Nice. I did the test and it estimated 1495 Elo.
Estimation on your ELO Rating test
Sort:
Yes, some interesting puzzle`s
I already recognize the first position from a Spassky vs. Tal game so it would be skewed in my case -_-
I just realised I misread the instructions. I need to do some of these positions again. 
Got 2020, woot! ^_^
Up to 1595.
The scoring code is broken. Here's how it works. For each problem, it has a list of several plausible moves and a rating for each, ranging from 1200 to 2600. The basic idea is that your final score is the avaerage of your move rating for each problem.
The way it computes this is for each problem it has a variable that gets the score for that problem. The variable elo1 gets the score for problem 1, elo2 gets the score for problem 2, and so on. Then it computes the final score as (elo1 + elo2 + elo3 + ... + elo10)/10.
When it scoring an individual problem, say problem 3, it does it like this. It first sets elo3 = 1000. Then it compares you move to the list of moves it has for problem 3, and if your move is on the list it sets elo3 to the score of your move.
Unfortunately, for problem 10, after it sets elo10 = 1000 and looks up your move on its list for problem 10, if it finds your move it store the score in elo9, not elo10.
Net result: problem 10 is always counted as scoring 1000, no matter what your move, and if your problem 10 move was better than 1000, the score you should have gotten on problem 10 replaces your score for problem 9.
There are three ways you can work around this.
Workaround #1. Enter an illegal move for problem 10, such as a1 a1. Then take your final score, multiply by 10, subtract 1000, and divide by 9. That will give you the average of your correct scores from the first 9 problems.
Workaound #2. Similar to #1, but enter the illegal move for problem 9. Then after you multiply by 10, subtract 1000, nd divide by 9, you will have the average score you got on 1-8 and 10.
Workaround #3. Calculate the score yourself. I'll give a link to a copy of the scoring key below. Look up the score for each of your moves (score 1000 for a move not listed), add them all up, and divide by 10.
Here is the scoring key: http://pastebin.com/c6x5qXAq
You could also wait until they fix the scoring JavaScript, but that bug was there at least a year ago when this test was posted on Reddit, so it doesn't look like it is going to get fixed.
I got 1630, not sure if this corresponds with my other stats on here but I've never actually played game of chess that has lasted more than 45 minutes.
I got 1520 which was higher than expected. Just about splits the difference between my live and turn-based rating here. Thought it would be lower than them both.
probably the same as this http://kopecchess.com/?page_id=55
With workaround #3 I get 1635.
1675. Which suggests my chess.com rating is inflated by around 50-100.
1805! Wow. Test confirmed inaccurate.
if no one believes i got that rating, im happy to give my answers
Problem with some / many of these is you can miscalculate like crazy, but choose a good looking move, and get a lot of points.
Looking at the answer key thing for diagram 9 it gives both Bg6 and Bh7 a score of 1800 but Bg6 throws away a bishop for absolutely nothing (??) and Bh7 is a meaningless check. I answered Bxh6 because hey, looks like an attacking move, why not. But it also loses (although not as instantly as Bg6) but is scored lower at 1700.
But the crazy thing is the totally fine move Rxb7, not at all losing like Bg6 and Bxh7, is given the lowest score of 1400. (By the way the answer, Bd4, makes a threat with a fun sequence).
A few other things I noticed. In diagram 2, I wanted to start with hxg with all the same ideas, but I thought this was a safer way to do it. Same thing with diagram 4, I wanted Bc2 with all the same ideas of e6 except I have an extra pawn so why not.
Thanks Wafflle, glad I read you post before taking the test, can you give out the answers for the rest please?
post #8 gave the scoring key. I didn't check every position with houdini, but I think the top pick for each position is the best move.
It's how it rates the other moves where I don't really agree.
waffllemaster wrote:
post #8 gave the scoring key. I didn't check every position with houdini, but I think the top pick for each position is the best move.
It's how it rates the other moves where I don't really agree.
I've added FEN for the position of each problem to the scoring key. Here it is:
This should make it a lot more convenient for people who want to take an engine to the positions to verify or take issue with the answers.
I've also appended to the bottom of the key a copy of the rating function from the page's JavaScript. This can be used to check my work to make sure I got the right answers. I've already found one mistake. In my original key, I had Rxa6 as one of the candidate moves, whereas it should have been Bxa6. That's corrected in the latest pastebin version.
http://www.chessmaniac.com/ELORating/ELO_Chess_Rating.shtml