What is meta? ×
Meta Stack Exchange is where users like you discuss bugs, features, and support issues that affect the software powering all 134 Stack Exchange communities.

(Skip this paragraph if you know FGITW) In the starting phase of a good question, several similar answers show up in a very short frame of time. One of them will happen to get the first upvote, and following votes may be influenced by that fact since some people don't have the time or muse to read all answers and restrict themselves to upvoting the first upvoted correct answer they find. This is known as "the fastest gun in the west".

I wonder, would displaying not the upvote-count itself but it's base 4 logarithm, rounded randomly&, decrease this problem? I picked 4 as the base because 2 would still provide too large a chance for the first upvote to show up as such, while 4 would give it a 75% chance of still being displayed as not-yet-upvoted, giving the other answers a fair chance. An answer with at least 4 upvotes will be displayed with at least a 1, 16 upvotes (and therefore a really good answer) with a 2, and even ridiculously upvoted ~400 times answers (which usually involve freehand circles or at least a very appropriate XKCD link) would get a 4 to 5.

Clicking the number should still reveal the real vote count (after reaching the rep needed unlock this), I assume$ high-rep users to vote responsibly anyway.

&e.g. take the fractional part of log4(nupvotes) as the probability for rounding up

$ yes, I am naive, thank you all the same


edit Taking Grace's critique into account, the fractional part of the answer-"score" should be displayed after a cast vote can no longer be undone to provide feedback, and maybe the rounding should be made real instead of random after that cast.

Also note that base 4 was just a first guess, maybe base 3 or base 2 with a preceding decrease or similar would be better.

edit2 Base 2 seems fairer and to be correct and avoid $-\infty$ for zero votes, the correct formula would be

with L := log2((abs(votescore)+1)  and r a random number in [0,1]  
score = sign(votescore) * { floor(L) + ( r > frac(L) ? 0 : 1) }

Here's an ugly table:

|votescore| || |score,min| || chance of rounding up
     1      ||        1     ||          0.00 %
     2      ||        1+    ||         58.50 %
     3      ||        2     ||          0.00 %
     4      ||        2     ||         32.19 %
     5      ||        2+    ||         58.50 %
     6      ||        2+    ||         80.74 %
     7      ||        3     ||          0.00 %
     8      ||        3     ||         16.99 %
     9      ||        3     ||         32.19 %
    10      ||        3     ||         45.94 %
    11      ||        3+    ||         58.50 %
    12      ||        3+    ||         70.04 %
    13      ||        3+    ||         80.74 %
    14      ||        3+    ||         90.69 %
    15      ||        4     ||          0.00 %
    16      ||        4     ||          8.75 %

+ indicates a 50+% chance of actually rounding up
share|improve this question
2  
I think a "solution" like this would vastly increase the number of votes cast on the site in a struggle to clarify the most appropriate answer to the question. This might mean other statistics would need to be offset accordingly, such as reputation requirements, badge awards, etc. Also, I don't think "Fastest gun" is a big problem for the site as long as the fastest gun is also quite a good shot. Good for them, they managed to answer the question correctly faster than anyone else, and so what if they edited after that to improve their answer? –  Andy E Aug 2 '10 at 13:31

2 Answers 2

up vote 6 down vote accepted

What will it look like when a user casts the vote? If your vote doesn't bring the value high enough to increment the visible score, it will appear as if the vote did not register. Which is bad.

The alternative is to display the correct score count after voting, which then becomes awkward as the count after voting is only going to be one different from the display prior to voting when the first vote is cast. Which is also bad.

The point of vote sort is so that the highest scored answers are the ones you see first, which (traditionally) means the most useful answers are given straight to the user. By collapsing multiple values into one score, then suddenly the extremely useful answers are on equal grounds as much lower - a score of 14 is the same as a score of 4 despite receiving a full 10 votes higher.

Overall, this causes a lot of problems with the perception of the voting system. It becomes awkward for the voters and it devalues the importance of those votes to other users (especially the majority with less than 1k that won't be able to do a split vote count).

share|improve this answer
    
Thank you for your feedback. It is absolutely justified and the reason why I made this question a [discussion] instead of a [feature-request]. Yet I hope you don't mind me trying to "fix" these issues: A vote should indeed have feedback, you are right, so maybe one could display the fractional part after a vote, too (but only after the vote can't be undone anymore). A 4x upvoted answer would be displayed as a 1 for sure, while a 10 would be displayed as a 2 with a 66% chance, plus if it displays as a 1 users are encouraged to clarify its "superiority" by further upvotes up to 16 votes. –  Tobias Kienzler Aug 2 '10 at 12:50
    
@Tobias If the number is changing randomly when they view it, then it is even more awkward. The number should only ever change when the actual votes are changing. It is extremely unintuitive for a vote count to fluctuate when absolutely nothing is happening. –  Grace Note Aug 2 '10 at 12:53
    
randomness is definitely a flaw... That was my first "solution" when I thought about the 4 vs 10 upvotes problem you mentioned. On the other hand, it kind of beats gaming which usually relies on determinism. So apart from that, what's your opinion on a pure log-view including the real fractional part? –  Tobias Kienzler Aug 2 '10 at 13:00

Just my opinion, but the system is designed to get answers to questions quickly, not distribute votes fairly across all answerers. This solution might make it less important to get your answer written quickly, but it would surely exacerbate the problem of answer sniping, where I simply crib my answer from someone else's.

In my experience, this isn't a real problem. Over time people who answer questions well are equally likely to be first as last answering a question. Generally, I've seen that better answers eventually migrate to the top. This is helped by the OP usually accepting the best, most helpful answer (though not necessarily the longest, most time-consuming one), which pins it to the top where it will likely receive even more votes.

Overall I think the FGITW problem is really only a problem for people who are motivated primarily to gain rep, not provide good answers. You can do both by providing a quick, correct answer, then expanding on it by adding examples and alternatives by editing. If you provide good answers, though, eventually the rep will come regardless of whether you're first or last.

share|improve this answer
    
+1 for "the system is [not] designed [... to] distribute votes fairly" –  Tobias Kienzler Aug 2 '10 at 13:04

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .