Every number Rookbook shows, in plain English — what it means, and how it's worked out.
These are the statistics behind Rookbook, a free chess.com game tracker that runs
entirely in your browser.
How the numbers are made
Material uses the standard count: pawn 1, knight/bishop 3, rook 5, queen 9.
Material is read only at settled moments. Rookbook replays every move and samples material only when no exchange is in progress: after a capture, nothing is counted until the recapture lands, or three plies pass with no reply.
That's why a queen trade doesn't register as a −9 crater followed by a +9 spike — the middle of a fair trade never gets counted. Peak, trough, swing, given back, hung pieces and blown leads are all built on this one ledger, so ordinary trades don't register — only real material changes do.
Material lost to a fork is excused separately: that's a tactic you walked into, not a piece left hanging.
Some formats sit out some stats. Results, openings and material count every game equally. A few analyses exclude formats that would distort them.
Bullet games sit out the time-pressure stats, because everyone is short of time in bullet — though filtering to bullet alone brings them back, since the comparison is fair again.
Daily games sit out think-time averages, where "seconds per move" is really days, and anything session-based — tilt, fatigue, sittings — because correspondence games don't happen in a sitting. Wherever an exclusion applies, the chart or panel says so.
Small samples are held back. Anything that splits your games into buckets needs a minimum before it will draw a conclusion: openings need three games before they appear, and several Insights panels stay quiet until there's enough to say something.
A 100% win rate across two games isn't a strength, and Rookbook tries not to dress it up as one.
Your numbers can change after an update. When the stats engine improves, Rookbook re-derives everything from your saved games automatically.
The newer numbers are the more honest ones — the changelog spells out what changed each time.
chess.com's Game Review score (0–100). It only exists once a game has been reviewed on chess.com — that's why some rows show "?".
An 85 means your moves tracked the engine's preference closely. Everyone's blitz accuracy runs lower than their rapid — compare within a format, not across.
On Missed Mates: how many of your moves the same mate stayed available for. More than one means you walked past it repeatedly, but it still counts as a single oversight.
Mate appears on move 20 and you finally play it on move 23 — that's one missed mate, three chances.
A win from 3+ material down — a game you stole.
You were down a full rook at your worst moment and won anyway: that's a comeback, and the ☆ Comebacks filter on the Game Log collects them.
On Missed Mates: mates you played divided by mates you had.
80% conversion means that when a forced mate sat on the board, you delivered it 8 times out of 10.
On Missed Mates: Stockfish's verdict right after the move you played instead of the mate. A big plus (or #) means the miss was harmless — still winning. Near zero or negative means that one move threw the win away on the spot.
The biggest fall from a high point inside a week or session — how far it dipped, even if it recovered.
A session in which you lost 25 Elo to start, then gained 100 back, still has a drawdown of 25.
Elo*
A rating marked with an asterisk is the post-game number: chess.com's archive didn't include the pre-game rating for that game, so the log shows where you landed instead of where you started.
Your material edge at the moment the board first drops to 4 or fewer pieces (pawns and kings not counted), measured outside of exchanges. This is what Endgame Conversion buckets on.
Entering at +2 means you were two pawns' worth up when the board thinned out — an endgame you're supposed to win.
Stockfish's verdict on the final position, from your side, in pawns. Displays cap at ±20⁺ — beyond that it's simply "completely winning/lost."
Fork
The piece that captured was attacking two or more big targets from where it landed. Material lost to forks counts as a tactic you walked into, not a piece left hanging.
A knight lands on c7 hitting your king and rook: whatever it wins next move is fork material, and it's excluded from the Hung-piece count.
The piece-value ledger, in pawns: pawn 1, knight and bishop 3, rook 5, queen 9 — kings aren't counted. Positive means you're up, negative down, always read at settled moments.
Mat +3 means you're up a minor piece, or three pawns. Mat −5 is a rook's worth down. Trading your queen for a rook reads as −4: nine out, five back.
A checkmate-in-one you had and didn't play. A mate that stays available for several moves counts as one oversight, not several.
Note that having mate available means you were already winning — so a miss usually costs moves, not the game. The Cost column on Missed Mates shows the exceptions.
Your best and worst material during the game, sampled only at settled moments — mid-exchange spikes (like the middle of a queen trade) don't count.
Peak +4, trough −2: at your best you were up a minor piece and a pawn, at your worst down two pawns.
Placement games
On a fresh ladder, chess.com moves your rating in outsized jumps (±40 or more per game) while it finds your level. Rookbook treats those swings as calibration, not skill — the Elo-per-game panel on Trends leaves them out, and sessions containing them are flagged.
On Trends: each point is the average of the last N games rather than one game's result, so the shape of a change shows through the noise. The window picker sets N.
On a 50-game window one disaster moves the line by about a fiftieth — a run of them bends it. Small windows react fast and jump around; large ones are steadier but slower to admit something has changed.
A run of games with only short breaks between them (no 60-minute gap). Sessions get their own table on Trends, with net Elo and the streak you quit on.
Games at 9:00, 9:20 and 9:45 are one session; a game the next morning starts a new one.
Settled moment
A position with no exchange in progress: after a capture, nothing samples until the recapture lands or three plies pass with no comeback.
How far material went against you before the game settled: the smaller of your best and worst moments. A game you led wire to wire scores 0, because nothing ever changed hands; up 8 with a one-pawn stumble scores 1; up 5 then down 4 scores 4.
Being crushed scores 0 as well — if you were behind from the first mistake to the last move, nothing ever changed hands either. So a 0 means one-sided, not good; check the result to know which way. The high numbers are blown leads and comebacks: up 6 and finishing 5 down is a swing of 5, and so is clawing back from 5 down to win by 6.
How many of your minor pieces never left their starting squares. A development report card.
Undev 2 in a 35-move game means a knight and a bishop watched the whole thing from home.
Three ways to lose a game you shouldn't have. "blew a lead" — you were up 3+ material at some point and lost anyway (no engine needed). "was winning" — the engine had you at +1.0 or better when the game ended, and you still lost: usually a resignation or a flag in a winning position. "not lost" — you resigned a position the engine didn't score as lost for you (better than −1.5, and the material ledger agrees). The game wasn't over; you just left.