Note: The answer here cannot be "neither 1) or 2); both add 200 and subtract 200 are incorrect, you really need to subtract 1000". Because the /distributions/ for both are available, in one I am three s.d. poor of mean and in the other I am right on top of mean.
"Elometer" results wildly different than rating?
Sort:
I saw a post with this "elometer", which uses a statistically validated model to predict your playing strength. Since there was some discussion about if real-world elo are slightly higher or slightly lower than chess.com elo, I decided to try it out.
My rating here at 15|10 bounces between 500 and 700. I wanted to see if I got something under (300-400, since some people say "add 200 points from your OTB to get chess.com"), something more like 700-900 (since some people say "subtract 200 points from your OTB to get chess.com")
Honestly I got demoralized really quickly and started blowing through it, since after awhile the positions started feeling hopeless and no-good-moves, similar to positions I constantly get crushed in on this site at the 500-700 elo level. "I just totally suck, I guess, but better not to just quit out."
Here is the result:
Can anyone even begin to explain this?
The only two options seem to be:
1) The chess test "elometer" is totally, totally broken in a basic way. (seems unlikely based on their methodology and previously validated results - the test is calibrated to match real-world elo and has been verified to do pretty well in that regard.)
2) Ratings on chess.com are totally fake / inaccurate and there is a pervasive, systemic reason for this.
I don't see a middle ground but I'm willing to entertain options.
Does anyone have any sort of experience taking the test they'd like to report? Or ideas about how this could be so wildly discrepant? (It's http://elometer.net/, btw)
There were some chess.com specific questions post-test. I've emailed the researchers to ask them about why they think the difference is so large and I will report back here if they respond.