New ELO test

Hi, I'm a Computer Science (& Math) student and I really like chess. I like doing small projects on the side of my studies to practice my programming. I would like to make an online ELO test. I know there is one that exists and I like it but with no offense intended it seems to be a little bit outdated. I have all the tools I'll need, but there's one thing that I'm missing: positions.

Here is my idea: I give the person a test, they choose the number of problems they have to solve (the more they solve, the more accurate the test). The test gives them a position, and they make the next move. Can someone confirm that this is a good methodology for doing what I'm trying to? For the record, I am NOT a good chess player. I just really like chess and computer science so I want to do something that incorporates both.

Advice is appreciated! Thanks

I think that the trick is that you'd need to give the positions ratings, and then consider a pass/fail equivalent to a win/loss against the posistion and simply run the ELO calculation on that basis.

The hard part is initially rating the positions.  I think ideal would be to have many players with known ratings complete the test and then run the ELO calculation in the other direction (adjusting the rating of the position as opposed to the player) in order to get some kind of reasonable measure.

I believe this is effectively the methodology chess.com takes with its tactic trainer, but both player and postion are re-rated after each attempt.

You'll want a big variety of positions, and test them on a wide range of players before the test can be as accurate as it can be.

Although in practice chess doesn't require equal amounts of defense, offense, calculation, endgames, openings, strategy, tactics, sacrifice, etc.  So even a very well put together test would not be accurate.  E.g. the actual rating of a player that plays the exchange variation of everything would probably depend more on endgames than attacking.

These tests can be fun though, no need worrying about how accurate they are.  I remember one rating me at >1900 back when I was probably <1500.  Although less exciting, you can also consider the output being strengths and weaknesses instead of a rating.

My idea for scoring the problems would be to put it into several engines with different ratings. I agree having hundreds or thousands of strong players would help with accuracy. Once I create a basic beta version, I will post it here to allow people to test it and tell me where it could improve