2

The USCF rating system as described by Professor Glickman includes a factor called K:

If m < 3, or if the player competes against any opponent more than twice, the "standard" rating formula that results in Rs is given by

Rs = R0 + K(S - E)

(source, emphasis mine)

Though at one point the K value was calculated for each player from the pre-event rating, it is now defined by one of two expressions depending on whether the tournament is "half-K" or "full-K:"

The value of K, which used to take on the values 32, 24 or 16, depending only on a player's pre-event rating, is now defined as...

(source: same page as above)

What does it mean if a tournament is half-K (besides that it changes the value of K)? Why would the tournament director choose to make an event half-K?

I also see that the USCF considered removing the half-K option, then did remove it, then considered using it for quick games in some way. However, none of those reports explain what being half-K actually does to a tournament.

  • Lowering K just makes the games in the tournament count for less in the rating system (halving K makes people's ratings change by half as much). I agree that it seems weird to just do this arbitrarily. It would have a little more justification if it made sense to consider a tournament to have less reliable data for rating purposes, e.g., if it consisted of fast games. – dfan Jun 29 '15 at 12:19
  • Far better to have more ratings based on actual speed of play rather than bunching them together in so few sets. My play varies widely depending on the speed and the two speeds I normally play are both in the same class. So what is my real rating? The high one at the faster speed or the low one at the slower one?? – yobamamama Jan 9 at 18:42
1

As dfan mentioned in his comment, K factor, is a factor for changing convergence speed of ratings. When you win against a strong player, your rating converges to your opponents rating, after some consecutive wins, you will have the same rating with your opponent. K factor simply changes how many games you need to converge we can say. Situations where it is used:

  • High K used, in first few matches of a player, to let the player reach actual Elo quickly
  • Low K used, in some tournaments which has non-standard time controls
  • Low K used, in matches between masters, to stabilize Elo changes.

etc.

There is no absolutely correct K value, it's really open to question how fast it should converge. But when it converges slower, Elo points are more stable. When it converges faster, people can reach their actual ratings faster.

I caused a lot of trouble to people just because of this dilemma. I had an initial rating of 1100(for some riddicilous reason, long story). And at the time my actual strength was about 1500. In time, my strength went up to 1900 but my rating was about 1400. Just because of this, my opponents lost lots of points(unfairly). I remember a game when I was 1300 something, and I won against a 2000 player. He lost a lot of points, and was very sad. If I were 1900, penalty would be much more less.

| improve this answer | |
  • 2
    It was always my understanding that "half-K" was an option that an organizer could choose, to try to entice very strong players to play in an event where playing conditions (such as short time controls, or the presence of many low-rated but improving players) might otherwise dissuade them from playing. This would mean that the desired strong players could play with fewer rating points at risk. In other words, more a marketing thing, not something that supports the statistical integrity of the rating system. It was probably made mostly obsolete when the separate quick system was established. – patbarron Jan 3 '16 at 4:32
  • It can be used the way you explained in your comment too. Anyway, for some reason, USCF wants to tweak Elo change speeds. And they invented K factor to do it. But not arbitrarily, there are some policies(or maybe traditions I should say) and organizers generally follow them. For example, fast games has lower K. – ferit Jan 3 '16 at 4:40
  • The K factor is part of Elo's original system. Used to be that K was a fixed value in the USCF implementation (for players with established rating under 2100, it was always 32, unless half-K was used). Now it's variable, depending on the length of your rating history, so newer players' ratings change faster. The USCF systems use the same K computation for regular, quick, and blitz time controls - they're just maintained as separate systems that don't affect each other (other than G/30 through G/60 events are rated in both systems - but each still calculated separately). – patbarron Jan 3 '16 at 5:04
  • "Both system" being both the regular and quick systems... – patbarron Jan 3 '16 at 5:06
  • Yes, you are right. – ferit Jan 3 '16 at 5:07

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.