Chessalyz May Update: Matty Got Specific
Matty's AI coaching now backs up his recommendations with real variations from the position. Three before-and-after examples inside.
Quick role-swap this month. Varun usually writes the Chessalyz updates because he handles the technical side while I focus on chess, but May’s headline improvement is on the coaching quality of Matty (our AI coach), so I’m taking this one. Varun will be back next month with the engineering deep-dives.
If you want to join the Chessalyz community, here’s the Skool group: https://www.skool.com/chessalyz-5973/. We’re over 110 members and growing!
What’s new
🎯 Matty’s feedback is more specific and concrete
📉 Fewer board visualization errors in the analysis
🔢 Matty now references real variations from the position
🐛 More bug fixes from your reports
The big one: Matty actually shows you the lines now
This is the update I’m most excited about. The previous version of Matty would give you reasonable advice, but it leaned on general principles and sometimes recommended moves without backing them up with concrete lines. That’s the chess coaching version of “trust me bro.” It’s also exactly the thing I push back on with my own students. The specific position overrides the general principle, and the way you prove a move is good is by showing what happens after it.
The new version of Matty does three things better:
The recommendations are more specific to the actual position on the board.
He references one or two short variations to back up his suggestions.
He has far less made-up moves (wrong piece on the wrong square, lines that don’t make sense, etc.).
The easiest way to show this is side by side. Here are three positions from real Chessalyz analysis sessions, with the older Matty 4.5 feedback next to the new Matty 5.0 feedback.
Position 1:
Black has a queen on g4 and a bishop on f3 generating threats. The student played Kf1 and said they “wanted to run their king away from the kingside”. Reasonable instinct, but there’s a real question about whether the king move is even necessary.
Matty 4.5:
New Matty 5.0:
The older version suggested Qf1 or Rxc6 and described them in terms of “reducing checking lanes” and “fighting back with activity.” Fine in the abstract, but the student doesn’t get to see what those moves lead to. The new version gives two full lines: Rxc6 h5, Qf1 h4, Nc5 and Qf1 h5, Rxc6 h4, Nc5 which both lead to the same position but through different move orders. Same ideas, but now the student can actually see the move order and the resulting position. The student is also given a question to ask themselves “What exact threat an I stopping, and what do I expect Black to play next if I do nothing?”
Position 2:
Different game, Black just played …Qd6 “to defend the e6 pawn”.
Matty 4.5:
New Matty 5.0:
The older answer mentions …Qe8 and …Nd4 as alternatives, which is fine. The new answer reframes the whole question (”Can I defend with tempo, or create a threat?”) and then backs up each suggestion with a real line: …Qe8, Nxf8 Kxf8, bxc3 Qh5 and …Nd4, Qe3 c2, b4 Nf5. The framing question is the part I really like, because it teaches the student how to think about defense in general, not just what to do here.
Position 3:
The student played Kf1 and said they “wanted to get their king off the g-file”. Reasonable instinct, but there's a real question about whether the king move is even necessary.
Matty 4.5:
New Matty 5.0:
The older answer pointed at exf4 and said “set up the position and play a few moves after exf4 to see how it changes piece activity.” Not bad advice, but pretty generic. The new answer explains why Kf1 doesn’t actually solve the problem (the pawn on g2 already blocked the file), gives the concrete continuation exf4 Rxf4, Ne1 Qf7, f3 and offers a second candidate move (b5) with its own line b5 axb5, cxb5 Bxf3, Bxf3. That’s the kind of answer a real coach would give in a lesson. The student is also given another question to ask themselves with “Can I hit back in the center or remove the pawn that is driving the attack?”
Why this matters
The reason I care so much about this upgrade is that at times high level ideas without concrete lines makes things less useful. Students don’t improve as much from being told “develop your pieces” or “play actively.” They improve more from seeing the moves that prove the point, and then internalizing the pattern. Concrete lines are what turn a guess into a calculation.
We’re not all the way there yet. Matty will still occasionally suggest a move that doesn’t quite work, and his variations are short rather than deep. But the gap between the older version and the new version is the biggest jump in coaching quality we’ve shipped so far, and the direction is the right one. I hope you enjoy the upgrades!
What’s next
We’re still working on:
Helping organize your Flashcards in well structured decks
Chessalyz curated decks for targetted practice in specific areas
Ability to add Flashcards from any game position
Micro demo videos for various functions of Chessalyz
How you can help
The thumbs up / thumbs down feedback we shipped in April is doing real work behind the scenes. Every rating helps us figure out where Matty is solid and where he’s still shaky.
If you’re using Chessalyz and Matty gives you an answer that’s off, please rate it. That’s how the coaching gets better.
🐛 Report bugs
👍 Rate Matty’s responses
💡 Suggest features
Try the new analysis on a recent game of yours and let me know what you think. The before-and-after on these three positions tells the story I’d been hoping to tell for months, and I’d love to hear whether the experience matches it for you.
Happy analyzing!
Dalton & Varun
Co-founders, Chessalyz.ai
Looks like a wonderful upgrade. Any chance of a future version avoiding run-on sentences? In the last example, "the knight lands on d4" is a new independent clause, and would ideally be preceded by a period or possibly a colon, rather than a comma.