Which moves go to the Blunders tab, and which to the Tactics tab?

This page answers one question: why is this move on this tab? It has three layers. Read the first one if you want the idea. Read the second if you want the exact rules. Read the third if you want to check a rule against a real game.

The same rules run in three places: the X-ray Coach Blunders tab, the X-ray Coach Tactics tab, and the website's Game Review. A test holds them to the same cards on real games (src/lib/server/tabApiParity.test.ts). The examples in layer 3 come from that same test data, so they can never say something the code does not do (src/lib/server/whichTabDoc.test.ts).

For the full blunder definition (thresholds, tiers, the 2-move window, what is excluded), see What counts as a blunder. This page only decides the tab.

Layer 1 — The idea, in plain English

This is our current definition. If we change the rules, we change this text.

We look at every move you played in a rated blitz, rapid or classical game, once our own engine analysis of that game is complete. We do not judge a move when you were already lost: seven pawns down or facing a forced mate. For each move we ask two questions.

Did the move make the position worse, for a concrete reason? The concrete reasons are: you hung a piece, you lost material, you walked into a mate in two or fewer, you allowed a stalemate, or you missed a tactic that wins inside two moves. Lost material must be worth more than one pawn. In the endgame one pawn is enough.

• If the move cost a lot (30 winning-percentage points or more) and there is a concrete reason, it goes to the Blunders tab.
• If the move cost a lot without a concrete reason, it is a big mistake. It counts in your statistics, but it gets no Blunders card: there is nothing concrete to show you. It can still be on the Tactics tab.
• If the move cost a little (at least half a pawn, but less than 30 points), it goes to the Blunders tab only when you gave material away and the best move would have saved at least half of it, when you walked into mate, when you missed a mate in one or two, or when you missed a tactic that wins material and you were not still clearly winning afterwards.

Was there a tactic on the board? If koga names a tactic for the best move (a fork, a skewer, a deflection, a checkmate pattern…), its line wins at least one pawn more than your move (T8), and you did not play it, the move goes to the Tactics tab. If koga names no one-move tactic for the best move, apart from a hanging piece, but the opponent's reply was a tactic that wins at least one pawn, the move goes to the Tactics tab too, marked "punished".

A move can be on both tabs. A missed fork that cost you the game is a blunder, and it is a missed fork.

Three things the rules deliberately leave out, and one rule about the clock:

• A hanging piece is not a tactic. A piece that was simply left en prise is a Blunders-tab matter. It is never listed on the Tactics tab. A free piece you missed gets a Blunders card only when the move cost a lot, or when you were still winning and started no trade. Otherwise it gets no card.
• One-move pawn grabs are not tactics. A pawn you could have taken in one move, or a pawn you gave away in one move, is often intentional. The Tactics tab never lists it. In the endgame, a pawn you gave away in one move can still be a blunder. A missed tactic worth exactly one pawn goes to the Tactics tab only.
• Winning positions stay quiet. When a small drop missed a tactic but you were still clearly winning after your move (three pawns or more, by the engine), we hide it. You may have simplified on purpose. Two exceptions. A missed mate in one or two always goes to the Blunders tab, when the mate is gone after your move. A free piece you missed without starting any trade is shown: on the Blunders tab when koga sees only a hanging piece, on the Tactics tab when koga names a tactic.
• The clock. Every card shows how much time you had, when the game recorded clocks. We list moves played in time trouble like any other move. The X-ray report's "Time pressure" switch changes your scores, not the Blunders or Tactics lists. One exception remains: the Blunders drilldown and the blunder counts in the X-ray report still leave out time-trouble moves unless the switch is set to Included. On the website, use the clock filter to hide them.

Layer 2 — The rules

Written in STE100: one fact per sentence, active voice. "You" is the player whose games are reviewed. "The engine" is Stockfish's evaluation. "Koga" is our tactic classifier. The rule numbers match the last column of the table in layer 3. The rules are checked in the order R2, R3/R4, R5, R6, R7, R8. The table names the first rule that decided the move, and the same code that decides the card writes that column.

Which moves are judged

R1. We judge your own moves only. We judge a move only when the engine evaluated the position before and after it. The game must be rated, standard chess, and blitz, rapid or classical.

R2. You cannot blunder in a lost position. If the engine already rated your position at 7 pawns down or worse, or the opponent already had a forced mate, we do not judge the move.

Big drops — the move made the position much worse

A move is a big drop when your winning chances fell by 30 winning-percentage points or more.

R3. A big drop with a concrete reason goes to the Blunders tab. The concrete reasons are: you hung a piece, you lost material, you walked into a mate in 2 or fewer, you allowed a stalemate, or you missed a 2-move tactic. A material reason must be worth more than one pawn. In the endgame one pawn is enough. A mate or a stalemate counts whatever the material. If koga also names a listed tactic (T1 to T5 and T8) for the best move, the move goes to the Tactics tab as well.

R4. A big drop without a concrete reason is a big mistake. It does not go to the Blunders tab. If koga names a listed tactic (T1 to T5 and T8) for the best move, or for the opponent's reply, the move still goes to the Tactics tab.

Small drops — the move made the position somewhat worse

A move is a small drop when it lost at least half a pawn but less than 30 winning-percentage points. Small drops were added on 2026-09-18 after GM Avo's votes.

R5. A small drop goes to the Blunders tab in two cases only. Case one: you gave material away, and the best move would have saved at least half of it. Walking into a mate in 2 or fewer counts as case one. Case two: koga names a tactic you missed that is worth more than one pawn (see R8), and you were not still clearly winning afterwards (see R7). In case two the move goes to the Tactics tab as well, when the tactic is a listed one (T1 to T5 and T8). A small drop that fits neither case is not shown. The table calls that "no concrete reason". We check case one before R6. A move that hung a piece and also missed a mate shows R5 in the table when the best move saved half, and R6 otherwise. The card is the same. When the engine gives no line for the best move, we keep the card. When it gives no line after your move, we do not. The table still says "saved half" or "lost it too" then. We read the line after your move for 2 moves. When a check or a capture continues past that, we read on to the end of the checks and captures (at most 6 more moves). A sacrifice that wins the material back is then not counted as lost (OKQBZpLA#44 22...Qg5, 2026-09-23). The best line is always read for 2 moves only.

R6. A missed mate in 1 or 2 always goes to the Blunders tab. This holds even when you were still winning afterwards. It does not hold when you still had a forced mate after your move: then you lost nothing. The move goes to the Tactics tab as "Checkmate" when koga names the mate. The Blunders card does not depend on koga: it reads the engine's mate.

R7. If koga names a tactic you missed, but you were still clearly winning after your move (3 pawns or more, by the engine), we hide the move. You may have simplified on purpose. There is one exception. If the best move won a free piece (3 points or more, at least 3 more than your move, and your move started no trade) we show it: on the Blunders tab if koga sees only a hanging piece, on the Tactics tab if koga names a tactic. A table row with rule R7 or R8 and an empty Tactics column is a tactic that koga named but the tab does not show: T2 to T4 or T8 removed it, or it lies deeper than one move and the tab shows one-move tactics by default.

R8. A missed tactic worth only one pawn goes to the Tactics tab only. One pawn is not a blunder outside the endgame.

What the Tactics tab lists

T1. The Tactics tab lists tactics that koga names. Koga looks at three moves: the stored best move, the move you played, and the opponent's reply to your move.

T2. "Hanging Piece" is never listed. It belongs to the Blunders tab. A pawn that one plain capture wins or loses is never listed as a tactic. It can still be a blunder in the endgame (R3, R5). A fork or a discovered attack on a pawn is a tactic and stays.

T3. Some of koga's labels are switched off because we do not trust them yet: Counter Threat, Windmill, Bombarding (koga's "bombardinio"), plain Promotion, Preparatory Move and Favorable Exchange.

T4. If your move and the best move show the same tactic on the same squares, the tactic does not explain your mistake. It is not listed. When koga gives no squares for a tactic, the same tactic name is enough.

T5. If nothing is left for the best move, but the opponent's reply was a tactic, the move is listed under that tactic and marked punished. The card then shows the opponent's reply. Exception: when koga named one of the switched-off labels of T3 for the best move, the reply is not listed either. We checked this on four real games on 2026-09-22 and listing the reply would have been wrong in all four. In kUard1pt move 21, 11.Qxd4, the reply Qa5+ looks like a fork of king and knight, but Nc3 parries both. The position is lost because White was a piece down and needed Bb5+ to keep the pressure. No card can name that. A reply is listed only when its line wins at least one pawn inside the 2-move window, read from the position after your move. Mate is always listed. That is the T8 floor applied to the reply, so kUard1pt move 21 is not listed: Nxc3 bxc3 is an even trade. A reply line that is missing, too short or cannot be read never removes an entry: the reply is still listed.

T6. A move played with fewer than 20 seconds on the clock is listed like any other move. The card shows the clock. Use the "Time pressure" switch (X-ray report scores) or the clock filter (website) to hide such moves.

T7. The "Pawn targets" switch on the X-ray Tactics tab hides tactics that only work because of a pawn. It is a viewing filter. The website lists these tactics, except the one-move pawn captures of T2.

T8. A tactic on the best move is listed only when that line wins at least one pawn inside two moves, counted net of what the played move wins in the same two moves — a recapture, a promotion or the same pawn taken with another piece is not a missed tactic (9jk1nY7l#41 Rxc5 vs Bxc5). Mate is always listed. A best move that wins nothing — a move that saves a draw, a defensive move, a positional move — is not a Tactics entry, even when koga names a motif on it. If the opponent's reply was a tactic, the move can still show through T5, under the same one-pawn floor. The line is read over two moves; a stored line shorter than that is re-searched first, and a line that cannot be read never removes an entry.

Layer 3 — The rules on real games

Five real games from the local corpus, with koga's real answers. The rows are a selection of the moves the rules flagged: at most three per rule, one per game. The Rule column names the rule from layer 2 that decided the move. Click the game link to see the position.

A selection of the five games' flagged moves — at most three boards per rule, one per game — the member's side at the bottom. red is the move you played, green is the stored best move.
The evaluations, best moves and tactic names below are all koga's own analysis (Stockfish 18, 100,000 nodes). It is deeper than Lichess but not ground truth: across these five games the two disagree on 13 positions that hold a real forced mate, and koga misses 8 of them (Lichess misses the other 5). Every one of the 13 was confirmed at depth 30. Treat the green arrow as "what koga suggested". Whatever these five games carry, production serves a MIX: as of 2026-09-22 some games rest on koga’s evals and some on Lichess’s, so a member’s own board may rest on either source. The mix moves as the queue works through the backlog — measure it, do not quote a number from here.

Of the rules a board can name, no move in these five games shows these:

• R2 already lost
• R3 allowed a stalemate
• R4 big mistake, tactic listed
• R5 walked into mate
• R7 free piece, hanging
• R7 still winning, hidden
• T8 best move wins no pawn