Pioneer54: I expect we'll disagree on this one. First of all, as someone who's been a software engineer for 30 years, I can assure you this would be remarkably easy to implement. I believe if someone is intelligent enough to use this site and play games, they'll very easily understand it. And there'd be no implicit penalty of anyone who kept playing a game, whether they came to the site earlier or later. And I've spoken to plenty of people here who do believe that inactivity is a justification for demoting ratings. (Personally I couldn't care less, but this is a solution to what seems to be a significant problem for many.) Ratings are indeed an inexact science, a fact mirrored by Fencer apparently using an extremely simple lookup table to calculate them.

Jason: I only play two games here at the moment, but I have a habit of playing a game, say chess, for two years and then giving it up for another five years, then playing it again and leaving it again. Whatever the exact time period, I'm always away from a game for longer than I'm with it. It would give me a similar effect to the one you suggest. I'd build up a rating and then it would decay a long way. But I'd be prepared to take that hit on my rating. It would be my choice. And many have told me they would think it fairer if ratings like that did slip while people were away. On the other hand, I do believe your point about 'concentrated' players doing better than 'dispersed' players is a lot more valid than any of Pioneer54's, and it would be the only reason of those mentioned so far that think I would support against my own proposed solution.

The problem: In many games rankings (notably near the top of the rankings for that game but not necessarily just at the top) there are some people who just stay there without playing. Perhaps when they joined they got lucky in a few games against some medium to high ranked players, got themselves a high BKR and thus a high ranking, and then decided not to risk their positions. Or perhaps they have not played any games for a year or more. Whatever the reason, very many other players find it frustrating that this 'deadwood' clutters up the ranking lists, especially when they don't respond to challenges or consistently decline them.

Old proprosed solution (that I don't support): Periodically purge these non-players from the rankings.

My proposed solution: In addition to the existing BKR, we have an 'Adjusted BKR' and that rankings are based on this Adjusted BKR. The existing BKR persists as "Unadjusted BKR' but rankings are no longer based on Unadjusted BKR.

What's the difference between them? Adjusted BKR is adjusted downwards by a small amount every day that a player is not on vacation or a weekend (and whether they have played or not). Therefore in order to keep your adjusted BKR high and therefore your ranking high, you have to keep playing games and winning them.

The amount can be proportional (like 0.08% per day) or linear (like 2 BKR per day). Since there are 104 weekend days per year and 30 Rook holidays per year, this would mean that a rating that was stuck at 2700 would decay down to around 2240 over a year using those figures if they didn't play in that time. Different figures can be used if those are considered too severe or too lenient. The linear system might be better because (a) it is easier to apply in software and (b) it stops 'bunching' in the medium to lower rankings. The proportional system may or may not be fairer.

I propose keeping the Unadjusted BKR but not using it for ranking. It can still be used to give an opponent an idea of someone's ability. If they are a medium ability player of around 1800 and they are thinking of challenging another player also around 1800, they will know whether they are playing someone about the same as them or whether this is someone who used to be a good player a year or two ago and just hasn't played since.

BKR could in future be displayed as 1800/2400 (= adjusted/unadjusted).

As well as being useful to very many players (from what I've heard) this system would also be an advantage to team captains who want to choose players who both high rated and who are 'current'. Doing this would be easy using Adjusted BKR whereas at the moment you have to use the Unadjusted BKRs plus look at each of the graphs.