Sample report: what RookSac shows for 300 real games
Before you run an analysis of your own, this page shows what a RookSac report contains. Every number below is computed from three sample reports bundled with the site: 100 recent rated games each for MagnusCarlsen, hikaru and GothamChess, fetched from Chess.com's public API and analyzed with Stockfish 17.1.
Last updated: September 29, 2026
What these samples are
The three reports are ordinary RookSac reports. They were made the way yours will be: the games were downloaded from Chess.com's public API, and each was analyzed in the browser with Stockfish 17.1 Lite at depth 14 (see how the analysis works).
- MagnusCarlsen: 100 rated games, 82 blitz and 18 rapid, played Aug 12–Sep 15, 2026.
- hikaru: 100 rated blitz games, played Sep 10–Sep 15, 2026.
- GothamChess: 100 rated blitz games, played Sep 9–Sep 15, 2026.
They are public games, used here only to show what a report looks like. Nothing on this page is a judgement of these players; a sample of 100 fast games says little about anyone's chess. The same reports are bundled with the app, so you can explore them interactively: open RookSac's Home screen, where the samples are listed by default, or restore one from its sample menu.
The report at a glance
| MagnusCarlsen | hikaru | GothamChess | |
|---|---|---|---|
| Games analyzed | 100 | 100 | 100 |
| Average rating (Chess.com) | 3,254 | 3,368 | 2,892 |
| Average opponent rating | 3,044 | 3,007 | 2,838 |
| Record (win–draw–loss) | 66–20–14 | 82–9–9 | 54–16–30 |
| Average accuracy | 90.4 | 89.3 | 84.9 |
| Median game accuracy | 91.4 | 89.5 | 84.9 |
| Lowest–highest game | 66.8–100.0 | 73.7–99.5 | 69.5–100.0 |
| Opponents' average accuracy | 86.7 | 84.1 | 82.3 |
| Average centipawn loss (ACPL) | 17.5 | 19.0 | 27.3 |
| Average game length (moves) | 48 | 44 | 39 |
The first thing to notice is how far from 100 these numbers are. Players rated above 2,800 on Chess.com average 84.9–90.4 accuracy in these games, and their opponents average 82.3–86.7. Accuracy is not the share of moves that were correct; it is an average of how much winning chance each move gave away, and fast games against strong opponents push it down. That is why every RookSac game also shows the opponent's accuracy: the useful comparison is against the people you actually play. The article What chess accuracy really means explains the calculation.
How the moves were labelled
Each move gets one label, from Brilliant to Blunder (the rules are on the move classifications page). The bars show the share of each player's moves that fell into each group. Book moves, which are known opening theory and not scored, are left out of these percentages.
| Per 100 scored moves | MagnusCarlsen | hikaru | GothamChess |
|---|---|---|---|
| Best or better (Brilliant, Great, Best) | 54.0 | 53.1 | 48.1 |
| Inaccuracy | 2.7 | 4.3 | 4.8 |
| Mistake or Miss | 2.1 | 2.0 | 2.7 |
| Blunder | 0.5 | 0.4 | 1.0 |
| Inaccuracy or worse, in total | 5.3 | 6.7 | 8.5 |
Even at the top of the rating list, between 5.3 and 8.5 scored moves in every 100 were an inaccuracy or worse. Your own numbers will be larger. The figure to watch is the error rate (inaccuracies, mistakes and blunders per 100 moves), because it is the number you can actually push down. RookSac also breaks it out by phase, time control and colour.
Where the accuracy is lost
| MagnusCarlsen | hikaru | GothamChess | |
|---|---|---|---|
| Opening accuracy | 95.0 | 93.9 | 89.5 |
| Middlegame accuracy | 87.0 | 85.9 | 81.7 |
| Endgame accuracy | 93.7 | 88.6 | 87.8 |
For all three players the middlegame is the lowest of the three phases, about 8 points below the opening. Two things are going on. Book moves count as fully accurate, which lifts the opening figure. And the middlegame is where positions are sharpest, with the most tactics and the most decisions that have no clear answer. Check whether your own report has the same shape: the phase breakdown is a good place to start when you look for what to study.
Openings
The report lists every opening you played with its record and average accuracy. These are the four most played opening families in each sample.
MagnusCarlsen
| Opening | Games | W–D–L | Avg. accuracy |
|---|---|---|---|
| Sicilian Defense | 23 | 13–5–5 | 91.2 |
| Queen's Gambit Declined | 11 | 7–4–0 | 92.5 |
| French Defense | 8 | 4–2–2 | 85.8 |
| Nimzo-Indian Defense | 8 | 6–2–0 | 89.3 |
hikaru
| Opening | Games | W–D–L | Avg. accuracy |
|---|---|---|---|
| Sicilian Defense | 22 | 17–3–2 | 88.2 |
| Nimzo-Larsen Attack | 14 | 13–1–0 | 90.8 |
| Queen's Pawn Game | 8 | 8–0–0 | 89.1 |
| French Defense | 7 | 5–2–0 | 87.0 |
GothamChess
| Opening | Games | W–D–L | Avg. accuracy |
|---|---|---|---|
| Trompowsky Attack | 15 | 7–3–5 | 84.0 |
| English Opening | 12 | 7–2–3 | 87.4 |
| Scandinavian Defense | 11 | 6–2–3 | 87.1 |
| Queen's Pawn Game | 9 | 6–1–2 | 85.2 |
Samples this small are prompts, not conclusions: a dozen games can't prove that an opening works or doesn't. The improvement plan uses the same idea but looks for an opening you play often where results or accuracy lag behind your other openings, and links each suggestion to the games where it happened.
Winning positions: converted or not
RookSac counts a game as having reached a winning position if the engine rated it at least two pawns in the player's favour at some point. It then checks whether that game was won.
| MagnusCarlsen | hikaru | GothamChess | |
|---|---|---|---|
| Games that reached +2 or better | 73 | 85 | 62 |
| Of those, won | 65 | 78 | 52 |
| Conversion rate | 89% | 92% | 84% |
Conversion is one of the more actionable numbers in a report, because a lost or drawn game that was once winning is a concrete, reviewable event: you can go to the moment the advantage disappeared.
Time on the clock
Most of these games are 3-minute blitz. Among the 9,980 scored moves played by these three players in 3-minute games, 15% were made with 30 seconds or less left on the clock, but those moves account for 22% of the inaccuracies-or-worse and 25% of the mistakes, misses and blunders. Put another way, the error rate was 10.6 per 100 moves when they were short of time and 6.6 per 100 moves otherwise.
| MagnusCarlsen | hikaru | GothamChess | |
|---|---|---|---|
| Moves made with 30s or less left | 20% | 7% | 18% |
| Share of all errors made then | 25% | 15% | 25% |
3-minute games only; errors are inaccuracies, mistakes, misses and blunders.
Your own report shows the same split, and for a 10-minute game or a rapid game it tells you at what point on the clock your play starts to slip.
One critical moment
A report is built from moments like this one. It is from a 3-minute game in the hikaru sample: Black (hikaru) to play on move 24, with 1:33 on the clock.
RookSac labels 24...Rc5 a Blunder: it took Black's winning chances from about 66% to 35%. The engine preferred 24...Rxc2!, a rook sacrifice. After 25.Qxc2 b3 the pawn attacks the queen, and after 26.Qc4 Qa2+ 27.Kc1 Rc8 the queen is pinned to its own king on the c-file and Black's attack goes on; Stockfish rates that line at about two pawns in Black's favour. The move played, 24...Rc5, cost roughly thirty percentage points of winning chance, and he spent only about 6 seconds on it.
Black went on to win the game anyway. A large swing does not decide a game by itself, and the report doesn't treat it as if it did. It ranks moments by what they cost, and the improvement plan looks for what repeats across many games, not for one-off shots. In your own report, each such moment comes with the played move, the engine's line, and a written explanation of what changed in the position.
What to do with a report like this
- Start with the dashboard. Compare your accuracy and error rate with your opponents'.
- Find the leaky phase. Check whether opening, middlegame or endgame costs you the most.
- Read the improvement plan. It turns the pattern into specific openings, motifs and endgames to study.
- Train on your own mistakes. The positions you misplayed become puzzles with hints and explanations.
- Repeat with new games. A month later, compare the same numbers and see what moved.
The guide walks through each screen in detail, and How to review your chess games gives a 20-minute routine for single games.