A Single Samurai Steve Diamond 0
185 Tuesdays with Molakesh the Destroyer Megan Grey
199 Trial By Fire Charles E. Gannon 10.90%
So, here I am laboring in the 15th floor. We were not able to get full ballot data but we do have some totals. That is not enough to be sure exactly how EPH would have come out but it's enough to make some well-founded guesses.
My first numbers (these are NOT about EPH, just about how prevalent puppy voting was):
There were 51 items that were on both Sad and Rabid slates. Of those, the lowest 4 vote totals were 150, 156, 158, and 160; while the highest 4 were 769, 487, 387, 372. Lowest and highest were MDSurridge, fan writer, and Guardians of the Galaxy, respectively.
For the 6 items that were on the sad but not rabid slate, the totals were
I'm sitting in the audience waiting for the Hugos. As soon as it's done and they release the ballot data, we'll be crunching the numbers and posting our findings here. We have a room ( in the 15th floor of the Doubletree, vegan room). Gonna be a long night for me.
If there's anyone at the con who has a nearby hotel room or similar space where we could meet to run the newly released 2015 data Saturday night into Sunday morning, please email me (firstname dot lastname at google's email service). Or you can of course buttonhole me here at the con...
As for the receipts: I'm not sure exactly how it works, but the payments don't go through until I finalize the order. But I don't see Cally Soukop on the list of people who have made orders.
Whoops, it seems that the order form link failed to post here. Here it is:
Please order your E Pluribus Hugo t-shirts ASAP using this link. They cost $8.04 per shirt, if you're there at the convention to pick them up. If not, please still order now, and we'll figure out what shipping costs later (probably a reasonable amount; but it certainly won't be free, since we have no profit margin on the shirts themselves).
In the notes on your order please tell us:
-Your name and email.
-If you will be picking the shirt up.
-If so, which day you'd like to get it.
-If not, your mailing address. We'll contact you by email to work out shipping payment.
Please order your E Pluribus Hugo t-shirts ASAP using this link. They cost $8.04 per shirt, if you're there at the convention to pick them up. If not, please still order now, and we'll figure out what shipping costs later (probably a reasonable amount; but it certainly won't be free, since we have no profit margin on the shirts themselves).
In the notes on your order please tell us:
-Your name and email.
-If you will be picking the shirt up.
-If so, which day you'd like to get it.
-If not, your mailing address. We'll contact you by email to work out shipping payment.
Please order your E Pluribus Hugo t-shirts ASAP using this link. They cost $8.04 per shirt, if you're there at the convention to pick them up. If not, please still order now, and we'll figure out what shipping costs later (probably a reasonable amount; but it certainly won't be free, since we have no profit margin on the shirts themselves).
In the notes on your order please tell us:
-Your email.
-If you will be picking the shirt up.
-If so, which day you'd like to get it.
-If not, your mailing address. We'll contact you by email to work out shipping payment.
I think that this debate is not going anywhere. Even if a longlist or lift-the-limit proposal is worth doing (which I happen to believe), there's no harm at all in waiting a year until we at least know whether EPH has passed. At that point, both sides can make their points on clear ground without having to pose multiple hypotheticals or risking giving ammunition to the anti-EPH concern trolls.
(That is, as long as EPH passes or fails with a clear enough margin this year so that the revote next year is pretty clearly foreordained. If not, we should wait two years to talk about longlists and such.)
I'll let the anti-longlist folks have the last word now if they want, and felice, I think you should do so too.
@369: Changing the rule would not change the social norms, so the number of extra nominations it would cause/allow would be minimal. The main reason to change it would be to remove the incentive for dishonesty, where if there are more than 5 things you'd like to nominate, you would strategically choose the more-viable ones even if they're not actually your top 5 choices.
But again, I'm not actually proposing this for this year, when the BM is likely to go into overtime anyway. This is more a minor fix that I'd support in an off year, not something worth worrying about in the current situation.
It's probably time to resolve the issues of T-shirts in the other thread.
@366: If that's supposed to be an argument for limiting the number people can vote for, then it's subject to a pretty obvious reductio.
My feeling is that if some people have higher standards, and some people have lower ones, it actually helps the system find the best candidates. This is shown clearly in simulations, as long as level of standards and overall genre/taste issues aren't ridiculously tightly correlated.
So it would be great if some people vote for only the 1 or 2 very best works they've seen in a year, while others vote for everything that's even halfway good. You wouldn't want everyone to vote like the latter group, but since that's not going to happen no matter what you do, constraining the voting system to prevent it is counterproductive.
| Year | Number of comments posted |
|---|---|
| 2015 | 514 |
Total: 514 comments. View all these comments on a single page. (May take some time to load.)
The most recent 20 comments posted to Making Light by Jameson Quinn:
Show all comments by Jameson Quinn.