Chess training tools we built for ourselves - then opened to everyone

PlayChessOnline.eu grew out of two passions that kept pulling us in the same direction: chess and machine learning. We wanted better training tools for our own improvement - deeper opening understanding, structured endgame practice, and practice partners that actually felt useful - so we started building them.

At the same time, we were deep into AI research, asking one question: can a neural network learn to play chess the way a human does? Not just strong moves - real style, real mistakes, real personality. Hundreds of training runs followed. Models that played nonsense, models that reverted to cold engine lines, GPUs running for days. Then one version played a move that genuinely felt... human. That was the turning point.

Once we had training tools we relied on every day and AI opponents with different configured behavior profiles, keeping it all private stopped making sense. We designed the platform, added lessons and courses - and opened PlayChessOnline.eu to everyone.

I am an independent developer who enjoys machine learning and chess. I built PlayChessOnline.eu as a long-term research and training project, testing many approaches before developing the custom models used here. Public model cards and validation reports will document what each system does, how it is evaluated, and where its limits are—without publishing the private training recipe, raw datasets, or training checkpoints.

Olivers writes the lessons, guides, and articles on PlayChessOnline.eu - someone who simply loves chess and enjoys explaining it in plain language. Every article is written to be read by a player, not skimmed by a search engine.

Each PlayChessOnline.eu character uses its own project-specific model and tuning choices to produce a different playing style.

• Practical Hunter

  Active, steady pressure. Converts advantages cleanly and grinds out endgames.

• Attacker

  Energetic, dangerous, loves initiative. Pounces the second your position loosens.

• Defender

  Calm and stubborn. Soaks up pressure, waits for your mistake, then strikes.

• Tactical Defender

  Absorbs pressure and counterpunches when your king or loose pieces wander.

• Endgame Challenger

  A patient closer who pushes you into the ending and grinds out the win.

• Beginner Teacher

  A friendly coach that plays principled chess and punishes mistakes gently.

• Opening Model

  A separate neural network trained on real opening play to guide your repertoire.

A classic Stockfish engine is also available inside the app - for moments when you want a pure analytical second opinion. But the heart of PlayChessOnline.eu is the human-like cast above.

• More human-like characters with new playing styles.
• A real online multiplayer mode, once our community is big enough.
• Deeper training modes and smarter post-game analysis.
• An optional newsletter so you can follow along without checking back.