(originally posted on my personal website track7 on november 9, 2011: http://www.track7.org/bln/steam-recommendation-useless)

the steam gaming service for windows and mac allows users to recommend games to their friends, and write a very short review to go with it. this seems handy but it’s actually a lot less useful than it could be.

the useful part of the recommendations is when one of your friends recommends a game. this of course requires you to have friends on steam as well as those friends being the type who like to write out their opinions on games. it also helps if your friends’ opinions are useful in determining if you’re going to like a game. finally, they have to be skilled at saying something useful in the limited space steam allows for recommendations.

if you want to know what the steam community as a whole thinks, too bad because all you get is a number of people who recommended the game. that number is almost completely useless. it completely ignores how well-known a game is (a game like team fortress 2 has a much larger pool of possible recommenders than one like doc clock: the toasted sandwich of time). even worse is it provides no method for someone who played the game but didn’t like it to warn everyone else not to waste their money. one of my steam friends has actually posted recommendations where the text essentially says not to buy the game, but anti-recommendations like that actually get counted along with the recommendations from players who did like the game, pushing the number in the wrong direction.

a couple simple changes could make steam recommendations considerably more useful. the first, which i see as the most important, is to add a simple for / against choice when someone posts a recommendation. that way if someone plays a game and doesn’t like it, they can recommend against it and explain why. instead of just showing X players recommend this game, steam could show Y of X (Z%) players recommend this game where the first number is only the recommendations for and the second is total recommendations for and against. the percent then helps to normalize across games with different-sized player bases.

while that change would allow the dissatisfied to be heard, it doesn’t tell potential buyers about the number of people who didn’t have a strong opinion either way. those people aren’t likely to post any sort of recommendation for or against the game, so they’re a little trickier to count. really though what you want is the total number of people who reasonably could have had an opinion, including those who posted a recommendation. it’s probably easiest to get the number of people who own the game, but that’s going to include everyone for free games as well as people who bought or otherwise obtained the game but haven’t played it yet. from what i’ve heard about some people’s buying habits during steam sales, and my own habit of buying all the humble bundles even though they come out with new ones before i’ve had time to try out all the games i got with the previous bundles, games that sit on an account waiting for their owner to have time to play them are common enough. instead i would get the count by how many accounts have some playtime logged to the game. you could maybe go with at least an hour of playtime, but i’ve actually completed some games in less than a half hour and still had an opinion.

with those two changes, steam could show a percentage bar with green on one end, red on the other, and yellow in the middle. the green end would show the percentage of recommendations for while red would show recommendations against. it could list those two percentage numbers along with the total number of players, but we probably don’t actually need the total number of recommendations.