Wild and crazy idea: Two ratings

I'm sure this will be spat upon, but after playing over 100 games here across probably 80 or so different opponents, I've noticed that the current rating calculation doesn't accurately gauge how close of a battle I'm in for...

I've met a few opponents ranked at least 200 points higher than I at the table and smoked them - I've also met people ranked significantly lower than I and had them put up one heck of a fight...

I realize errors are common on both sides of the table (definitely my side, anyway) but that cannot possibly explain the skill differences. Which brings me to my thought:

Give each player two ratings:

One rating would be calculated by every ranked game they've played - (exactly as it is now).

Have the second rating be calculated by every ranked game they've ever played - excluding games ended by one side claiming a timeout victory.

I think this would give a more accurate representation of the player's true skill...

Thoughts?

not sure, but not the worst idea ever for sure (I've got that one wrapped up forever).

You could display a current rating (say last 30 games) and an overall win %. That might be a better indicator of level of play to expect from the opponent. Haven't thought it through much though.

Originally posted by Villain
I'm sure this will be spat upon, but after playing over 100 games here across probably 80 or so different opponents, I've noticed that the current rating calculation doesn't accurately gauge how close of a battle I'm in for...

I've met a few opponents ranked at least 200 points higher than I at the table and smoked them - I've also met people ranked ...[text shortened]... think this would give a more accurate representation of the player's true skill...

Thoughts?

I like your reasoning, but I think 2 ratings might overcomplicate things.
How about if we just keep the one rating and show how many games were won or lost by timeouts.
This could either be a percentage of the total number of games, or just as a number of games.
This would give people some clue as to how many games were actually won or lost by checkmate or resignations.

Originally posted by martin williams
I like your reasoning, but I think 2 ratings might overcomplicate things.
How about if we just keep the one rating and show how many games were won or lost by timeouts.
This could either be a percentage of the total number of games, or just as a number of games.
This would give people some clue as to how many games were actually won or lost by checkmate or resignations.

This would also help - while not as descriptive, it would still be a step up in my opinion...

Hmmm. Not so sure about this. There are plenty of people out there who just drag there feet, or don't play at all when a game is definately lost. So this would devalue games that were genuinely won, and not just taken by 'Pressing The Win Button'.

One rating would be calculated by every ranked game they've played - (exactly as it is now).

Have the second rating be calculated by every ranked game they've ever played - excluding games ended by one side claiming a timeout victory

WDF? You just described them exactly the same. Only one doesn't have time outs counted as wins, plus timeouts are apart of the game, if you can't make one move in like 3 days, then play longer time control games

Originally posted by Pawn Artist
WDF? You just described them exactly the same. Only one doesn't have time outs counted as wins, plus timeouts art of the game, if you can't make one move in like 3 days, then play longer time control games

..go back to sleep

I hesitate to suggest this, but I think ratings may be somewhat affected by players tending to play against opponents from their own time zones. This means European players play European players more often than U.S. players, etc. IF--and I say IF--this is so, then regional variations of, say, as much as 100 points in ratings could accumulate. This may explain why, say, a Turkish player at 1450 could wipe the floor with a 1550 player. (Names withheld to protect the red-faced.)

C.I.