Where did you get a huge database of 1.7M+ chess problems?
All positions were calculated from the scratch by special algorithms we developed.
Those algorithms written at most on Python and use chess games available freely on the net to find
positions we think are interested enough for man to solve.
Can I get this database?
By now, not. We will update you about that. All our service uses centralized server to get blunders from
and we are working on unified API to communicate with server. We will publish this API when it will be stable.
Can you describe this algorithm in words?
When player makes a mistake in the even game which costs him at least 2 pawns or equivalent material,
we move this position to special database. After that, another algorithm checks if this position is beautiful
enough to solve. We don't rely on some static rules like mate in N or sacrifices, but instead of this use
Stockfish score as only criteria. We calculate principal variation a player must find, when all his moves
are completely forced. When forced moves ends, we cut the variation.
Sometimes I dissagree with given solution. Can you explain this?
If you feel like that, check your solution with modern engine. At most cases you will understand your mistake.
In a very rare cases, algorithm indeed gives not optimal solution. The reason for that is the time limit, algorithm
have for each positions to analyze. This time is long enough for most cases, but for some positions
it is not true. Those positions are usually with rating 2000 or above. You can read about
Horizon effect to understand this phenomenon.
In the future, as service will grow, we will recalculate those postions to get more accurate lines.
What are the rating?
When new player signs up, he receives an Elo rating of 1500. Chess problems also have a rating calculated by
special algorithm. Easy positions are rated 1200, extremelly hard are about 3000. That all give us a scale.
When user solves a position right, his rating gets up, if he fail, his rating gets down. The greater the
difference between the ratings of the player and the position, the greater the change in the ratings in the case
of success or failure.
Please read aboute Elo rating to understand the math behind.
Sometimes I get positions far away from my rating. Why you give me so hard positions to solve?
Indeed, player should receive positions not too difficult and not to easy. This functionality is
ready to be implemented on the server and probably will be introduced in the near future.
I want more features.
We have a lot of ideas to where we can extend this service, you should'nt tell us about. Contact us
if you really want to participate in Open Source project.