User talk:Faebot/SandboxX

Feb 31st?!

From the table:

Dschwen (talk · contribs) 	16823 	1556 	2005-02-31 	Remove

This cannot be. Correct to Mar.1st, or Feb.28th, or something… -- Tuválkin ✉ 16:55, 18 August 2013 (UTC)

 Done --Fæ (talk) 21:57, 18 August 2013 (UTC)

Weighted totals

I tried weight each vote with all three additional parameters — account creation date, number of edits, and number of uploads. Imported to Excel and added to each row the following three six columns: ^{(log(x+1) added, as suggested below. -- Tuválkin ✉ 17:52, 19 August 2013 (UTC))}

=SQRT(TODAY()-D1)*IF(E1="Keep";1;IF(E1="Remove";-1;0))
=SQRT(B1)*IF(E1="Keep";1;IF(E1="Remove";-1;0))
=SQRT(C1)*IF(E1="Keep";1;IF(E1="Remove";-1;0))

=LOG(TODAY()-D1)*IF(E1="Keep";1;IF(E1="Remove";-1;0))
=LOG(B1+1)*IF(E1="Keep";1;IF(E1="Remove";-1;0))
=LOG(C1+1)*IF(E1="Keep";1;IF(E1="Remove";-1;0))

And I fixed the date data thusly: ^{(updated) -- Tuválkin ✉ 17:25, 19 August 2013 (UTC)}

user	date used	remark
Dschwen	2004.11.19	corrected by Fæ
Krzysiu	2005.09.30	date of 1st edit
Blue Elf	2005.10.23	date of 1st edit
SatuSuro	2005.12.19	date of 1st edit
-jkb-	2004.11.14	date of 1st edit
Jklamo	2005.09.26	date of 1st edit
Colin	2005.11.21	date of 1st edit
Avenue	2005.01.09	date of 1st edit
Geni	2005.02.14	date of 1st edit
Ghirlandajo	2005.06.24	date of 1st edit
David Gerard	2004.08.29	date of 1st edit
Kaganer	2004.11.30	date of 1st edit
Inisheer	2005.06.20	date of 1st edit

The results are interesting:

weight	metric	total	wins
²√x	account creation date	0−1752,61	de-Bureaucrat
²√x	number of edits	00+975,53	Keep
²√x	number of uploads	00+339,90	Keep
㏒ₑ(x+1)	account creation date	00−102,20	de-Bureaucrat
㏒ₑ(x+1)	number of edits	000−60,77	de-Bureaucrat
㏒ₑ(x+1)	number of uploads	000−32,05	de-Bureaucrat
—	simple tally	74−104=−30	de-Bureaucrat

-- Tuválkin ✉ 17:16, 18 August 2013 (UTC)
-- Tuválkin ✉ 21:35, 18 August 2013 (UTC)
-- Tuválkin ✉ 17:25, 19 August 2013 (UTC)
-- Tuválkin ✉ 08:05, 20 August 2013 (UTC)
-- Tuválkin ✉ 00:00, 22 August 2013 (UTC)
-- Tuválkin ✉ 20:24, 22 August 2013 (UTC)
-- Tuválkin ✉ 13:21, 23 August 2013 (UTC)
-- Tuválkin ✉ 16:16, 24 August 2013 (UTC) (UPDATED)

Yes, very interesting. We should seek consensus for policy on that. Can you link it to the de-crat page to work in real-time?--Canoe1967 (talk) 17:32, 18 August 2013 (UTC)

Canoe, feel free to link here, but this is, at most, the basis for a policy solidly based on consensus, as you said. I wouldn’t like to see these figures (which anyone could think up and compute; I claim nothing whatsoever) being used to disrupt even more the Russiavia/Jimbo/Pricasso debacle.

The principle «one man, one vote» may be, time honored as it is, too simple for the virtual meritocracy we try to be, but I fear that any limitation of that rule which tries to be more accurate in reflecting said merid of each voter (instead of cruder, arbitrary limitations, such as a preset minimum of edits of a trial period), such as the weighting above, might be misused (the temptation of inflating the number of edits or uploads…), or at least might be accused of being an invitation for such abuse, not to mention the endless possibilities concerning the used metrics (subtract wiki-holidays, undone edits, deleted/superceded contributions?, use some other coefficient instead of a square root?…) — in the end making the whole process less clear and effective, and ultimately less fair.

-- Tuválkin ✉ 21:29, 18 August 2013 (UTC)

more weighted totals

weighted by…	total	wins	-Fae	wins	-Fae+Rus(?)	wins
account creation date	−938	de-Bureaucrat	−943	de-Bureaucrat	−925	de-Bureaucrat
number of edits	+617	Keep	−312	de-Bureaucrat	+273	Keep
number of uploads	+149	Keep	−222	de-Bureaucrat	−76	de-Bureaucrat

The first column is as above (figures are slightly different, due to a different time of computation). Why taking the square-roots (as already done above by User:Tuvalkin)? Because of the out of the pattern values for User:Fae (don't discuss why, this is not the place). Without square-roots, Fae makes the choice by himself. If we consider only all the others, this leads to the second column. If additionaly Russavia votes 'keep', this leads to the third column. In other words, choose your result and deduce the according 'mathematical' rule (another example for the book: 'How to lie with statistics'). Nevertheless, an useful exploration. Pldx1 (talk) 14:41, 19 August 2013 (UTC)

I vote for the +Fae weighted solution, where my opinion outweighs everyone else collectively. LOL - at least it would make for a very short !vote... --Fæ (talk) 15:30, 19 August 2013 (UTC)

edit conflict I agree with not using weighting to evaluate votes, as said above. However, the square root (or any other such parameter) is routinely used to normalize distances and similar factors. Also, removing outliers in a vote is even worse (less fair) than overweight their value by paremetrizing. -- Tuválkin ✉ 15:40, 19 August 2013 (UTC)

Ok with all you have said. In the same vein: reverse problem. Someone asserts that logarithm is, in our situation, the most fair normalizing function. Deduce her opinion. Pldx1 (talk) 16:15, 19 August 2013 (UTC)

Cannot use simple $log$, as some (may) have zero uploads. Add one, perhaps? -- Tuválkin ✉ 17:28, 19 August 2013 (UTC)

Additionally, deduce if she is using edits or uploads. Cordially. Pldx1 (talk) 17:51, 19 August 2013 (UTC)

This might help with the calculazions. --Túrelio (talk) 15:10, 23 August 2013 (UTC)