EVE General Discussion

Thought I'd go see what talented people can do with the character creator. Was amazed to see my name up there with all you pretty people. Thank you for making my day.

If that means what I think it means, then frikkin' YAY!

+1 like


Edit: This idea just struck me, and it could be awful or restrict the use of your site more than you'd like, but if you made it so voters had to register and made voting history public once match history is in place, it might further 'encourage' objective, unbiased, non-random voting. Maybe we could even sort by the highest volume voters? I'd be interested in examining the 'tastes' of those who cast the most votes...

If that's too prohibitive, maybe there could be the 'public' contest as it exists currently, and a separate, parallel division in which only registered voters (with public voting records) are allowed to participate?

Maybe Nova should reserve the top spots for those characters and exclude them from the voting, but maybe it is only your opinion that they belong in the top 12 and not everybody's.



They don't appear more often than any other avatar.



I don't think you understand the voting process, it's not like that 10 voters vote on 10 individual cycles, all voters vote on one cycle at any time. Besides, nobody is prevented from voting against a pretty looking avatar.

It's Avatar PVP at it's best

One improvement would be to remove the character names on the voting pairs as I think some voters might be biased by name and/or reputation.

Fair enough.


That's how "random" works. Pure random without controlled seed actually means that the random distribution is not random at all. If you throw a dice, 20 times, and get 5 occurrences of 2s, the end result would still be "2 appears more often than others". Regardless of the fact that throwing dice rolls are purely random.




While I won't pretend that I know how the site was written. I'm a software developer and I've designed similar algorithms before. A straight random occurrence pairing does not work without proper evaluation mechanics in place to control the RnG seeds. It doesn't matter if the site managed to lock a single voting instance to a particular user (which is almost impossible without hwid records, ip records are almost useless since a lot of broadband connections uses dhcp anyway) because the sample are so huge that luck factor affects the final results more than the actual results from pairings.

To put it simply it's like this, there are almost seven thousand avatars on that site, how much time does one spent, say each day on the site? maybe a few minutes, maybe an hour or two. How many avatars are included on each activity? a dozen? a hundred? That's less than 1% of the full data sample. Even when someone spent twice the time and do it for a full week he/she'll barely reach a quarter of the avatars on the database.

With plain RnG seed and basic win:loss evaluation, those avatars who were lucky enough to pop out on those instances gets the points (or lose them in the case of downvotes), seeing how similar the win:lose ratios on top 100-200 avatars (around 120 wins and 35 losses) you can easily see that RnG luck has more to do with the actual top score than the 'value' of the avatars in itself.




Agreed.

You bring up good points about random matching shortcomins, but after trying a few different approaches, random works better and is more reliable, also there are less bugs with it. Randomized approach seems to provide the least-interference type of voting. And I think it has worked ok so far.


The very first revision of New Eden Faces would send you 1000k+ characters to your browser and let you vote for characters independent of other users. As you can imagine that spiralled out of control on the first day. Some time later the server code would maintain the list of voted characters to prevent voting for the same character multiple times. As of October 2013 it is now handled on the database layer. It is not possible to vote for the same character more than any other character per round. Voting rounds are global. If you voted for Makalu Zarya just now, everyone else will have to wait until this round is over before someone sees Makalu Zarya again. So it's a random matching but with a condition that either Character A or Character B have not been voted on yet.

As a voter I wish i could vote and see every character pair, but with that you cannot have balanced votes.



I will add the winning and losing streak counts to the character profile page, but again as I explained above it's best to leave randomized voting as is. If I am to change it again I will surely screw it up and introduce more bugs with this new slightly more intelligent match pairing. There are lots of easy features I could implement first, and if I have time in the future I will consider on improving match pairing and voting algorithms.



I am planning to add a page with handpicked avatars and maybe update it once a month or until I see better avatars.


Coming in the next update, but as a side-effect you should know you won't have a way to open character profile page. If it's not a big deal then I don't mind removing names when voting. (Speaking of bias, back in the day it would actually show you a winning percentage with a nice progress bar right next to character names)

Poor falcon. His feels are hurt!