Game Ranks are Only Half of Your Score

[Mike_Crosley]

So, your ranking score is only half right and here's why. When you win, you get points, duh. If you play against a higher ranked player, you get more points. Win against a lower ranked player, you get fewer points. Seems like a good way to rank a player's progress, right? Well, its only half the story.

For example, I really like the game Can't Stop. So far I've played 1448 games. Of those, I've won 761 (as of this writing). However, my ranking score is only 260. While 260 is a nice rank (and I'm happy about it), what's missing? Well, the fact that I've won 74 more games than I've lost. As a percentage, I've won 5% more games than I've lost or, 55% of the games I've played. That may sound like a small percentage, but its not! Its VERY significant in any two player game.

Despite the ranking score, there's a lot that goes in to blurring the accuracy of its representation. You could have disconnected when you played against a highly ranked player and took an especially bad hit to your score but not due to how well you played. You could have run across a streak of mismatched players so your rank took a beating, but your percentage changed only marginally because overall you didn't have as big of losses as the rank might indicate.

I started noticing this when I thought "I fancy myself really good at this game, so the best way to know that is 'Do I win more than I lose at it?'" Well, the ranking doesn't tell me that at all. It just tells me "Overall, when you did win, you beat better ranked players. When you did lose, you lost to lesser ranked players, also overall." (or vice versa depending on if your rank is in the toilet or doing well).

So I think there should be a PERCENTAGE WON statistic alongside the ranking value. Together these would tell us not only have you beaten (or been beaten by) more higher ranked players than lower but also "and overall, you have X% wins".

The way I determine mine is I take my number of wins, double it, and subtract that from the total number of games played. If the value is positive, I'm ahead of a 50/50 split by that many games. If its a negative value, I'm behind a 50/50 split by that many games. Currently my score, again, is 74 which tells the story a lot better than the ranking score.

Thoughts?

[robinzig]

You can do whatever calculation you want if it makes you happy, as this clearly does. But "percentage of wins" (which although slightly different is basically what your chosen measure amounts to) wouldn't work if actually used as the official metric on the site, as it's easily gamed by picking on much weaker players and avoiding strong ones. A 95% win rate in 2p games is much less impressive if all your games are against beginners! (I'm not saying you do this, not at all - just that this would become encouraged if BGA dropped ELO for a metric like this.)

While I take your point about randomness including disconnections etc that you can't do anything about, over time these should balance out. While there are a lot of justified complaints about BGA's use of ELO for multiplayer games, and for team games in particular (where it's pretty nonsensical), for 2p games ELO is actually probably about as accurate a measure of performance/skill as is likely to be found.