WHAT HAPPENS WHEN THE ANALYSIS ENGINE IS UNREACHABLE

By the ChessDiary team

ChessDiary's move-by-move analysis normally runs on Stockfish — the same open-source engine used in professional preparation. But Stockfish runs on a separate server, and any server can occasionally be slow to wake up, overloaded, or briefly unreachable. When that happens, ChessDiary doesn't show you an error and ask you to try again later. It quietly switches engines instead.

Here's exactly what changes when that happens, and why the fallback exists in the first place rather than just failing outright.

1. WHY THE SERVER GOES QUIET IN THE FIRST PLACE

The Stockfish engine runs on a separate server, not bundled into the app itself — a full engine installation is too heavy to ship inside a mobile app or load in a browser tab. Servers like this are often configured to spin down during quiet periods to save on hosting costs, which means the very first analysis request after a period of inactivity can take longer to wake the server up — occasionally long enough to trigger the fallback.

2. WHAT TRIGGERS THE FALLBACK

When you tap Analyze, ChessDiary sends your game's PGN to the Stockfish server and waits up to 90 seconds for a response. If that request fails for any reason — the server is cold-starting, temporarily down, or simply doesn't respond in time — ChessDiary catches the failure automatically and re-runs the same analysis through Gemini AI instead, using the same PGN. From your side, this happens as one continuous action; there's no separate retry step to trigger manually.

3. WHAT ACTUALLY DIFFERS BETWEEN THE TWO ENGINES

Stockfish's numbers come from genuine engine search — it calculates millions of positions to arrive at a centipawn evaluation and a best move, the same way engine-assisted opening prep works. Gemini's analysis, by contrast, is a language model's assessment of the game: still built from real chess understanding and still useful for spotting mistakes and blunders, but not a brute-force search, so its centipawn estimates and best-move suggestions are less precise than Stockfish's. Both return the same report structure — quality labels, comments, motifs — so nothing about how you read the results changes, only how they were produced.

4. HOW TO TELL WHICH ONE RAN

After analysis finishes, a short confirmation tells you which engine actually produced the report — Stockfish or Gemini AI. If you get a Gemini AI result and want a Stockfish-grade evaluation instead, simply re-running the analysis later, once the server is responsive again, will overwrite it with a fresh Stockfish-based report.

5. EVEN A STOCKFISH REPORT USES GEMINI

The two engines aren't entirely separate even on the normal path: after Stockfish finishes its evaluation, ChessDiary sends the result to Gemini a second time specifically to tag tactical motifs — forks, pins, hanging pieces — on top of Stockfish's raw quality and centipawn data. So a full Stockfish-engine report is really Stockfish's search plus Gemini's pattern recognition layered on top, not one engine working in isolation.

6. WHY FALL BACK INSTEAD OF JUST FAILING

A dedicated chess engine server costs money to run continuously and can go quiet under low traffic to save on hosting. Rather than let that show up to you as a broken feature, ChessDiary treats Gemini as a genuinely usable second option — not a perfect substitute for engine search, but accurate enough that a temporarily unreachable server doesn't mean you walk away without an analysis at all.

In practice, most analysis requests complete on Stockfish without any fallback ever triggering — this mostly comes up on the first request in a while, or during a brief spike in server load. It's a safety net, not the common path.