Stats — Scarlett's Super Secret Chess Trainer

⏱ Time on task

	Today	This week	This month	All time
Drills	—	—	—	—
Mystery	—	—	—	—
Total	—	—	—	—

Active engagement time: tab visible + input within 15 s. Covers all drills, races, reviews and mystery. Granularity: 1 minute.

⏱ Time by drill

	This week	All time

Per-drill active minutes from the heartbeat. Tracking for the newer drills (Visualise, Sharpen, Radar, Gauntlet, GTM, …) starts 2026-06-18 — earlier minutes weren't recorded per drill.

Weekly progress (last 12 weeks)

Week	Loading…
Loading…

Each cell shows attempts · solve rate. — = no activity. Drills in cycle order.

Flash Recall — recall over time 🧠

Week	Level	Done	Recall %	Pieces placed
Loading…

Pieces placed (correct / shown) is the truest progress signal — accuracy % drops at a level-up just because the board has more pieces, even when she's recalling more. Watch the correct count climb. Current level is bold.

⏳ Post-solve engagement (last 28 days)

Time from move submitted → Next clicked. Longer = more reading/thinking. Target: mostly 5–30s. Under 5s = skipping feedback; over 30s = analysis mode.

Drill	Count	Avg time	<5s	5–30s	>30s
Loading…

Calculating vs guessing — Winning/Equal/Losing reads

Drill	Reads	Matched engine	Right move, wrong read

"Right move, wrong read" = she played a good move but misjudged the resulting position — the sign she's finding moves by feel rather than calculating to a conclusion.

Today's cycle — corrects per block

How many correct answers she needs in each drill before the cycle rotates to the next one. The cycle runs continuously in order D3 → D5 → D2 → D8 → D1 → repeat. There's a silent safety cap of 2× the correct-target so she's never stuck on a bad day.

Total per cycle: 0 correct answers

🥈 Silver: 0 correct (2 cycles)

🥇 Gold: 0 correct (3 cycles)

Cycle 2 race config

Settings for the accuracy-at-ceiling race that fires after every even-numbered drill cycle. Time cap is a failsafe only — it never blocks cycle progression.

Target correct (10–25)

Min puzzle rating

Max puzzle rating

Time cap (minutes)

Show solution on wrong

🎯 Launch accuracy race demo → Opens the main app and starts a cycle 2 accuracy race immediately

👑 Convert & Survive — per-mode stats

Mode	Attempts	Passes	Pass rate	Level / detail
Loading…

🗺️ Opening Gauntlet — repertoire coverage

The bot plays the real Lichess field, so it no longer forces coverage — this shows Cold-test performance appears first, followed by which prepped lines she's actually faced, the ones she's never met (drill those via “Prep an Opponent”), and the lines she's currently failing.

Loading…

⚠️ Do it together

Every position that needs working through together, in one list — auto-flagged after 3 fails, failed Training-Game reviews, or whenever Scarlett taps "I don't understand". She can't see a flagged position in her drills until you solve it together; it clears automatically when she gets it right.

📊 Training Games — report

Teaching-log diagnostics: what triggered an interruption, what she fixed and which patterns recur. This card is not a strength estimate; the immutable cold/coached tracker below owns that job.

🤖 Training Games — mistake diagnosis

Every move she blunders in a Training Game, logged by type and by how long she thought first. After ~10 games this says whether she's moving too fast or looking-and-not-seeing.

📈 Independent-move transfer tracker

Fixed-node analysis of her original move before any interruption. Cold and coached games are never pooled; corrections and abandoned games never enter the strength denominator. Download the full PGN archive: ⬇ all games (PGN).

🎯 Original decisions vs real 5+0 cohorts

Real rated-standard Lichess 5+0 decisions, analysed with the identical engine contract. Raw and position-adjusted rates are shown separately; no rating is inferred when coverage or cohort ordering is insufficient.

Rewind vs play-on

She always plays on from a caught mistake (fight the worse game) — unless it leaves her worse than −2 pawns, in which case she takes it back (rewind) once she's found the right move.

Enhancement flags

All flags are off by default. Scarlett sees no changes until you enable a flag here. Use Lee's profile to test each feature before enabling for everyone.