red thumbs down - my two cents

[ivan the tolerable]

I can understand the difference between a personal log of who you do/don't like playing with, but for an open rating system red flagging people publicly there should be a bit more restraint to it. Everyone has thumbs down, most have quite a few of them actually, and you simply can't identify the truly unsportsmanlike and offensive players from ones that simply don't play as rapidly as some would like, or simply win a lot. Its become a meaningless measure.

I would like to see the private up/down thumb system stay the same, but a separate public up/down rating be a separate thing you can also choose to do.

There are people I don't want to play with, but will not red flag as they are not doing anything offensive, I just don't like the way they play. So in short, both rating systems are useless when combined - i cannot get a complete private list of players i do/do not like to play with, and i cannot tell from red thumbs what players are bad actors and which ones simply win a lot or think sometimes.

[zahlman]

My opinion: a system with more than a binary rating simply does not work. There is a reason that Youtube stopped providing a 5-star rating scale and moved to like/dislike: because in practice, people almost always only vote 5 or 0 anyway, and if you're one of the few who doesn't, your vote effectively does not count as much as others, so you are effectively penalized for not having a strong opinion.

Thumbs down are IMHO important as well because the reports process is too inconvenient to bother with unless somebody did something really wrong.

The main problem is that getting early thumbs-down hurts a new player too much. This can be fixed directly, by not showing the up/down totals, and instead translating those values into some kind of popularity score for display. This can be made to follow any formula, so it could be tweaked to make an early thumbs-down not count so much. For example, it could use the idea of confidence intervals. Do something like http://www.evanmiller.org/how-not-to-so ... ating.html, except use the upper bound of the confidence interval instead of the lower bound.

Another thing that can be done is to make the thumb-votes count less when they come from somebody with a lower score, so that more "trusted" people have a bigger effect and "revenge" voting ("I red-thumb you because you red-thumb me", or for example somebody who just likes to give red thumbs to others - such a player is probably not very well liked himself ) has lower impact. Along those 'trust' lines, the score could also be weighted for people who have been around longer and accumulated more total thumbs, though it'd have to be very slight weighting (e.g. using a logarithm) to be fair IMHO.