Rookbook — Chess.com Game Tracker

v3.63 now · pocket rook15 Jul 2026

Rookbook installs like an app now. A web-app manifest and icons make it installable from the browser — Add to Home Screen on phones, an install button on desktop — opening fullscreen with its own icon, no browser bars.
And it opens offline: a small network-first service worker means the installed app always shows the newest version when online, and falls back to the cached one when there's no signal — your saved stats travel with you. Fetching new games and engine work still need a connection, as they should.

v3.62 · a second pair of eyes15 Jul 2026

Repeat visits stopped re-downloading everything. The old no-cache meta tags (a relic of local-file iterating) are gone; the host's ETag revalidation takes over — one tiny "did it change?" request per visit, a 304 when it didn't, and new uploads still appear immediately.
The home-screen icon actually works now. iOS ignores inline data-URI touch icons, so apple-touch-icon.png ships as a real file beside the page — add-to-home-screen gets the rook, not a screenshot.
og.png exists: a 1200×630 share card in the site's own palette, ready for the social-preview tags the moment the address is final.
A note for the script-less: visitors with JavaScript off now get one polite sentence explaining why a browser-only analyzer needs a browser that runs code, instead of a silent dead page.
The format dot came back to the Game Log. It only appeared when the visible list mixed time classes — every other table shows it always, so a single-format view just looked broken. Unconditional now: green rapid, blue blitz, gold bullet, on every row.
The sidebar's sample-data call-out calmed down: the green box is now a quiet two-line hint with a single accent stripe — visible when it matters, silent about itself.
Credit where due: the About box now tips its king to my chess coach, whogeeyao on chess.com.
rookbook.net is home. The social-preview tags are live — shared links now carry the rook, the wordmark, and "the stats chess.com doesn't show you." No placeholders left anywhere in the file.

v3.61 · the log shows its hand15 Jul 2026

The Game Log opens richer by default: stars, Opp #, Opp Acc, Acc ±, Undev, Swing, Eval, Clock and Avg think are all on from the first visit — the Columns button still tailors it, and anyone who already chose their own set keeps it.
About went on a diet: The columns, Things worth knowing, and Under the hood are gone — the glossary and the pages themselves explain better than a modal ever did.
A visitor counter, quietly: GoatCounter is wired in — free, cookie-less, and it sees page views and referrers, nothing else. Your games stay yours; only the visit is counted.
The demo banner got simpler: one line — "Look up a chess.com username to get started." — no exit button (fetching your games or clicking the logo already leaves the demo). And since the demo's mate scans come preloaded, no Scan button can appear there either.

v3.60 · launch prep15 Jul 2026

The front page stands alone. An empty game log ("No games loaded yet.") trailed below the welcome — a boot-order slip, not a design choice. The workspace is still preselected on the log; it just stays behind the door now.
The demo's Missed Mates page opens full. Eight hand-picked sample games ship with their mate scans baked in — one where the same mate was walked past five times before being spotted, one with two distinct misses, one where the opponent had mate against you twice and let you off — and the rest of the sample pool is pre-scanned in the background at load. No Scan button, no waiting.
Sample games stopped lying about their endings: "checkmate" is now only ever the ending when the movetext actually finishes in one, and a game whose moves end in mate names its own winner.
The footer slimmed down: the Glossary link left (the sidebar is the door — and it now stays in the phone's nav bar too), and a quiet buy me a coffee link took its place for anyone who wants to tip the jar.

v3.59 · the glossary opens doors15 Jul 2026

Glossary entries that name a feature now link to it — "Blown Leads," "Missed Mates," the sessions table, Endgame Conversion and friends are one click from their definitions, in the same green dotted style the coach uses. The Session entry goes one better: it lands directly on the sessions table, not just the Trends page.
Week by week's Time column reads like a clock: "3h 24m" instead of "3.4h" — nobody thinks in decimal hours.
Overlay leads the Elo chart's badges — the cross-format view sits first, then each ladder.
Fork shows you a fork: hover the example and a board appears — a knight landing on c7 with check, arrows on both targets. Board arrows learned to travel in pairs for it (one piece, two threats needs two arrows). And an old latent bug surfaced and died: a caption wider than the board stretched the arrow layer with it, sliding every arrow off its square — the overlay now pins itself to the grid, and long captions wrap instead of stretching the card.

v3.58 · boards for the dossier15 Jul 2026

Scout's last-10 gained Mat ± with the final-board hover — shown from the scouted player's side, so + means they finished up material.
Missed Mates one-liner now reads "Every position in which you missed delivering a game-ending checkmate," and the one-oversight rule moved out of the intro and into the tooltip of the number it explains — Mates missed — with the "eventually spotted" count folded into Mates found's tooltip.

v3.57 · half-open doors15 Jul 2026

The coach clamps past five lines: long lists show ~4½ lines with a fade — the half-visible fifth line is the invitation — and "show all N" opens the rest.
Scout's last-10 table grew up: time of day on the date, plus Acc and Clk ± — the same reads the main log offers.
Trends' twelfth panel: Time per move — rolling seconds per move, which also evens out the grid's empty corner. Read it against Accuracy: pace down while accuracy holds is confidence; both down is carelessness.
Insights gains the Both / White / Black chips — "do I tilt more as Black?" is a real question, and now one click answers it.
Conversion & swindles aligned: every row reads the same way — count, then percentage of the honest denominator. Swindles are now measured against games you were down 3+ in, and Converted against your total 3+ leads: the numbers that actually mean something.
Blown Leads and Missed Mates lead with one line beside the header: "Games where you blew a 3+ material lead." and "Every position in which you missed ending the game with a checkmate." The methodology notes stay below, shorter.

v3.56 · the full last ten15 Jul 2026

Scout shows their last 10 in full: a table under the W/L strip — date with the format dot, their color, who they played and at what rating, the opening (hover for its briefing), result, how it ended, moves, and what it did to their rating. Above it, a one-line tally of the openings seen across those games.

v3.55 · badges & briefings15 Jul 2026

The Elo chart's ladders are clickable badges again — each shows its color, current rating and game count; Overlay puts every ladder on one chart. Until you click one, the chart follows the global Time filter. A page-level dropdown was the wrong tool for a chart-level choice.
The Time dropdown starts unchecked, matching the Ending dropdown — nothing ticked means everything, so isolating one format is a single click. It also grew the same column headers (games · rating) and hint line.
Week by week grew two columns: Time — hours at the board, both clocks summed from move timestamps — and Missed, forced mates walked past across that week's scanned games.
"The Italian Game is your weapon" explains itself: opening names in coach lines carry the same hover briefing as the Openings page — the moves, the idea, how to play it.
Openings-faced tables cap at 420px with their own scrollbar instead of running the page long.
Best-session calibration note reworded: "includes games played during elo calibration, where elo swings are greater."

v3.54 · the dictionary look15 Jul 2026

Glossary reads like a dictionary now: the term sits in its own left column, the definition hangs beside it — the stacked layout made entries blur together. Stripes still cover each full entry, and it collapses back to stacked on phones.
Compare's chart panels moved up beside the table — one row, table left, bars right, wrapping on narrow screens — instead of a lonely table with the charts below the fold.

v3.53 · compare gets company15 Jul 2026

Compare grew two chart panels: "The gaps, drawn" — win rate, accuracy, castling, time losses and hung pieces as paired bars, where a 12-point gap looks like one — and "How the games end," each player's share of mates, resignations, flags and draws.
The all-green column bug: your side of the compare table was styled green-and-bold since its earliest version, which made every one of your numbers look like a "better value" highlight. Both columns read neutral now; only the winning number wears green. Columns also hug their content instead of spreading across 640px.
Panels cap at three across. On very wide screens a fourth column squeezed in and shrank every chart below comfortable reading size — Insights and Endgames now stop at three and let each panel breathe.
Blown Leads footer honesty: M✓ on a page of losses never meant "you delivered mate" — it means a forced mate was sitting on the board for you when you lost. The decoder now says so.

v3.52 · legibility round15 Jul 2026

Rated vs unrated, actually tightened: the win-rate bars duplicated the table's Win % row (an old changelog claimed they were removed; they weren't) — gone now — and the table sizes to its content instead of forcing a 640px-wide scroller into a 420px panel.
The vs-you ring can actually appear now: opponent cards cached before the flag existed never refreshed themselves — old-format recent-games data now refetches once on hover. (The ring still only shows when one of their last 10 games was against you.)
Glossary stripes: alternating row shading so the eye can track entries, and the whole page reads slightly larger than the rest of the app.
The changelog's early history joined the fold: v1.0 through v3.18 were plain text below dozens of collapsible rows — they're now one collapsible entry, "the early days," so they're findable and expand-all opens them.
Blown Leads footer restructured: the games-and-points total moved above the table; the decoder below reads one line per topic (Verdict / Eval / Peak); verdict definitions just point at the hover, where they've lived since v3.50. And "Opp Elo" replaces "Their Elo" here and in the Missed Mates and Queen-loss tables.
Scout and Compare toolbars align with the filter bar (they were inset 4px less), and text in their username fields stays clear of the ▾ caret.

v3.51 · loose ends15 Jul 2026

Stale evals can be scrubbed. On Blown Leads: "Doubt a verdict? Re-check saved evals." clears any saved engine number that contradicts the material ledger hard (3+ material one way, engine claiming fine the other) — then Evaluate positions redoes just those, fresh. Saved evals outlive the code that wrote them; now there's a broom.
Missed Mates: next to "They missed" there's now "…and you won" — games where an opponent had mate against you, didn't play it, and you completed the swindle.
Opponent names show more characters before truncating, in the game log and every other table (110→150px).
Trends: the "you're mixing formats" warning now sits right under the filter bar instead of floating below it.

v3.50 · verdicts on trial15 Jul 2026

"not lost" fixed. A resigned game could earn the verdict on the engine's word alone — and a stale saved eval could call a K-vs-K-and-3-pawns board "not lost." The material ledger must now agree (not 3+ down at the end), and the verdict's hover shows the engine's actual number so you can see its reasoning — or catch it lying.
The ending dropdown starts empty. Nothing ticked = every ending, so isolating one is a single click instead of eight unchecks. It also grew column headers (games · net Elo) and a hint line.
Scout and Compare search history got the real dropdown — same caret, hover rows, per-name remove and clear-all as the username field, replacing the browser's plain autocomplete.
Compare table: the better value is now marked on the number itself (green, bold) instead of tinting the whole cell, and the table keeps 14px off both edges.
Missed Mates: Think sits between Cost and Mat then, where it reads as part of the moment; and the page now honors the Win / Loss / Draw filter chips — the logical place for "show me only the misses that mattered."
Opponent hover: if one of their last 10 games was against you, that chip wears a ring (hover it: "this one was against you"). Applies to lookups from now on — already-cached cards refresh on the next Look up opponents run.
Glossary: Material (Mat ±) defined, with the piece values and worked examples.

v3.49 · the hover card diet15 Jul 2026

Opponent hover, redone — less going on. The rating sparkline and the stacked career bar are gone. What's left: name and meta, every ladder rating on one line (Rapid · Blitz · Bullet, the format you played them in highlighted, peak on the right), any fair-play warning, your head-to-head, their last games as W/L/D chips with the net swing, and the career record as one plain line — raw W/L/D and the win percentage.

v3.48 · honest numbers15 Jul 2026

Rating diffs that contradict the result are treated as unknown. chess.com can adjust your rating between games (fair-play refunds, mostly), and Rookbook diffs consecutive post-game ratings — so a loss could show as +8. A rated loss can't gain points and a rated win can't lose them: when the diff says otherwise, the game now shows an asterisked rating and no Elo ±, and a session whose change is unknowable shows ? with an explanation, instead of a wrong number and a phantom zero drawdown.
The Hung Pieces page is gone — built in v3.47, judged bloaty, removed. Its one keeper moved: Missed Mates' new column is now Think, the seconds you spent on the move where you missed the mate (under 3s in red — you never looked), replacing "Clock then".
The ending dropdown shows net Elo per ending next to each game count — one glance says which way of losing actually costs you.
Scout and Compare remember the names you typed — the last 8 searches appear as a dropdown on their username fields.
The Blown Leads verdicts explain themselves in place: hover "was winning", "blew a lead" or "not lost" in the table and the definition is right there.
Banned opps moved to the very end of the log's summary strip — it was sitting in front of Taken and Gave Away.

v3.47 · the follow-through15 Jul 2026

"Any ending" is a multi-select now — same checkbox panel as the Time dropdown, tick several to combine (checkmate + time, say). Each ending shows its game count.
Hung Pieces is a page (experimental). Summary strip, a coach line, the full table of gifts — piece, move, swing, seconds of thought before the hanging move, your clock at that moment — plus "When you hang them," the honest blunder-vs-time-pressure split, and a by-piece survival chart. If it doesn't earn its keep it comes back out.
The Banned ⛔ chip is gone. In its place: a "Banned opps" tile at the end of the log's summary strip — it only appears when there are any, and clicking it toggles the view down to just those games. It resets when you leave the log.
Missed Mates grew two columns — Clock then (your time right after the move you played instead: was the miss time pressure?) and Elo ±. Blown Leads grew Acc.
Sessions fix: a session's length now includes the first game itself — game timestamps are end times, so a one-game session used to read "1m" no matter how long the game ran.
Format face-off: the best value in each comparable row (win rate, accuracy, castling) is highlighted — pace and length are styles, not scores, so no crown there — and every table inside a panel now keeps 14px clear of both edges (queen autopsy's star included).
Opponent hover reworked: reads top-down by relevance — who they are (flag, member-since, career games), any fair-play warning right under the name, your head-to-head, their recent form, career totals last.
Result chips read Win / Loss / Draw, the Time dropdown's games and ratings sit in aligned columns, the scan button says "Scan N new games", scout/compare empty-states dropped the bold, the endgames grid tucked up under the filter bar, and the sidebar toggle no longer nudges 2px on collapse.

v3.46 · thirty things15 Jul 2026

The glossary reads one entry per row — the two-column layout is gone — and most entries grew a worked example ("ending 1:30 up against their 0:30 is a Clock ± of +1:00"). New entries too: Chances, Conversion, Cost, Elo*, Given back, Placement games, Rating gap, and the three Blown-Leads verdicts, spelled out.
"was winning" and "not lost" defined — on the Blown Leads page itself and in the glossary. "not lost" = you resigned a position the engine scored as equal or better; the game wasn't over, you just left.
Format face-off on Insights: Rapid vs Blitz vs Bullet, side by side — games, win rate, accuracy, seconds per move, share of your clock budget used, castling, game length. Ignores the Time filter on purpose.
World tour on Insights: your record by your opponent's flag, fed by Look up opponents.
Banned ⛔ filter chip in the marks group — one click shows every game against an account that's since been closed. Fair-play closures already carried the ⛔ on the name; now they're findable.
The Elo chart follows the Time filter — its own row of time chips is gone. One class ticked up top = that ladder; anything wider = every ladder overlaid, with a passive legend showing each ladder's current rating.
The Hung pieces module left Insights. It tried to do a page's job in one panel. The data still feeds the coach and Compare; a dedicated page is the plan.
The missed-mates coach line speaks plainly: "You missed a checkmate on 38 occasions across 355 scanned games (9% of them). This has cost you N games."
Time-class dots now mark the Date column in the Blown Leads, Missed Mates and Queen-loss tables, same as the Game Log. And Rapid, Blitz, Bullet are capitalized everywhere they're shown.
Queen loss autopsy: columns tightened (the table no longer stretches to fill the page), plus Final (hover for the board), Acc, and Elo ±.
Placement games handled: ±40+ single-game swings are chess.com finding your level, not you climbing — the Elo-per-game trend leaves them out, and a Best session containing them says so.
The long haul (Endgames) no longer counts bullet in its low-clock row — a 1-minute game ends under a minute by definition.
The Time dropdown shows each class's games and current rating.
Status lines pulse while working — fetches, evals, lookups and scans visibly think instead of looking frozen. And Stop buttons keep the button's size instead of shrinking.
Elo* explained: hovering an asterisked rating now says what it is — the post-game number, shown when the archive didn't include the pre-game one. Glossary entry to match.
Scan for missed mates is a real button now — solid green, not a ghost — and the mates and blown-leads tables grew ~200px taller.
Footer: three lines, one type size, one color, even spacing — and the Glossary link returned (phones hide the sidebar's reference section, so the footer is their door).
Sidebar no longer nudges its items when collapsed or expanded, the openings feed got breathing room off the left edge, the trends panels stopped spending a third of their height on the title block, scout and compare empty-states use matching type, and the opponent hover card got wider and airier.
FAQ: a proper write-up of how material swings are counted — settled moments, why trades don't register, and why forks are excused.

v3.45 · header, decluttered15 Jul 2026

The glossary link left the top header. The sidebar entry is now the door.

v3.44 · reference, shelved properly15 Jul 2026

Glossary moved into its own quiet section at the bottom of the sidebar — dimmed until hovered, separated from the analysis pages — and it stays out of the phone's top bar entirely (the footer link is the mobile door).

v3.43 · the glossary settles down15 Jul 2026

The glossary is a page now — its own sidebar entry, two columns on desktop, no popup. The footer link routes there too.
Every changelog entry carries its date, pinned to the right edge so they all line up.

v3.42 · reference desk15 Jul 2026

Glossary: every number in Rookbook, defined in one place — settled moments, forks, drawdowns, endgame entries, the lot. Linked from the footer.
FAQ joined the About box: free?, privacy, the accuracy "?", why stats improve after updates, lichess, rate limits.
The changelog folds: entries collapse individually (newest open), an expand/collapse-all toggle up top, and every entry is dated from here on.
Trends: the Time filter now leads the whole row, ahead of the rolling window.

v3.41 · aligned & sticky15 Jul 2026

The Time filter is sticky: set it once and it follows you across every page — visiting a page without the control no longer silently resets it.
Dates right-align in every table — sessions, blown leads, missed mates, endgames, weekly — and the best-session star sits in its own slot so it can't push the date out of line. The log's star column tightened to 16px.
The swindler line gives its denominator now ("won 29 of the 54 games you were down 3+ in — 54%"), two of the three "cheapest thing to fix" lines found new words, and the Elo chart dropped the game count from its corner — the chips already say it.

v3.40 · time takes the wheel15 Jul 2026

The Time filter now leads every filter surface — the game log bar, Trends' inline row, Openings' header, and every tab in between. It decides which chess you're looking at; everything else narrows within that.
Thin views explain themselves. Filter down to 8 blitz games and Insights now says so: most panels unlock around 10–20 games, coach lines at 15–30, and they fill in as you play.
The Elo chart's calendar mode dropped its empty volume band — the chart tightens by 30px when the bars aren't there.

v3.39 · one dropdown to rule them15 Jul 2026

Every dropdown in the app is now the same component — same mono face, same border, same hover — in two sizes: form (sidebar) and toolbar (filters, Scout, Compare, Review). Compare's odd-ones-out joined the family, and the Time multi-select matched its metrics to the rest.

v3.38 · the good night, framed15 Jul 2026

The Best session line now names the date and start time, and that session's row is starred and tinted green in the table so you can find the whole night at a glance. Worst session gets its time too — fair's fair.

v3.37 · no more "please refetch"15 Jul 2026

The stats engine now upgrades your saved games by itself. Every move-list stat is versioned; when the parser improves, Rookbook quietly re-derives peaks, troughs, endgame entries, queen losses and hung pieces from the PGNs already in your cache — no network, a few seconds, a status line while it works. The zwischenzug fix (and every earlier parser upgrade) now reaches old games automatically.

v3.36 · the zwischenzug fix15 Jul 2026

Interrupted trades no longer fake troughs. A queen trade with a check in the middle used to read as "trough \u22129" — the parser only recognized instant same-square recaptures. Captures now open a pending exchange that stays unsettled until the recapture lands (however delayed) or three plies pass with no comeback. Peaks, troughs, swings, endgame entries and hung pieces all read through the same lens now, so mid-trade phantoms are gone from every stat at once. The final position always settles — a queen hung right before a resignation still counts. Refetch to reprocess old games.

v3.35 · pocket-sized15 Jul 2026

Every popover answers to a tap. Phones can't hover, so the board tooltips, opponent cards and info cards now open on tap, pin in place, and close on a second tap or a tap anywhere else. Desktop hover is unchanged.
Boards fit the screen. Every chess board — hover boards, the review board, blown-lead pairs, opening cards — scales to viewport width under 600px instead of overflowing.
Phone polish: modals fit the screen, table cells and header rows tighten, comparison and scout-mini text sizes adjust. Wide tables still scroll sideways — that's the right answer for 30 columns.

v3.34 · the dormant tip jar15 Jul 2026

A Buy Me a Coffee link is wired into the footer and the About box — invisible until the COFFEE_URL constant gets a value. One paste activates it everywhere.

v3.33 · launch prep15 Jul 2026

Social-card meta tags: links to Rookbook now unfurl with a title and description on Discord, X, iMessage and the rest. An og:image slot is staged for when the domain exists.

v3.32 · scout, rearranged15 Jul 2026

Scout leads with the good news: How they win before how they lose, their best openings before their worst. The welcome line slimmed to one sentence.

v3.31 · credit where it's due15 Jul 2026

The coach hands out more earned greens. Four new positive-capable lines — comeback wins, bouncing back after losses, clean hands (no hung pieces), and the Elo climb — plus the display floor rose from two good lines to four. Thresholds are honest; nothing is participation-trophy green.
Scout's welcome rewritten ("jump into a game and plug in your opponent's username…") with even line spacing, and Compare suggests people you just played when no repeat opponents exist.
W/L strip hovers name the time class everywhere. The log's left edge tightened again. Three British spellings sent home.

v3.30 · a lighter left edge15 Jul 2026

Time-class dots in the log: when your view mixes formats, each row's date carries a small dot in its ladder's color — same palette as the Elo chart. One format in view, no dots, no noise.
The About folds up. Six collapsible sections; Getting started opens by default, the manual waits until asked. Nothing was cut.
The log's left edge tightened — the flag column slimmed from 26px to 16px and the first cell hugs closer.

v3.29 · small doors15 Jul 2026

A quiet line under the Elo chart points to Trends for accuracy, blunder rate and everything else over time.
Scout and Compare's opponent suggestions now refresh the moment a fetch finishes, not just when you revisit the tab.

v3.28 · order & invitation15 Jul 2026

The coach reads top to bottom now: green, then yellow, then red — the two always-on lines (blown leads, missed mates) used to tack on at the end, which is how yellows landed under the reds.
Empty pages that invite: on a real account, Scout offers the people you just played as one-click targets, and Compare lists your most frequent opponents. No sample data outside the demo.
Trends decluttered: the mixing warning lost its padding for real this time, the "each line is a moving average" caption is gone, and the filters sit right beside the rolling-window chips.
Ending filter now lives beside Result, where it belongs. Coach links switched from green text to ink with a green underline — green text kept reading as "good news" inside red lines.

v3.27 · forks, freshness & calendar time15 Jul 2026

Hung pieces learns what a fork is. When the piece that captured was attacking two or more big targets from its square, that's a tactic you walked into — counted separately, not as a hang. Needs a refetch to classify old games.
Fetch freshness fixed: the current month's archive now bypasses the browser's HTTP cache, so a game finished minutes ago actually arrives. Same-second games no longer get skipped. Opponent lookup re-queues rate-limited players instead of silently recording nothing, with a visible "slowing down" status.
Elo chart: By game / By date. Calendar mode spaces games by real time and holds the line flat through days you didn't play. The per-game result bars stay in By-game mode (where they align 1:1 with the line) and bow out in date mode.
Repertoire options get Stockfish's number — each reply under "Your options after…" shows its depth-14 eval from your side, computed once and cached forever.
Checkmate in the log's How column names the killer on hover. The Ending dropdown moved left next to the castle filter. Demo Scout/Compare show exactly one demo banner.

v3.26 · jump, scout, hunt15 Jul 2026

Review's click-to-jump fixed. The "what went wrong" moments and the eval curve carried jump targets that were never wired to anything — only the move table worked. All three jump now.
Scout remembers your last five real scouts as clickable links, and the dossier opens with their last 10 as a W/L strip (hover any letter for the story). Mini-insight labels got room to breathe and the bars stopped shouting.
Demo missed mates, for real this time. The sample-game simulator plays like a hunter now — it chases with checks, takes an available mate a little under half the time, and walks straight past the rest. The demo scan finds real oversights, and a quarter of demo games end in checkmate.

v3.25 · the scout gets a brain15 Jul 2026

Scout: Mini insights. The opponent's games run through the same lenses you get — win rate when up 3+ material, when behind, when they hang a piece, entering the endgame ahead, right after a loss, short and long games — colored against their own average. Red rows are doors.
Scout reshuffled: how they lose and how they win side by side, then their worst and best openings — each opening now carries the full hover card from the Openings page (board, first moves, the idea, how it's played).
Endgame conversion, defined and split. "Ahead" now separates +2–3 from +4-or-more, and the panel states exactly where it measures: the first quiet moment the board drops to ≤4 pieces (pawns and kings excluded), material tallied 1/3/3/5/9 from your side. Hover the header for the fine print.
Compare tightened: the table caps at 640px and gains Avg opponent Elo, Blown leads, Comebacks, Hung pieces %, Wins by mate %, and Draws %.
Week by week: dates right-aligned, and a Totals row anchors the bottom. The Elo peak marker is now a ring on the actual peak game with its label alongside. The landing page's Insights blurb is back to the original wording.

v3.24 · hung pieces15 Jul 2026

New: Hung pieces. The parser now spots the worst clean gift in every game — material captured and never won back — no engine needed. The Insights panel shows how often it happens, the five worst (piece, move, opponent, links), your win rate with and without gifts, and how long you thought before each one. Blitzed a bishop in 1.8 seconds? It'll say so. Needs one full refetch to fill in.
Elo chart marks your peak — a dashed line and label at your high-water mark on each ladder (single-class view).
Bar colors explained in-app: a legend under the coach spells out the green / yellow / red logic.
The landing page's Insights blurb fits its card again.

v3.23 · the missing fields15 Jul 2026

The real bug, found. The parser computed endgame entry material, the queen-loss ledger, and the peak-lead board for every fetched game — and then mapApiGame dropped all four fields on the floor. Endgame conversion, the Queen loss autopsy, and Blown Leads' before-boards now light up for real accounts. Re-fetch once and they fill in.
Demo mode got a full education: sample games now carry real movetext and think times, so Review, the Repertoire tree, and Missed Mates (auto-scanned) all work in the demo. Demo dossiers on Scout and Compare wear a clear demo banner — even after you've loaded a real account.
Scout reads both sides now: "How they win" (mate, resignation, or the clock) and "Their best openings" (the lines to steer away from) join the dossier.
Week by week shows the year and gains Blown, Time L and Castle columns. Sessions' accuracy column shows each session's ± against your overall.
New demo banner wording, scrollbars in house gray with a green hover, the phantom status dot no longer appears without text, and demo mate scans stay out of your browser storage.

v3.22 · alignment & honesty15 Jul 2026

"+61.7" explained and tamed. The math was right — Stockfish reports huge centipawn scores in completely won positions it hasn't proven mate in — but the display now caps at ±20⁺, with the raw figure in the hover.
Rated vs unrated moved to Insights, where it always belonged (it had been quietly rendering at the bottom of Blown Leads). When every game in view is rated, it says so instead of vanishing. Insights also gains the Rated/Unrated filter.
Records lined up. Cleanest games, trophies and biggest swings now share fixed columns — opponent Elo, accuracy, date — and the view links wear the house green.
Openings header split in two: grouping, window and filters on one line; the engine pass and its status on their own line below. The "Last" date shows its dotted underline so you know it opens the game.
Tilted chip retired from the log filters (the tilt column and insights stay). Hover boards grew to 54px squares. About lists every page on its own line. Footer signs off with a simple "Contact".

v3.21 · a round of feedback, front to back15 Jul 2026

The knight is a horse now. Proper muzzle, jaw, upright ears, eye and nostril — the old one read as a chicken (charming, but wrong).
Demo mode covers Scout and Compare. Open either page in the demo and Rookbook conjures a sample opponent to scout and a sample rival to compare against, clearly labeled.
Insight bars tell the truth in color: green when a split beats your overall win rate by 3+, red when it trails by 3+, yellow between. Accuracy split gains Rated/Unrated rows.
Endgame conversion works in the demo — sample games now carry entry-material data. Blown Leads shows when the peak happened (mv 14/38). Scout's dropdowns are labeled and default to the time class you play.
Review tidied: the stats strip runs four across, and the move table stops stretching across ultrawide screens. Charts: day-of-week, game length and accuracy-by-length share one row, and the length chart's x-axis reads 10-19, 20-29, …
Coach headers enlarged, the trends mixing note lost its padding, and footers no longer hug the sidebar.

v3.20 · filters, everywhere they belong15 Jul 2026

Filter by time class anywhere. A new Time dropdown in the filter bar — tick one format for a clean single-ladder read, or several to combine. It follows you to Charts, Insights, Trends, Openings, Blown Leads, Missed Mates and Endgames, and the mixed-ladder warning on Trends now points straight at it.
Filter labels unified: Both → All, Any → All. The game count on the Elo chart's chips is readable when the chip is selected.
Openings keeps its own filters. Rated/Unrated and Time now sit in the page's header row, and the "filters don't apply here" note is gone. Trends' inline filters moved toward the center of the row.
Coach links wear the house green instead of the browser's default purple.
The About box caught up with the app — all thirteen pages, the engine passes, demo mode, and what's saved where.

v3.19 · navigation & polish14 Jul 2026

Elo chips show your rating. The time-class picker on the Elo chart now reads "rapid 829 (20)" — the current rating on each ladder, games in parens.
Module headers enlarged everywhere — chart titles, Sessions, Week by week, Missed Mates, Blown Leads, Openings, and insight cards all step up to the same readable size.
Trends filters moved home. Both/White/Black and Rated/Unrated now sit right beside the rolling-window chips (now 10 / 25 / 50); the explanatory note and the detached filter bar are gone from that page.
The coach always talks about Blown Leads and Missed Mates — a count and a link when there's something to see, credit when there isn't, and a nudge to scan if you haven't. Any coach line that mentions endgames now links straight to the Endgames page.

v1.0 → v3.18 · the early days→ 14 Jul 2026

v3.18 · readability round

New chart: Accuracy by game length. Your average accuracy bucketed by how long the game ran, with the dashed average line. Sliding down to the right is fatigue; a red left edge is quick blunder-fests.
Coach lines alternate a subtle background stripe — long insight lists are much easier to scan.
The sidebar's "see sample data" hint is now a proper green call-out card instead of a whisper.
W/L by day of week and game length distribution grew taller.

v3.17.b

Staunton pieces, take two. The flat geometric set read as too abstract — the boards now draw rounded staunton-style pieces in the chess.com vein: charcoal black pieces, warm-white white pieces, proper collars and coronets. Same set on every board and every browser.
Endgame conversion refuses to lie. Games restored from an old cache don't carry entry-material data, and the panel was quietly falling back to the biased final-material number — which is why it still said 94%. It now shows a re-fetch prompt instead, and only ever computes from honest data.
Sidebar footer removed.

v3.17 · autopsies & honesty

Queen loss autopsy on the Blown Leads page. Every analyzed game where your queen was captured, judged on the material ledger across the moment it fell: fair trades are excused, gifts are listed — move, what took it, the swing, and how the game went. Re-fetch to fill it in.
Endgame conversion got honest. "Ahead on material" is now measured when the position first simplifies, not at the final position — the old number was survivorship (games that end with you ahead were already won).
Trend badges stop screaming. The change readout now compares against the first full rolling window, so a single game at the start can't produce a "-100%" artifact.
Castle mates are their own category in How Mates Happen, pinned next to pawn mates on the trophy shelf.
Opponent hover drops the join year.

v3.16 · the big round

Compare, properly. Pick a range and a time class right on the page — both sides get the same window, so it's apples to apples. And the Compare button becomes a working Stop while fetching.
Two new trends: Blunder rate (rolling % of games you were winning and lost anyway) and Missed mates per game (after a scan). These are your improvement metrics, normalized per game.
Week by week gains Drawdown, Avg Opp, and Avg Moves columns.
Openings: the Last date now links straight to that game, and the engine-move hover names whose move it is ("engine plays for White").
Pawn mates always show in How Mates Happen (even at zero — aspirational), and castle-mates are credited to the rook.
Opponent hover shows the country written out (Germany, Scotland) — the flag emoji didn't render everywhere and is gone.
Sidebar gains a footer with the version and a changelog link. Footer text reordered under the copyright.

v3.15 · a round of feedback

Elo chart with few games no longer stretches three points across the whole width or bleeds bars into the y-axis — small samples cluster compactly on the left.
Sessions: lengths over 59 minutes read as hours (1h 15m), and hovering a W/L/D letter now shows the opponent, move count, accuracy, and how it ended.
Blown Leads reordered: Verdict → Lost by → the material arc (Peak / Final / Given back) → Eval. Missed Mates gains the opponent's Elo and puts Result up front.
Tilt score is now four conditions — the 11pm–3am clause is gone. Whether late-night chess hurts you is a question for your data, and the "late-night chess is costing you" insight still answers it when it's true.
Country flags in the opponent hover card (after "Look up opponents"), and opponent names got a visible dotted underline so you know they're hoverable.

v3.14 · feeds & polish

Before & after boards in Blown Leads. Hover the Peak column to see the board at the moment of your biggest lead; hover Final for how it ended. The distance between the two is the story.
Blown Leads and Missed Mates now show the game's time next to the date. Both feeds run newest game first, and when one game has several missed mates, the latest miss leads.
Demo mode says it once — the banner got bolder, the duplicate status line is gone.
Insights section headers (Psychology, How games end, …) and card text are bigger and easier to read.

v3.13 · charts, sharpened

Win-rate bars. W/L by hour and by day of week now scale bars by win rate, with a dashed line marking your overall average — green days beat it by 3+, red days trail it. Hover any bar for the full W/L/D record.
Accuracy by hour gets the same dashed average line, and both hour charts are taller.
Elo over time now spans the full page width.
Streaks recentered. The zero line sits in proportion to your best and worst runs, so one monster win streak no longer squashes the loss side flat.
Game length and rating gap swapped places.
Unreviewed games show ? in the accuracy columns — hover to see why: chess.com only computes accuracy once you run Game Review there.

v3.12.b

Try before you type. The landing page now has a "Peek at a sample account" link that opens demo mode straight to Insights.

v3.12.a

One voice on resignations. The material-based "resign too early" coach line now yields entirely once any resignation has an engine eval — no more two lines double-counting the same habit. When no evals exist yet, it hedges honestly and points you to Evaluate positions.
Landing page copy rewritten.

v3.12 · the coach gets organized

Coach box sorted good→bad so the wins lead.
Percentages added to the "a of b" counts, weekdays spelled out in full.
Coach lines now link straight to Missed Mates and Blown Leads where relevant.

v2.9–v3.11 · the big build

Stockfish evaluation. Run the engine on every game's final position — powers the flags column, premature-resignation verdicts, and engine-verified thrown-away games.
New panels: Missed Mates, Blown Leads, Endgames, Openings tree, Trends, Compare players, Scout an opponent, and a game Review board.
Demo mode. Click any menu item before loading a username and Rookbook shows you around with sample data.
Flagged-games filter, column picker, day-of-week and accuracy-by-hour charts, and a long list of smaller fixes.

v2.8.e

Get in touch. Bug reports and suggestions to rookbookchess@gmail.com — on the landing page, in the footer, and in the about box.
The mascot leads the landing page: logo and wordmark side by side, at a size where you can actually see him.

v2.7 · the rook shows up

A logo. The rook and his book, cropped from the source art, recolored for dark mode, and embedded in the page — so it survives being uploaded anywhere. Also the favicon and the iOS home-screen icon.
Fixed: a fragment of the old favicon was rendering as stray text at the top of the page.
Fixed: clicking the logo flashed the front page and bounced straight back — the link was fighting its own click handler.

v2.6 · cleanup and a bug review

A regression suite. 30 tests, each one a bug that actually shipped: the queen-trade "+9 lead", the inverted checkmate, the triple-counted missed mate, the sawtooth eval, the walkovers dragging down averages. They exist so those can't come back quietly.
Conventions written down. Four separate bugs came from getting perspective or sign wrong. The rules now live at the top of the file, and everything refers back to them.
Dead code removed, duplicate functions removed, thresholds collected in one place, and silent failures now say something instead of quietly producing a wrong number.
Endgames fleshed out, with hover definitions for what actually counts as a rook endgame, a pawn endgame, and the rest.

v2.5 · review, properly

Fixed the blunder count, twice. First it judged a move by the eval before and after it — which measures search depth, not the move, and flagged 1…Nc6 as a mistake. Then it counted moves in an already-lost game as fresh blunders. Now each move is judged against the engine's choice, and once a game is decided it stops counting.
Bigger board, an eval curve where you're always the bottom half, hoverable per-move segments, and the moves as a table with what each one cost and how long you took.

v2.4 · addressable

Every tab and every game review has a URL. Middle-click to open in a new tab, bookmark a game, share a link, use the back button.

v2.1–2.3 · a front door

A landing page. Newcomers used to arrive at fourteen tabs of somebody else's sample data.
Blown Leads got its own tab — it was promised on the front page and buried at the bottom of Insights.
Fewer columns by default, filters that label themselves, a sticky table header, and a reset button so people can clear their data.

v2.0 · Position

Paste a FEN or a PGN and the engine gives you the top three moves, with arrows on the board.
Fixed: hovering a table of opponents fired dozens of API requests and got you rate-limited — which is why the buttons sometimes stopped responding.

v1.5–1.9 · the engine gets useful

Openings tell you what to play. Stockfish works out the best move in each of your openings, names the line it belongs to (the Austrian Attack, the Alapin), and gives you the plan behind it.
Full game review with an eval bar, an eval curve, and blunder marks — with your think time beside each one.
Nothing is lost on refresh. Games, engine analysis, mate scans and opponent lookups all persist, and fetching only pulls what's new.
Scout — point it at an opponent and get a game plan, each item with a "how".

v1.1.c

Openings tables relabelled: they show what your opponents played against you, not what you chose. An opening is named after both sides' moves.

v1.1.b

Opening explainers. Hover any opening for its moves, its idea, and how to play it. Plus a "what you're actually playing" briefing with boards.
Renamed to Rookbook.

Earlier

Scout — point it at any opponent and get a game plan: do they crack on the clock, do they tilt, which openings beat them.
Save fetch — caches your games with engine evals and mate scans, so reopening loads instantly.
Columns — show or hide any of 30+ data points; notes and stars per game.
Missed Mates — every forced mate you walked past, with the board and what you played instead. No engine needed.
Tempo — per-move think times: do you blitz your winning positions?
Review — step through any game, with move times color-coded.
Endgames, Sessions, Trends, Compare, Charts, Insights — tilt curves, blown conversions, clock wars, heatmaps.
Engine evaluation of final positions, run in your browser.

Bugs worth remembering

Peak material counted mid-exchange snapshots — a queen trade read as "+9 lead". Now only settled positions count.
Checkmate sign was inverted, so every loss-by-mate looked like a win you threw away.
Missed mates were counted once per position, not per oversight — one blunder read as three.
Rating-gap buckets used post-game ratings, which baked the result into the bucket.
The Elo chart drew rapid, blitz and bullet as one line, inventing cliffs that never happened.

A closer look at your chess.com games.

Glossary