Chess DNA Research
We analyze real games with a real engine and publish what we find, including the parts that make our own product look less impressive.
Everything here comes from games players actually lost, run through Stockfish at depth 18, not from coaching folklore. Two pieces are original research on our own analysis data: what separates one rating band from the next across 6,288 games, and an honest look at whether an improvement app moves anyone's rating. Two are deep dives on the mistakes that come up most often in that data, throwing away won positions and repeating the same blunder. Where a number is small or a sample is thin we say so in the article rather than in a footnote.
Most chess advice is a strong opinion with no denominator behind it. We publish the denominator. Every claim on this page traces back to a specific set of analyzed games, and each article states its sample size, its engine depth, and what it cannot conclude. If a result did not survive us trying to break it, it is written up as a result that did not survive.
If you want the same treatment on your own games instead of the average player's, Chess DNA analyzes your recent Chess.com or Lichess games free and shows which patterns are costing you the most rating.
Original research
Our own analysis data, with the method and the limits stated up front.
- Do Chess Improvement Apps Actually Work?App efficacy
We measured our own users, then attacked the result five ways before publishing it. What survived the attacks, and what we will not claim.
- What Each Elo Level Struggles With6,288 games, 103 players
Stockfish depth 18 over 6,288 real games. Blunder rates by rating band, what changes at each level, and why two players at the same rating rarely share a top weakness.
The patterns behind the numbers
The two mistakes that show up most in the data, and what to do about each.
- Why You Keep Throwing Away Won GamesConverting wins
Converting a won position is a separate skill from reaching one. Where the wins actually leak, and the drill that plugs it.
- Why You Keep Blundering in ChessBlunder patterns
Most players blunder for the same three reasons. How to tell which one is yours, and how to stop repeating it.
Frequently Asked Questions
Where does the data in these articles come from?
From games imported by Chess DNA users from Chess.com and Lichess, analyzed move by move with Stockfish 17 at depth 18. Every article states how many games and how many players sit behind its numbers. Nothing is simulated and no games are generated for the purpose of an article.
Is this independent research?
No, and we say so on every page. Chess DNA builds a chess analysis product, so we have an obvious interest in the conclusion that analyzing your games helps. That is exactly why the efficacy piece spends most of its length trying to break its own headline number instead of defending it. Read the limitations section of each article before quoting it.
Can I cite or reuse these findings?
Yes. Link back to the article you are citing and quote its sample size along with the number, since a rate from 6,288 games and a rate from 28 users are very different kinds of evidence. If you want a breakdown we have not published, or the method behind one, email yuval@chessdna.app.
How often is this updated?
When there is a new result worth publishing, not on a content calendar. Existing articles are revised in place when more games change a number, and the updated date at the top of each one reflects the last real revision.
