What's costing ▚▚▙♥︎╱╯ games
The short version
- ▚▚▙♥︎╱╯'s costliest tactical blind spot is not punishing the opponent's mistakes: 47 times in 34 games, costing about 5.2 games' worth in practice.
- The slow-burn habit that costs ▚▚▙♥︎╱╯ most is passive or misplaced piece moves (50 times in 42 games). Nothing hangs, but the position turns.
- 88% of ▚▚▙♥︎╱╯'s mistakes came in sharp positions, where most moves lose, which were 58% of the positions: there, 15% of moves were mistakes; elsewhere, one move in 36 was. ▚▚▙♥︎╱╯ took a median of 8 s on them, against 7 s on the others.
- A player about 300 points stronger would have avoided only about 22% of what ▚▚▙♥︎╱╯'s mistakes cost: most are hard for them too. What they've most left behind is pawn moves that weakened the position.
- Of 30 losses, 9 were one-move oversights; 16 turned on a single move that wasn't an obvious blunder, and 4 came from several mistakes adding up, with no single decisive move.
- ▚▚▙♥︎╱╯ had a winning position (80%+) in 16 games that weren't won.
- At the move that decided each game ▚▚▙♥︎╱╯ didn't win, there was still a median of 7:49 left on the clock, and the move took 6 s. There's time to spend on the moments that matter.
- ▚▚▙♥︎╱╯ had a forced mate in 5 or fewer 36 times and let 6 go, 1 of them mate in 1. Opponents let 2 of 8 go against ▚▚▙♥︎╱╯.
- ▚▚▙♥︎╱╯'s accuracy falls the more games played in a day: about 84% in the first game, 74% by the 5th.
- Opponents let 34% of ▚▚▙♥︎╱╯'s mistakes go unpunished, so results understate how often these patterns happen.
▚▚▙♥︎╱╯'s pre-move check
Not a calculation routine. These are the few questions that would have caught most of ▚▚▙♥︎╱╯'s costly mistakes, built from ▚▚▙♥︎╱╯'s own games.
- After each opponent move, also scan what they just left loose: the checks and captures on offer, and their unguarded pieces.
- Before letting go of a piece: is it protected where it lands, and what was it guarding where it stood?
- When something is free, spend ten seconds on why. Look at their best reply to taking it.
- Before trading, ask whose piece is better and which endgame the trade heads into. Trade freely when ahead, carefully when not.
What separates ▚▚▙♥︎╱╯ from the next level
Stockfish says what was best. It can't say what a stronger person would have done, and most of the best moves it finds aren't ones people at any club level find. Maia-2, a model of how people move at each rating, can. Here each of ▚▚▙♥︎╱╯'s mistakes is set against how likely a player about 300 points stronger would have been to make it, in the same position.
Of what ▚▚▙♥︎╱╯'s mistakes cost, a player about 300 points stronger would have avoided about 22%, and one about 500 stronger 31%. Most of these mistakes are ones stronger players make too, in the same positions. What they've learnt is narrower, and it's where to start:
| Blind spot | Mistakes | Avoided 300 points up |
|---|---|---|
| Pawn moves that weakened the position | 28 | 34% |
| Trades that helped the opponent | 19 | 32% |
| Not answering what their last move threatened | 17 | 24% |
| Letting winning positions go | 38 | 24% |
| Passive or misplaced piece moves | 50 | 20% |
| Endgame technique | 31 | 20% |
| Taking material that was bait | 15 | 19% |
| Not punishing the opponent's mistakes | 47 | 15% |
Near the bottom: not punishing the opponent's mistakes. Stronger players still make these nearly as often, so they're the slower ones to improve.
Which level ▚▚▙♥︎╱╯'s moves look like
▚▚▙♥︎╱╯'s moves are most like those of a player rated about 1,430 on chess.com, against the 1,750 ▚▚▙♥︎╱╯ was rated on average in these games. That's which moves are chosen, not how well: a player's moves can be ahead of their results, or behind them.
Maia-2's ratings are lichess's. They're put on chess.com's scale by ▚▚▙♥︎╱╯'s opponents: the level their moves look like, against the ratings they had.
The opening is clearly ahead of the middlegame: the middlegame is where ▚▚▙♥︎╱╯'s play looks most like a lower-rated player's.
Where the next level would have played differently
Mistakes that are common at ▚▚▙♥︎╱╯'s level and much rarer 300 points up.
Nothing tactical, but Qg3 kept a much better position. ▚▚▙♥︎╱╯'s win chance went from 55% to 39%.
A player about 300 points stronger would most often have played Qg3 (40% of the time). Qxf6 is a 40% choice at ▚▚▙♥︎╱╯'s level, and 15% for them.
Qh4# was available (16.Qh4#), and was missed. ▚▚▙♥︎╱╯'s win chance went from 100% to 84%.
A player about 300 points stronger would most often have played Qh4# (71% of the time). Nf7+ is a 31% choice at ▚▚▙♥︎╱╯'s level, and 23% for them.
17...Qxf6 recaptures, and the exchange loses material: the knight on f6 wasn't safe there. ▚▚▙♥︎╱╯'s win chance went from 74% to 49%.
A player about 300 points stronger would most often have played Bc5 (57% of the time). Nxf6+ is a 33% choice at ▚▚▙♥︎╱╯'s level, and 23% for them.
Trends
▚▚▙♥︎╱╯'s 100 games from Mar 16, 2026 to Nov 7, 2026, oldest on the left. Single games swing a lot, so each chart compares the earlier 50 games with the latest 50, and only calls it a change when the gap is bigger than game-to-game luck would explain.
Rating
risingHere the line is ▚▚▙♥︎╱╯'s rating after each game, the dots are the opponents' ratings, and the strip underneath shows wins, losses and draws. The rating moves about 9 points a game, so over 100 games a drift of about ±52 happens by luck alone. This change is bigger than that. What ▚▚▙♥︎╱╯'s results were worth against these opponents (performance rating): 1780 in the earlier 50 games, 1964 in the latest 50.
Accuracy
no clear changeHow close the moves came to the engine's, per game. The opponents (grey) averaged 80% and then 78%.
Serious mistakes per 100 moves
no clear changeMoves that cost 10 points of win chance or more, out of the real decisions made. From 1458 moves in the earlier half and 1319 in the latest.
Threats not answered
no clear changeOf the times the opponent already had a threat on the board, how often the reply ignored it. From 220 threats in the earlier half and 230 in the latest.
Winning positions converted
no clear changeOf the games that reached 80%+ win chance, how many went on to be won. That's 34 games in the earlier half and 45 in the latest.
Time per move
slowerThe median thinking time per move in each game (after the first five moves).
Does the time of day matter?
▚▚▙♥︎╱╯'s games split by when they started, in the time zone this report was made for (America/Toronto). Games before 4am count as the night before. Accuracy is the steadiest guide: results and rating also depend on who ▚▚▙♥︎╱╯ was paired with.
Later in the day, ▚▚▙♥︎╱╯'s accuracy rises from about 80% in games around 6am to 81% around 3pm. That's well within luck: no real difference.
Game after game, ▚▚▙♥︎╱╯'s accuracy drops from about 84% in the first game of the day to 74% by the 5th. Luck alone would make a change that big less than 1 time in 20, so it's real.
| Games | Accuracy | Vs other games | Score | Rating | |
|---|---|---|---|---|---|
| Time of day | |||||
| Morning (4am to noon) | 78 | 81% | +1.0 | 65% | +194 |
| Afternoon (noon to 5pm) | 13 | 76% | −4.7 | 73% | +52 |
| Late (9pm to 4am) | 9 | 85% | +4.4 | 78% | +50 |
| Day of the week | |||||
| Monday | 11 | 77% | −4.2 | 64% | +22 |
| Tuesday | 13 | 80% | −0.2 | 65% | +41 |
| Wednesday | 13 | 72% | −9.3 | 77% | +67 |
| Thursday | 20 | 86% | +6.6 | 72% | +89 |
| Friday | 22 | 80% | −0.3 | 68% | +68 |
| Saturday | 13 | 82% | +1.4 | 58% | +1 |
| Sunday | 8 | 85% | +4.3 | 56% | +8 |
| Games into the day | |||||
| First game of the day | 48 | 83% | +4.9 | 70% | +156 |
| 2nd or 3rd | 33 | 81% | +1.3 | 74% | +143 |
| 4th to 6th | 14 | 73% | −8.5 | 36% | −27 |
| 7th to 10th | 5 | 71% | too few games | 80% | +24 |
Tactical blind spots: one move that turns the game
Recurring kinds of tactical mistakes, ranked by what they cost in practice: the score they took off ▚▚▙♥︎╱╯, against opponents of the level ▚▚▙♥︎╱╯ meets, who don't see everything (1.0 ≈ one game's worth). The engine's count, as if every mistake were punished, is beside it. A mistake can appear under more than one heading.
Not punishing the opponent's mistakes
The opponent left something hanging or allowed a tactic, and the next move played on without taking it.
Practise on lichess: Seize the advantage Hanging pieces
Rxd2 was available (18...Rxd2 19.Bc4 Rxd1 20.Bxf7+ Kg7 21.Raxd1), and was missed. ▚▚▙♥︎╱╯'s win chance went from 93% to 13%.
Nxg2 was available (18...Nxg2 19.d4 exd3 20.Kh1 Rg6 21.Qf5), and was missed. ▚▚▙♥︎╱╯'s win chance went from 95% to 38%.
Qxd5 was available (8...Qxd5 9.Qxd5 Nxd5 10.O-O-O Nc7 11.Bh4), and was missed. ▚▚▙♥︎╱╯'s win chance went from 89% to 42%.
Leaving pieces unprotected
A move put a piece on a square it couldn't hold, walked away from a piece it was guarding, or opened a line, and the opponent simply took material.
Piece moved: knight 50%, bishop 25%, queen 8%.
Practise on lichess: Hanging pieces
20.Qxc4 wins the queen on c4: it moved to a square it couldn't hold. ▚▚▙♥︎╱╯'s win chance went from 92% to 4%.
20...Bxe7 wins the knight on e7: it moved to a square it couldn't hold. ▚▚▙♥︎╱╯'s win chance went from 74% to 9%.
The move opened the line for 10.Bxd8, which wins the queen on d8. ▚▚▙♥︎╱╯'s win chance went from 65% to 6%.
Taking material that was bait
Something that looked free was captured, and the capture walked into a tactic.
Practise on lichess: Deflection Attraction
The move opened the line for 10.Bxd8, which wins the queen on d8. ▚▚▙♥︎╱╯'s win chance went from 65% to 6%.
8...Qa5+ forks the king and bishop. ▚▚▙♥︎╱╯'s win chance went from 62% to 13%.
17...Rd8 pins the bishop to the queen. ▚▚▙♥︎╱╯'s win chance went from 60% to 14%.
Forks: one piece hitting two at once
The punishment was a double attack, most often a knight or queen landing on a square that hits two pieces at once (or the king and a piece).
Punishing moves included 5 knight moves, 3 long-diagonal moves, 2 backward moves.
Practise on lichess: Forks
17...Qxf2# forks the king, rook and knight. ▚▚▙♥︎╱╯'s win chance went from 84% to 0%.
8...Qa5+ forks the king and bishop. ▚▚▙♥︎╱╯'s win chance went from 62% to 13%.
15...Nxf2+ forks the king and queen. ▚▚▙♥︎╱╯'s win chance went from 52% to 4%.
Also seen
| Pattern | Times | Games | Cost in practice | Engine's count |
|---|---|---|---|---|
| Not answering what their last move threatened The opponent's previous move set up a threat (to win material or to mate), and the reply didn't deal with it. That isn't being out-calculated: the threat was already on the board. | 17 | 15 | 2.2 | 5.9 |
| Recapturing on autopilot After a capture, the piece was taken back automatically, when an in-between move, a different recapture or a bigger capture was much better, or the recapture itself walked into trouble. | 11 | 11 | 2.0 | 3.0 |
| Quiet moves not seen coming The move that punished it was not a check or capture. It was a quiet move that set up an unstoppable threat. These are the hardest to see because nothing looks forcing. | 11 | 11 | 0.7 | 4.1 |
| Walking into attacks on the king The move allowed a forced mate or a mating attack. | 2 | 2 | 0.5 | 1.4 |
| Pins, skewers and discovered attacks Pieces were lined up on a file, rank or diagonal with an enemy rook, bishop or queen behind or in front, and the opponent used the line. | 4 | 4 | 0.4 | 1.4 |
| Pieces getting trapped A piece ended up with no safe squares and was rounded up. | 2 | 2 | -0.0 | 0.6 |
Positional habits: the slow-burn kind
Mistakes where nothing hangs, but the position turns. These build the 'no-win situation' that only shows once it's too late.
Letting winning positions go
The position was clearly winning (75%+), and a single move gave a big part of it back.
Piece moved: bishop 21%, knight 21%, pawn 18%.
28...Qc5+ 29.Kf1 Qxd6 30.Qe4+ f5 31.Qe7+ wins material. ▚▚▙♥︎╱╯'s win chance went from 97% to 6%.
20.Qxc4 wins the queen on c4: it moved to a square it couldn't hold. ▚▚▙♥︎╱╯'s win chance went from 92% to 4%.
17...Qxf2# forks the king, rook and knight. ▚▚▙♥︎╱╯'s win chance went from 84% to 0%.
Endgame technique
Serious mistakes once most pieces were traded.
Practise on lichess: Endgames Rook endgames
The bishop on d6 was already under attack; 44...Rxd6 takes it. The threat already on the board was Rxd6. ▚▚▙♥︎╱╯'s win chance went from 86% to 14%.
The bishop on f3 was guarding the rook on g2; once it moved, 35.Kxg2 took it. ▚▚▙♥︎╱╯'s win chance went from 96% to 50%.
Nothing tactical, but Kf3 kept a much better position. ▚▚▙♥︎╱╯'s win chance went from 50% to 7%.
Passive or misplaced piece moves
Knight, bishop, rook and king moves that didn't hang anything but handed over the initiative: retreats to passive squares, a piece wandering away from the action, the king stepping into the open.
Piece moved: bishop 34%, knight 30%, rook 22%. 6 were retreats.
Nothing tactical, but Kf3 kept a much better position. ▚▚▙♥︎╱╯'s win chance went from 50% to 7%.
Nothing tactical, but Kh1 kept a much better position. ▚▚▙♥︎╱╯'s win chance went from 50% to 14%.
Nothing tactical, but Rd5+ kept a much better position. ▚▚▙♥︎╱╯'s win chance went from 50% to 19%.
Trades that helped the opponent
Exchanges started or offered that made things worse: a good piece swapped for the opponent's bad one, a trade into a lost endgame, or lines opened toward one's own king.
Nothing tactical, but Rd8 kept a much better position. ▚▚▙♥︎╱╯'s win chance went from 97% to 31%.
Nothing tactical, but exd3 kept a much better position. ▚▚▙♥︎╱╯'s win chance went from 68% to 25%.
Nothing tactical, but O-O-O kept a much better position. ▚▚▙♥︎╱╯'s win chance went from 85% to 58%.
Also seen
| Pattern | Times | Games | Cost in practice | Engine's count |
|---|---|---|---|---|
| Pawn moves that weakened the position Pawn moves that lost nothing right away but left holes behind, loosened the king's cover or fixed the pawn structure badly. Pawns can't move back, so the damage stays. | 28 | 22 | 1.8 | 5.4 |
| Queen adventures Queen moves that didn't hang anything but cost ground: the queen leaving the defence, going pawn-hunting, or landing where the opponent's pieces could chase her and develop with gain of time. | 11 | 11 | 1.0 | 2.6 |
Missed mates
Every time either side had a forced checkmate in 5 moves or fewer, and whether it was seen through. A mate played out move by move counts once, and as missed if it was let go at any step. The opponents play at ▚▚▙♥︎╱╯'s level, so their misses show which mates are easy to overlook: the ones to look for in ▚▚▙♥︎╱╯'s own games.
How often each side missed
| Mate in | ▚▚▙♥︎╱╯ missed | Opponents missed |
|---|---|---|
| 1 | 1 of 7 (14%) | 1 of 3 (33%) |
| 2 | 1 of 7 (14%) | 0 of 2 (0%) |
| 3 | 2 of 4 (50%) | – |
| 4 | 1 of 4 (25%) | 1 of 2 (50%) |
| 5 | 1 of 14 (7%) | 0 of 1 (0%) |
| Any | 6 of 36 (17%) | 2 of 8 (25%) |
▚▚▙♥︎╱╯ and the opponents miss mates about as often as each other (within luck).
Which mates get missed
By how the mate starts, both sides together: one side alone has too few to tell. Checks are the moves everyone looks at first.
| Starts with | Times | Missed | Other mates | ▚▚▙♥︎╱╯ | Opponents |
|---|---|---|---|---|---|
| a check | 34 | 21% | 10% | 5 of 26 | 2 of 8 |
| a capture | 2 | 0% | 19% | 0 of 2 | 0 of 0 |
| a quiet move | 8 | 12% | 19% | 1 of 8 | 0 of 0 |
Bold rows are clear patterns, not noise.
Played instead of the mate: by ▚▚▙♥︎╱╯, a capture 2 times, a different check 1 time and something else 3 times; by the opponents, a capture 1 time and a different check 1 time.
▚▚▙♥︎╱╯'s missed mates
Of these 6, ▚▚▙♥︎╱╯ went on to lose 0 and still won 6. A game with two missed mates counts twice.
2 more
| Mar 17, 2026 vs ◤▘▀◤╱▒ Mate in 5: 19.Ne4 Rg8 20.Nxf6+ Kd8 21.Nxg8 Ke8 22.Rh8 f5 23.Nh6#; ▚▚▙♥︎╱╯ played 19.Rde1, and still won | |
| Mar 17, 2026 vs ◤▘▀◤╱▒ Mate in 3: 24.Rh8+ Ne8 25.Nd5 f5 26.Bc7#; ▚▚▙♥︎╱╯ played 24.Re1, and still won |
Mates opponents missed against ▚▚▙♥︎╱╯
Of these 2, they went on to lose 1 and still won 1. A game with two missed mates counts twice.
Sharp positions: where ▚▚▙♥︎╱╯'s mistakes are
A position is sharp when most moves in it lose: fewer than a quarter of the legal moves keep it (within 5 points of win chance of the best). Every position ▚▚▙♥︎╱╯ moved in after the opening, with the game still open, was searched move by move to tell. Mistakes aren't spread evenly: they come where there are few ways to get it right.
▚▚▙♥︎╱╯ took a median of 8 s on a sharp position, about the same as the 7 s on the others. That's where extra time would pay: nearly all the mistakes are there, and the rest of the moves could go quicker to make room for it.
▚▚▙♥︎╱╯ found one of the safe moves in 78% of the sharp positions. By Maia-2, a typical player at ▚▚▙♥︎╱╯'s level finds one about 56% of the time, and one about 300 points stronger 59%.
How to tell
The signs, visible at a glance, that most often went with a sharp position in ▚▚▙♥︎╱╯'s games:
| Sign | Positions | Sharp when it's there | Of all sharp positions |
|---|---|---|---|
| When the opponent's last move attacked one of ▚▚▙♥︎╱╯'s pieces | 278 | 79% | 22% |
| When lots of pieces are attacking each other | 650 | 75% | 47% |
Between them, these flag 55% of ▚▚▙♥︎╱╯'s sharp positions, and a position with one of them is sharp 75% of the time, against 58% overall. When one is there, check each candidate move against the opponent's best reply before playing it.
When to slow down
Each row compares ▚▚▙♥︎╱╯'s mistake rate in a situation with the rate everywhere else. Bold rows are clear patterns, not noise. There's no need to calculate every move, only these. A high rate alone mostly says the situation is hard, for anyone. The last column takes that out: ▚▚▙♥︎╱╯'s mistakes there against those a player of the same level would be expected to make in the very same positions. Above 1× is worse than the level, below is better, and it's in bold when it's clear.
| Situation | Moves | Mistake rate | Otherwise | Risk | Time spent | Against the level |
|---|---|---|---|---|---|---|
| When the opponent has a threat that isn't a direct attack on a piece | 159 | 19% | 7% | 2.7× | 6 s | 1.15× |
| When only one move holds the position | 253 | 13% | 7% | 1.8× | 4 s | 0.82× |
| When ▚▚▙♥︎╱╯'s king is still in the centre after move 12 | 103 | 13% | 8% | 1.7× | 10 s | 1.21× |
| When lots of pieces are attacking each other | 814 | 10% | 7% | 1.4× | 9 s | 0.92× |
| When clearly winning | 717 | 10% | 7% | 1.4× | 7 s | 0.91× |
| When the opponent's last move attacked one of ▚▚▙♥︎╱╯'s pieces | 391 | 10% | 7% | 1.3× | 7 s | 1.02× |
| When two or more of ▚▚▙♥︎╱╯'s pieces are undefended | 1537 | 7% | 8% | 0.9× | 4 s | 1.02× |
| Right after the opponent captures something | 674 | 7% | 8% | 0.9× | 4 s | 0.96× |
| With under a minute on the clock | 30 | 7% | 8% | 0.9× | 4 s | - |
| When three or more of the opponent's pieces aim at the area around ▚▚▙♥︎╱╯'s king | 16 | 6% | 8% | 0.8× | 8 s | - |
| When ahead in material | 340 | 6% | 8% | 0.8× | 8 s | 1.14× |
| In the endgame | 521 | 6% | 8% | 0.7× | 6 s | 0.86× |
| In the opening (moves 1-10) | 989 | 6% | 9% | 0.6× | 3 s | 1.12× |
| When behind in material | 452 | 3% | 9% | 0.4× | 5 s | 0.57× |
| When in check | 122 | 2% | 8% | 0.3× | 5 s | 0.51× |
In the riskiest situation (when the opponent has a threat that isn't a direct attack on a piece), ▚▚▙♥︎╱╯ spends a median of 6 s per move, about the same as the 5 s spent otherwise.
Against players of the same level
The same comparison by the kind of move ▚▚▙♥︎╱╯ made and how long it took: in 1,772 positions searched move by move, ▚▚▙♥︎╱╯'s mistakes against those a typical player of the same level (by Maia-2) would make in them. Bold rows are clearly different from the level.
| Moves | Mistake rate | Expected at this level | Against the level | ||
|---|---|---|---|---|---|
| Kind of move | |||||
| Moving a pawn | 400 | 10% | 10% | 1.06× | |
| Moving a knight | 283 | 11% | 11% | 1.02× | |
| Moving a bishop | 308 | 14% | 11% | 1.27× | |
| Moving a rook | 364 | 5% | 8% | 0.66× | |
| Moving the queen | 226 | 10% | 11% | 0.95× | |
| Moving the king | 191 | 7% | 7% | 0.97× | |
| Capturing | 373 | 8% | 10% | 0.85× | |
| Giving check | 181 | 6% | 10% | 0.59× | |
| Moving a piece back | 240 | 9% | 9% | 1.01× | |
| Time taken | |||||
| Under 3 s | 298 | 8% | 9% | 0.91× | |
| 3-10 s | 776 | 9% | 9% | 1.02× | |
| 10-30 s | 557 | 12% | 10% | 1.12× | |
| 30 s or more | 141 | 7% | 11% | 0.62× | |
However long ▚▚▙♥︎╱╯ took, the mistakes came about as often as the positions would lead one to expect: longer thinks went to harder positions, and did about as well there as a typical player would.
How ▚▚▙♥︎╱╯'s games are lost
For every loss: the last moment the position was fine, the move that did the most damage after it, and whether the game went in one move or slipped away.
| Kind of loss | Games | What it means |
|---|---|---|
| One-move oversight | 9 | a single move lost the game, and their very next move took the material or mated |
| One move turned it, and it wasn't obvious | 16 | a single move swung the game, but the punishment took a combination or a quiet move to find |
| Several mistakes added up | 4 | no single move decided it: the position got worse over several moves until it was lost |
| Flagged in a playable position | 1 | ran out of time with the game still going |
In the 20 losses that weren't simple oversights, the decisive position was an only-move situation 5 times, and the opponent already had a threat on the board 7 times. ▚▚▙♥︎╱╯ spent a median of 6 s on the move that decided it. These are the positions that felt normal but weren't.
Games where ▚▚▙♥︎╱╯ reached a winning position (80%+) and didn't win: 16.
Each loss
| Game | Win chance | Kind | Decisive move |
|---|---|---|---|
| vs ╳▜▟▄●▙ (1953) Nov 7, 2026 · Indian Game · was winning | One move turned it, and it wasn't obvious | 21...Bxc3 97%→31% · 13 s Better was Rd8. | |
▚▚▙♥︎╱╯ played Bxc3 (red); Rd8 (or Rb8) was better (green). Their best reply: 22.Qxc3 Qxc3+ (orange). They found it. | |||
| vs ╳╱▛▝▘▄ (1879) Oct 31, 2026 · Indian Game · was winning | One move turned it, and it wasn't obvious | 58...Bd7 49%→5% · 12 s 59.Re7+ forks the king and bishop. | |
▚▚▙♥︎╱╯ played Bd7 (red); Bxh3 (or Kf7) was better (green). Their best reply: 59.Re7+ Kg6 60.Rxd7 Rc2+ 61.Ke3 Rc3+ (orange). They found it. | |||
| vs ♠︎▄╱♣︎╲╰ (1890) Oct 16, 2026 · French Defense · was winning | One-move oversight | 19...Qc4 92%→4% · 8 s 20.Qxc4 wins the queen on c4: it moved to a square it couldn't hold. | |
▚▚▙♥︎╱╯ played Qc4 (red); Qc6 (or Qf5) was better (green). Their best reply: 20.Qxc4 g5 21.Bg3 Kg7 22.Be5 Bd6 (orange). They found it. | |||
| vs ◤◢▝○╲▘ (1907) Oct 15, 2026 · Closed Sicilian Defense · was winning | Flagged in a playable position | 35.Qb8+ 86%→84% · 4 s Better was Qd5. | |
▚▚▙♥︎╱╯ played Qb8+ (red); Qd5 (or Qb3) was better (green). Their best reply: 35...Kh7 36.Qd8 Re8 37.Qd7 Re7 38.Qd4 (orange). They found it. | |||
| vs ▟╭▜▙╲◣ (1926) Oct 14, 2026 · Sicilian Defense · was winning | One move turned it, and it wasn't obvious | 28.Rd6 97%→6% · 3 s 28...Qc5+ 29.Kf1 Qxd6 30.Qe4+ f5 31.Qe7+ wins material. | |
▚▚▙♥︎╱╯ played Rd6 (red); Rxd8 was better (green). Their best reply: 28...Qc5+ 29.Kf1 Qxd6 30.Qe4+ f5 31.Qe7+ (orange). They found it. | |||
| vs ♥︎▟◥▘▌▟ (1901) Oct 9, 2026 · Caro Kann Defense · was winning | One move turned it, and it wasn't obvious | 36.Qc6 56%→12% · 40 s Better was Qb7+. | |
▚▚▙♥︎╱╯ played Qc6 (red); Qb7+ (or Qc3+) was better (green). Their best reply: 36...Rh4+ 37.Kg1 (orange). They found it. | |||
| vs ▖╳╰╳▒▐ (1828) Oct 2, 2026 · Birds Opening · was winning | One move turned it, and it wasn't obvious | 26...dxe4 63%→11% · 0 s Better was Bxe4. | |
▚▚▙♥︎╱╯ played dxe4 (red); Bxe4 was better (green). Their best reply: 27.Qd1 Bf5 28.d5 Rb1 29.Qh5 Bg6 (orange). | |||
| vs ▌●▐╭▟♦︎ (1878) Sep 25, 2026 · French Defense | One move turned it, and it wasn't obvious | 13...Qa5 73%→16% · 4 s 14.Nb3 Bxf2+ 15.Rxf2 Qb4 16.Nc5 Bc8 wins material. | |
▚▚▙♥︎╱╯ played Qa5 (red); Qa3 was better (green). Their best reply: 14.Nb3 Bxf2+ 15.Rxf2 Qb4 16.Nc5 Bc8 (orange). They found it. | |||
| vs ▄▜╯▌◥╳ (1728) Sep 20, 2026 · Vienna Game · was winning | Several mistakes added up | 18...Bd8 87%→78% · 34 s Better was Be6. It went over 1 moves: 18...Bd8 (−8) | |
▚▚▙♥︎╱╯ played Bd8 (red); Be6 (or Bd7) was better (green). Their best reply: 19.Nxa7 Bd7 20.Nd6 b5 21.a4 bxa4 (orange). | |||
| vs ▄╭▌╲◢♣︎ (1742) Sep 3, 2026 · Sicilian Defense | One-move oversight | 11.Be5 62%→5% · 6 s 11...Bxe5 wins the bishop on e5: it moved to a square it couldn't hold. | |
▚▚▙♥︎╱╯ played Be5 (red); Bxg7 (or Bb3) was better (green). Their best reply: 11...Bxe5 12.h4 Rb8 13.Bb3 Rb4 14.h5 (orange). They found it. | |||
| vs ▟▄╮π▖╳ (1755) Sep 3, 2026 · Alapin Sicilian Defense | One-move oversight | 7...Ne4 49%→4% · 3 s The move opened the line for 8.Bxd8, which wins the queen on d8. | |
▚▚▙♥︎╱╯ played Ne4 (red); h6 (or Be7) was better (green). Their best reply: 8.Bxd8 Kxd8 9.Ne5 Nxe5 10.dxe5 Kc7 (orange). They found it. | |||
| vs ▒╰╭▖○╯ (1724) Aug 22, 2026 · Alapin Sicilian Defense | One-move oversight | 14...Nh5 54%→9% · 17 s The move opened the line for 15.Bxe7, which wins the bishop on e7. | |
▚▚▙♥︎╱╯ played Nh5 (red); Re8 (or a5) was better (green). Their best reply: 15.Bxe7 Re8 16.Qe2 Qb8 17.Qxh5 Qxb2 (orange). They found it. | |||
| vs ▄◤▀▒♣︎π (1760) Aug 8, 2026 · Sicilian Defense | Several mistakes added up | 3...e6 46%→31% · 2 s Better was cxd4. It went over 4 moves: 3...e6 (−15), 4...exd5 (−5), 10...Nh5 (−5), 11...Bd7 (−13) | |
▚▚▙♥︎╱╯ played e6 (red); cxd4 (or Nxd4) was better (green). Their best reply: 4.d5 Na5 5.d6 a6 6.e5 b6 (orange). They found it. | |||
| vs ○▛▜▖╯╰ (1769) Aug 1, 2026 · Three Knights Opening · was winning | Several mistakes added up | 20.Rxa7 90%→89% · 2 s Better was Qe4. It went over 0 moves: | |
▚▚▙♥︎╱╯ played Rxa7 (red); Qe4 was better (green). Their best reply: 20...Rxf5 21.Ra8 Rxa8 22.Qxa8+ Rf8 23.Qe4 (orange). They found it. | |||
| vs ╯▟╭▙▘▘ (1723) Jul 21, 2026 · French Defense | One move turned it, and it wasn't obvious | 6...Qxd5 41%→22% · 3 s Better was Qe7+. | |
▚▚▙♥︎╱╯ played Qxd5 (red); Qe7+ (or Nd7) was better (green). Their best reply: 7.Nxd5 Bd6 8.Ne2 Nc6 9.Bf4 Be6 (orange). They found it. | |||
| vs ♦︎▛▖π♠︎▚ (1708) Jul 18, 2026 · Caro Kann Defense · was winning | One move turned it, and it wasn't obvious | 55.Kg4 50%→19% · 7 s Better was Rd5+. | |
▚▚▙♥︎╱╯ played Kg4 (red); Rd5+ was better (green). Their best reply: 55...Rc3 56.Rf5 Rxc2 57.Rxf6+ Ke5 58.Rxf4 (orange). They found it. | |||
| vs ○▞▜♥︎▌▐ (1709) Jul 10, 2026 · Caro Kann Defense | One-move oversight | 15.Bg5 38%→7% · 7 s The bishop on e3 was guarding the knight on c5; once it moved, 15...Bxc5 took it. | |
▚▚▙♥︎╱╯ played Bg5 (red); a4 (or Be2) was better (green). Their best reply: 15...Bxc5 16.Rfe1 Rd8 17.Rad1 Ne7 18.Bb1 (orange). They found it. | |||
| vs ▟●♥︎♣︎◣▀ (1722) Apr 5, 2026 · Pirc Defense | One move turned it, and it wasn't obvious | 19...Bc6 47%→5% · 12 s 20.e4 traps the queen on d5. | |
▚▚▙♥︎╱╯ played Bc6 (red); f6 was better (green). Their best reply: 20.e4 Qxd6 21.exd6 h6 22.Nf3 g5 (orange). They found it. | |||
| vs ╲▗◣●▜▘ (1693) Mar 23, 2026 · Sicilian Defense | One move turned it, and it wasn't obvious | 18.g4 55%→24% · 8 s Better was Re1. | |
▚▚▙♥︎╱╯ played g4 (red); Re1 (or b3) was better (green). Their best reply: 18...Nxf3 19.Rg2 hxg4 20.Rxg4 Rh8 21.Rf4 (orange). They found it. | |||
| vs ♣︎▖◢▜♠︎╯ (1692) Mar 22, 2026 · Caro Kann Defense | One move turned it, and it wasn't obvious | 15.Kh3 52%→4% · 2 s 15...Nxf2+ forks the king and queen. | |
▚▚▙♥︎╱╯ played Kh3 (red); Kg3 was better (green). Their best reply: 15...Nxf2+ 16.Kg3 Nxd3 17.cxd3 Kg8 18.Rf1 (orange). They found it. | |||
| vs ▟◥●▀▝╰ (1691) Mar 20, 2026 · Philidor Defense · was winning | One-move oversight | 44.Be4 86%→14% · 6 s The bishop on d6 was already under attack; 44...Rxd6 takes it. | |
▚▚▙♥︎╱╯ played Be4 (red); Bd3+ was better (green). Their best reply: 44...Rxd6 45.Bh7 Kb5 46.Kg3 Kxb4 47.h4 (orange). They found it. | |||
| vs ◣♠︎▘▛◤╭ (1686) Mar 19, 2026 · Italian Game · was winning | One move turned it, and it wasn't obvious | 18...Qf6 95%→38% · 24 s Better was Nxg2. | |
▚▚▙♥︎╱╯ played Qf6 (red); Nxg2 (or Nd3) was better (green). Their best reply: 19.d4 Bd6 (orange). They found it. | |||
| vs ♥︎▒╱▒╳▐ (1672) Mar 18, 2026 · Van t Kruijs Opening | One-move oversight | 4...Nf6 56%→0% · 2 s 5.Qxf7# is checkmate. | |
▚▚▙♥︎╱╯ played Nf6 (red); g6 (or e6) was better (green). Their best reply: 5.Qxf7# (orange). They found it. | |||
| vs ▐▛╯▜▄◥ (1719) Mar 18, 2026 · Giuoco Piano Game | One move turned it, and it wasn't obvious | 11.Bxf7+ 35%→8% · 14 s 11...Rxf7 recaptures, and the exchange loses material: the bishop on f7 wasn't safe there. | |
▚▚▙♥︎╱╯ played Bxf7+ (red); Bd5 was better (green). Their best reply: 11...Rxf7 12.Bxa7 Nxa7 13.Nc3 Bb7 14.a4 (orange). They found it. | |||
| vs π●▚╳▘◢ (1666) Mar 17, 2026 · Reti Opening | Several mistakes added up | 12.Ne2 30%→4% · 3 s The knight on g5 was already under attack; 12...Qxg5 takes it. It went over 4 moves: 4.O-O (−9), 6.b3 (−8), 10.Ng5 (−8), 12.Ne2 (−26) | |
▚▚▙♥︎╱╯ played Ne2 (red); Qg4 (or Nf3) was better (green). Their best reply: 12...Qxg5 13.Bc1 Qd8 14.c3 e5 15.Ng3 (orange). They found it. | |||
| vs ▝◤▞▛╯π (1683) Mar 17, 2026 · Kings Gambit | One move turned it, and it wasn't obvious | 14...Bb6 39%→16% · 8 s Better was Rad8. | |
▚▚▙♥︎╱╯ played Bb6 (red); Rad8 was better (green). Their best reply: 15.Nxb6 Rad8 16.Nd4 Qc5 17.Rxf4 Qxb6 (orange). They found it. | |||
| vs ◣▐▄♦︎◤▚ (1660) Mar 17, 2026 · Pirc Defense | One-move oversight | 9...Nxe4 65%→6% · 14 s The move opened the line for 10.Bxd8, which wins the queen on d8. | |
▚▚▙♥︎╱╯ played Nxe4 (red); exd4 (or Qa5) was better (green). Their best reply: 10.Bxd8 Rxd8 11.Bb3 Nb6 12.Nd2 a5 (orange). They found it. | |||
| vs ▘♠︎◥▞▙◢ (1655) Mar 16, 2026 · Vienna Game | One move turned it, and it wasn't obvious | 37.Ke3 50%→7% · 0 s Better was Kf3. | |
▚▚▙♥︎╱╯ played Ke3 (red); Kf3 (or Ke2) was better (green). Their best reply: 37...Ke5 38.Kf2 Kd4 39.Ke2 Kc3 40.Ke3 (orange). They found it. | |||
| vs ▟▖♠︎╳♣︎╲ (1661) Mar 16, 2026 · Vienna Game · was winning | One-move oversight | 12.Kg1 77%→23% · 1 s 12...gxf3 wins the knight on f3, which the move left without enough protection. | |
▚▚▙♥︎╱╯ played Kg1 (red); Nh2 was better (green). Their best reply: 12...gxf3 13.Bb3 Ne5 14.Na4 Ba7 15.Nc5 (orange). They found it. | |||
| vs ▜◣▌╰╰○ (1705) Mar 16, 2026 · French Defense · was winning | One move turned it, and it wasn't obvious | 14.Rfe1 59%→11% · 6 s 14...Bxf2+ forks the king and rook. | |
▚▚▙♥︎╱╯ played Rfe1 (red); Rae1 (or Nfd4) was better (green). Their best reply: 14...Bxf2+ 15.Kf1 Bxe1 16.Rxe1 Nac5 17.Nxa7 (orange). They found it. | |||
Did ▚▚▙♥︎╱╯'s opponents use an engine?
Two kinds of evidence. First, whether chess.com has closed the opponent's account for fair-play violations. Second, how often they played Stockfish's first choice in positions that were still undecided. High accuracy on its own means little: once one side has blundered, almost any move wins for the other, so those moves aren't counted.
Games to skip
| Game | Evidence | Top move | Their mistakes |
|---|---|---|---|
| vs ╲▖▝▟▀▒ (1721) Mar 16, 2026 · ▚▚▙♥︎╱╯ won | Account closed for fair-play violations. In this game they played the top move 53% of the time (34% typical) | 53% of 36 | 3 |
▚▚▙♥︎╱╯'s most engine-like opponents
For comparison: the unflagged games where the opponent played the top move most often, against what a typical opponent of ▚▚▙♥︎╱╯'s would have managed in the same positions.
| Game | Typical | Top move | Their mistakes |
|---|---|---|---|
| vs ♦︎▛▖π♠︎▚ (1708) Jul 18, 2026 · ▚▚▙♥︎╱╯ lost | 34% | 50% of 48 | 2 |
| vs ▟╭▜▙╲◣ (1926) Oct 14, 2026 · ▚▚▙♥︎╱╯ lost | 37% | 59% of 22 | 2 |
| vs ○▐♥︎▜╱╮ (1681) Jul 21, 2026 · ▚▚▙♥︎╱╯ won | 34% | 52% of 23 | 1 |
| vs ▟╭π▗╭▌ (1667) Mar 18, 2026 · ▚▚▙♥︎╱╯ won | 35% | 50% of 20 | 2 |
| vs ▟◥●▀▝╰ (1691) Mar 20, 2026 · ▚▚▙♥︎╱╯ lost | 35% | 45% of 31 | 6 |
▚▚▙♥︎╱╯'s most engine-like games
The same test on ▚▚▙♥︎╱╯'s own moves: the games where ▚▚▙♥︎╱╯ played the top move most often, against the same typical opponent in the same positions. How far these games stray from typical shows how far an opponent's can without an engine.
| Game | Typical | Top move | Mistakes |
|---|---|---|---|
| vs ▙◥╱♣︎╯▗ (1722) would be flagged Sep 22, 2026 · ▚▚▙♥︎╱╯ won | 37% | 71% of 51 | 4 |
| vs ▟╱▒◣╮▟ (1821) would be flagged Oct 4, 2026 · ▚▚▙♥︎╱╯ won | 42% | 94% of 16 | 0 |
| vs ▛◣▌◤▖▙ (1695) would be flagged Jul 23, 2026 · ▚▚▙♥︎╱╯ won | 38% | 72% of 29 | 1 |
| vs ▝▀▘▞○╲ (1701) would be flagged Jul 24, 2026 · ▚▚▙♥︎╱╯ won | 36% | 80% of 15 | 0 |
| vs ╳╱▛▝▘▄ (1879) Oct 31, 2026 · ▚▚▙♥︎╱╯ lost | 34% | 60% of 42 | 3 |
By the same test, 5 of ▚▚▙♥︎╱╯'s 55 games long enough to judge would be flagged too. ▚▚▙♥︎╱╯'s most engine-like opponent who wasn't flagged, ♦︎▛▖π♠︎▚ (50% of 48), would rank 17th among them.
Players at ▚▚▙♥︎╱╯'s level pick the engine's first choice about 36% of the time in undecided positions. ▚▚▙♥︎╱╯ plays it 52% of the time. Someone copying an engine agrees with it 55–77% of the time, depending on how strong an engine they use. A game is flagged when the opponent reaches 60% over 15+ moves and that is far beyond chance. Someone who only uses an engine for a few key moves won't show up in a single game. If you're suspicious, report them on chess.com: its fair-play team looks at all of a player's games. Accounts are often closed weeks later, so each run checks them again.
▚▚▙♥︎╱╯'s clock
Playing fast is fine. The problem is spending the same few seconds on a sharp position as on a quiet one.
| Time on the move | Share of moves | Mistake rate | In a sharp position |
|---|---|---|---|
| under 3 s | 23% | 7% | 30% |
| 3-10 s | 43% | 9% | 22% |
| 10-30 s | 27% | 11% | 26% |
| 30 s or more | 7% | 7% | 27% |
Sharp: they had a threat on the board, or only one move held. ▚▚▙♥︎╱╯'s long moves are in sharp positions no more often than the quick ones (26% of the moves over 10 s, against 24% of quicker ones), so the extra time isn't following the position. Long moves like that are often time away from the board rather than thinking, so more mistakes on slow moves doesn't mean thinking hurts. The risky move is the one made on coming back, before the opponent's last move has sunk in: look at it before playing.
How hard were the tactics that hurt ▚▚▙♥︎╱╯?
For each of ▚▚▙♥︎╱╯'s 65 tactical mistakes: how many moves the opponent needed before the material was gone or the mate landed.
| Their very next move Material was simply taken, or it was mate in one. A quick safety glance before moving catches these. | 36 | 55% |
| Two moves deep A threat, then the capture: a fork, a pin, a trapped piece, mate in two. Pattern practice makes these jump out. | 20 | 31% |
| Three or more moves A real combination. Even strong players miss these, so they're the least worrying kind. | 9 | 14% |
Which threats ▚▚▙♥︎╱╯ misses
Every time the opponent already had a real threat on the board (to win material or mate), did ▚▚▙♥︎╱╯ deal with it? Here that's split by the shape of the threat. If one shape gets missed far more often than the rest, that's a blind spot in how ▚▚▙♥︎╱╯ scans the board.
| Kind of threat | Times | Missed | Other threats | Ratio |
|---|---|---|---|---|
| Long-diagonal threats (bishop or queen from 3+ squares away) | 30 | 17% | 3% | 5.8× |
| Quiet threats (the threatened move isn't a capture or check) | 30 | 3% | 4% | 0.9× |
| Knight threats | 85 | 2% | 4% | 0.6× |
| Threats from a backward move | 43 | 0% | 4% | 0.0× |
Bold rows are clear patterns, not noise: bigger than game-to-game luck would explain, allowing for how many shapes are compared.
Knights or bishops
Trading a knight for a bishop, or a bishop for a knight, comes up in most games. Which of ▚▚▙♥︎╱╯'s does more, and which of the opponents' does more damage?
- No clear difference yet between knights and bishops, ▚▚▙♥︎╱╯'s or the opponents', so neither trade stands out as costly. Decide each one on the position: a bishop is usually the better piece in an open position with pawns on both sides of the board, a knight in a closed one, or on a square no pawn can chase it from.
| Knights | Bishops | |
|---|---|---|
| ▚▚▙♥︎╱╯'s knights and bishops | ||
| Mistakes per 100 moves with itEvery real decision made moving one. | 8.5566 moves | 10.7457 moves |
| Winning shots it had, and ▚▚▙♥︎╱╯ tookAfter the opponent's mistake, the best move won material or mated, with this piece doing it. | 9 of 1182% | 10 of 10100% |
| The opponents' | ||
| Times theirs won material from ▚▚▙♥︎╱╯Per 100 moves ▚▚▙♥︎╱╯ made facing one. | 0.511 times | 0.716 times |
| Their threats with it that ▚▚▙♥︎╱╯ missedThreats to win material or mate. | 2 of 852% | 5 of 658% |
| Trades | ||
| Trades ▚▚▙♥︎╱╯ started giving it up, that were mistakesA knight for their bishop, or a bishop for their knight. ▚▚▙♥︎╱╯'s other moves: 8%. | 0 of 160% | 1 of 254% |
| With it against their other one, ▚▚▙♥︎╱╯'s scoreGames where that lasted 6 or more of ▚▚▙♥︎╱╯'s moves, against the win chance when it began. | 92% from 70%6 games | 48% from 55%26 games |
Bold rows are clear differences between knights and bishops, not noise, and a bold trade goes wrong clearly more often than ▚▚▙♥︎╱╯'s other moves. A shot, or material lost, is put down to the piece that makes the fork, pin, skewer or discovered attack, or else takes the material.
Openings
▚▚▙♥︎╱╯'s score in each opening, counting a win as 1 and a draw as ½, and the average win chance after move 10. Below about 40% after move 10 means leaving the opening worse off than at the start.
By name
The name the site gave each game. It follows the line the game turned into, so games that began the same way can end up under different names.
| Opening | Colour | Games | Score | Win chance after move 10 |
|---|---|---|---|---|
| French Defense | Black | 13 | 73% | 59% |
| Sicilian Defense | White | 13 | 73% | 64% |
| Sicilian Defense | Black | 7 | 86% | 47% |
| Caro Kann Defense | White | 6 | 33% | 46% |
| Indian Game | Black | 6 | 58% | 52% |
| Vienna Game | White | 6 | 67% | 62% |
| French Defense | White | 4 | 75% | 58% |
| Pirc Defense | White | 4 | 88% | 38% |
| English Opening | Black | 3 | 83% | 40% |
| Kings Gambit | Black | 3 | 67% | 49% |
| Kings Pawn Opening | White | 3 | 100% | 61% |
| Alapin Sicilian Defense | Black | 2 | 0% | 17% |
| Four Knights Game | White | 2 | 100% | 57% |
| Giuoco Piano Game | White | 2 | 50% | 49% |
| Kings Indian Defense | Black | 2 | 100% | 56% |
By first moves
The same games by the moves they began with, each side's first three, however they were named later. Each row is the games that went that way, under the row they branched from; moves all of them went on to play are folded in. Lines with fewer than 3 games are left out.
| As white | Games | Score | Win chance after move 10 |
|---|---|---|---|
| 53 | 70% | 56% | |
| 19 | 71% | 57% | |
| 12 | 67% | 59% | |
| 6 | 67% | 56% | |
| 3 | 67% | 71% | |
| 5 | 70% | 63% | |
| 3 | 50% | 58% | |
| 15 | 70% | 62% | |
| 13 | 73% | 63% | |
| 6 | 58% | 58% | |
| 3 | 50% | 61% | |
| 5 | 80% | 60% | |
| 4 | 75% | 62% | |
| 6 | 33% | 46% | |
| 5 | 40% | 47% | |
| 4 | 75% | 58% | |
| 4 | 100% | 51% | |
| 3 | 100% | 57% |
| As black | Games | Score | Win chance after move 10 |
|---|---|---|---|
| 28 | 61% | 50% | |
| 13 | 73% | 59% | |
| 5 | 60% | 51% | |
| 3 | 67% | 29% | |
| 4 | 75% | 64% | |
| 3 | 67% | 55% | |
| 10 | 55% | 41% | |
| 5 | 40% | 28% | |
| 3 | 33% | 29% | |
| 5 | 40% | 45% | |
| 3 | 67% | 49% | |
| 10 | 65% | 48% | |
| 3 | 100% | 53% |
Habits
| Castled by move 10 Most coaches aim for nearly always. | 69% |
| Never castled (games of 15+ moves) A king left in the middle is a target once lines open. | 8% |
| Queen moved in the first 6 moves Early queen trips usually cost time as she gets chased. | 4% |
| Two or more minor pieces undeveloped at move 12 Get the knights and bishops out before going on adventures. | 9% |
| Phase | Mistakes per 100 moves |
|---|---|
| Opening (989 moves) | 5.6 |
| Middlegame (1267 moves) | 10.3 |
| Endgame (521 moves) | 6.0 |
Every game
Newest first. Open a game for its serious mistakes (10%+ win chance) on boards or step through the whole game move by move, with the engine's view of every position.
Download all 100 games as PGN, or , to open in another chess program, with the engine's evaluation and the clock after every move.
Nov 7, 2026 · vs ▚○◣▗╯▐ (1919) · won by resignation · 76% accuracy (vs 70%) · 4 mistakes
Nov 7, 2026 · vs ●π▟╭◥▗ (1912) · won by resignation · 99% accuracy (vs 82%) · 0 mistakes
Nov 7, 2026 · vs ╳▜▟▄●▙ (1953) · lost by resignation · 63% accuracy (vs 87%) · 2 mistakes
Nov 7, 2026 · vs ▞♣︎▒╯╳π (1942) · won by resignation · 70% accuracy (vs 72%) · 4 mistakes
Nov 5, 2026 · vs ♥︎π▜◥▜▝ (1899) · won by resignation · 97% accuracy (vs 87%) · 0 mistakes
Nov 4, 2026 · vs ▄▞○▀╭π (1865) · won by resignation · 86% accuracy (vs 71%) · 2 mistakes
Nov 3, 2026 · vs ╲╰▖╱▚▙ (1880) · won by resignation · 98% accuracy (vs 84%) · 0 mistakes
Oct 31, 2026 · vs ╳♦︎╱▙▘▀ (1877) · won by resignation · 62% accuracy (vs 55%) · 6 mistakes
Oct 31, 2026 · vs ╳╱▛▝▘▄ (1879) · lost on time · 80% accuracy (vs 88%) · 3 mistakes
Oct 31, 2026 · vs ╯π╮╯╳╲ (1963) · won by resignation · 99% accuracy (vs 83%) · 0 mistakes
Oct 23, 2026 · vs ▒◢◢▝╳◤ (1854) · won by resignation · 97% accuracy (vs 75%) · 0 mistakes
Oct 16, 2026 · vs ♠︎▄╱♣︎╲╰ (1890) · lost by resignation · 46% accuracy (vs 83%) · 1 mistake
Oct 15, 2026 · vs ◤◢▝○╲▘ (1907) · lost on time · 95% accuracy (vs 89%) · 1 mistake
Oct 14, 2026 · vs ▟╭▜▙╲◣ (1926) · lost by resignation · 53% accuracy (vs 83%) · 3 mistakes
Oct 14, 2026 · vs ▛╳▒▞♣︎▟ (1889) · won by checkmate · 52% accuracy (vs 67%) · 3 mistakes
Oct 13, 2026 · vs ▙▟▐▙╱▗ (1879) · won by checkmate · 70% accuracy (vs 62%) · 1 mistake
Oct 12, 2026 · vs ♠︎╯π▗▜╰ (1847) · won by resignation · 92% accuracy (vs 80%) · 2 mistakes
Oct 9, 2026 · vs ♥︎▟◥▘▌▟ (1901) · lost by resignation · 86% accuracy (vs 94%) · 2 mistakes
Oct 9, 2026 · vs ▄π▞▗◣▀ (1852) · won by checkmate · 97% accuracy (vs 85%) · 0 mistakes
Oct 9, 2026 · vs ◥╮●◥▐╮ (1866) · drew by agreement · 75% accuracy (vs 77%) · 4 mistakes
Oct 9, 2026 · vs ╱○╲♣︎╭╮ (1866) · won by checkmate · 79% accuracy (vs 71%) · 4 mistakes
Oct 9, 2026 · vs ▌▙╱▘♦︎▟ (1869) · won by resignation · 96% accuracy (vs 90%) · 0 mistakes
Oct 8, 2026 · vs π╯▄◢╲▛ (1807) · won by resignation · 74% accuracy (vs 53%) · 6 mistakes
Oct 8, 2026 · vs ▄╭▙♦︎╳◣ (1850) · won by resignation · 98% accuracy (vs 88%) · 0 mistakes
Oct 4, 2026 · vs ▟╱▒◣╮▟ (1821) · won by resignation · 98% accuracy (vs 86%) · 0 mistakes
Oct 4, 2026 · vs ╳▀▙◢▗π (1805) · won by resignation · 93% accuracy (vs 67%) · 0 mistakes
Oct 4, 2026 · vs ╮▖▜◢╳╮ (1819) · won by resignation · 72% accuracy (vs 66%) · 3 mistakes
Oct 2, 2026 · vs ▖╳╰╳▒▐ (1828) · lost by resignation · 70% accuracy (vs 84%) · 4 mistakes
Oct 1, 2026 · vs ▀▒▌◢♠︎◤ (1831) · won by resignation · 96% accuracy (vs 88%) · 1 mistake
Sep 28, 2026 · vs ▚▐▝▒○╮ (1788) · won by checkmate · 83% accuracy (vs 78%) · 3 mistakes
Sep 25, 2026 · vs ▌●▐╭▟♦︎ (1878) · lost by resignation · 58% accuracy (vs 91%) · 1 mistake
Sep 25, 2026 · vs ▌▛╳●◣◥ (1774) · won by checkmate · 83% accuracy (vs 79%) · 7 mistakes
Sep 25, 2026 · vs ╰♥︎◣▚▖▙ (1753) · won by checkmate · 100% accuracy (vs 54%) · 0 mistakes
Sep 25, 2026 · vs ◥♥︎●▛◢▐ (1751) · won by resignation · 53% accuracy (vs 69%) · 3 mistakes
Sep 25, 2026 · vs ○●♦︎♦︎▒▐ (1746) · won by resignation · 97% accuracy (vs 71%) · 0 mistakes
Sep 25, 2026 · vs ▛╲╭◢♥︎○ (1765) · won by resignation · 99% accuracy (vs 78%) · 0 mistakes
Sep 23, 2026 · vs ▌▌╲▟♠︎▟ (1738) · won by resignation · 54% accuracy (vs 32%) · 4 mistakes
Sep 23, 2026 · vs ◢◢╱▝▌▟ (1709) · won by resignation · 83% accuracy (vs 79%) · 2 mistakes
Sep 22, 2026 · vs ▙◥╱♣︎╯▗ (1722) · won by resignation · 85% accuracy (vs 84%) · 4 mistakes
Sep 20, 2026 · vs ▄▜╯▌◥╳ (1728) · lost by resignation · 84% accuracy (vs 75%) · 2 mistakes
Sep 20, 2026 · vs ╭▝▗▞╲▖ (1683) · won by resignation · 90% accuracy (vs 81%) · 1 mistake
Sep 5, 2026 · vs ▚▛╯▟◤▞ (475) · won by checkmate · 98% accuracy (vs 78%) · 0 mistakes
Sep 3, 2026 · vs ▄╭▌╲◢♣︎ (1742) · lost on time · 69% accuracy (vs 83%) · 2 mistakes
Sep 3, 2026 · vs ▘▛▟▗▌▖ (1724) · won by resignation · 83% accuracy (vs 71%) · 2 mistakes
Sep 3, 2026 · vs ▟▄╮π▖╳ (1755) · lost by resignation · 55% accuracy (vs 99%) · 1 mistake
Aug 25, 2026 · vs ▒▚▚▄▗▜ (1716) · won by resignation · 94% accuracy (vs 79%) · 0 mistakes
Aug 22, 2026 · vs ▒╰╭▖○╯ (1724) · lost by resignation · 71% accuracy (vs 89%) · 2 mistakes
Aug 8, 2026 · vs ▄◤▀▒♣︎π (1760) · lost by resignation · 76% accuracy (vs 87%) · 2 mistakes
Aug 1, 2026 · vs ♦︎▌▐╳╲▚ (1708) · won by resignation · 87% accuracy (vs 77%) · 2 mistakes
Aug 1, 2026 · vs ○▛▜▖╯╰ (1769) · lost by abandonment · 96% accuracy (vs 83%) · 0 mistakes
Jul 24, 2026 · vs ▝▀▘▞○╲ (1701) · won by checkmate · 96% accuracy (vs 89%) · 0 mistakes
Jul 24, 2026 · vs ♦︎▖π◥●▒ (1689) · won by resignation · 86% accuracy (vs 80%) · 3 mistakes
Jul 23, 2026 · vs ▛◣▌◤▖▙ (1695) · won by checkmate · 85% accuracy (vs 83%) · 1 mistake
Jul 22, 2026 · vs ╱╯▒▘▛▞ (1705) · won by checkmate · 98% accuracy (vs 92%) · 0 mistakes
Jul 21, 2026 · vs ○▐♥︎▜╱╮ (1681) · won on time · 91% accuracy (vs 85%) · 1 mistake
Jul 21, 2026 · vs ╯▟╭▙▘▘ (1723) · lost by resignation · 84% accuracy (vs 99%) · 3 mistakes
Jul 18, 2026 · vs ♦︎▛▖π♠︎▚ (1708) · lost by resignation · 84% accuracy (vs 91%) · 3 mistakes
Jul 10, 2026 · vs ○▞▜♥︎▌▐ (1709) · lost by resignation · 79% accuracy (vs 96%) · 2 mistakes
Jul 9, 2026 · vs ╯▟▘╯▛▛ (1696) · won by resignation · 95% accuracy (vs 53%) · 0 mistakes
Apr 5, 2026 · vs ▟●♥︎♣︎◣▀ (1722) · lost by checkmate · 83% accuracy (vs 91%) · 4 mistakes
Apr 4, 2026 · vs ▟▞♠︎▘◣♥︎ (1702) · drew by repetition · 81% accuracy (vs 82%) · 6 mistakes
Apr 4, 2026 · vs ╳▝╭▗π▟ (1680) · won by checkmate · 93% accuracy (vs 83%) · 1 mistake
Apr 3, 2026 · vs ╱▛▗▜▌▙ (1688) · won by resignation · 84% accuracy (vs 76%) · 4 mistakes
Apr 3, 2026 · vs ○◥╭▖○╲ (1692) · won by checkmate · 68% accuracy (vs 57%) · 2 mistakes
Mar 23, 2026 · vs ╲▗◣●▜▘ (1693) · lost by resignation · 83% accuracy (vs 92%) · 2 mistakes
Mar 22, 2026 · vs ▀♦︎▄▗╰▒ (1676) · won by checkmate · 90% accuracy (vs 77%) · 3 mistakes
Mar 22, 2026 · vs ♣︎▖◢▜♠︎╯ (1692) · lost by resignation · 58% accuracy (vs 74%) · 3 mistakes
Mar 22, 2026 · vs ▐♦︎╮▐♣︎▀ (1679) · drew by repetition · 80% accuracy (vs 84%) · 2 mistakes
Mar 20, 2026 · vs ♣︎▝╲▞◥▝ (1630) · won by resignation · 77% accuracy (vs 76%) · 2 mistakes
Mar 20, 2026 · vs ▘▐▙♥︎▀▌ (1685) · drew by agreement · 95% accuracy (vs 95%) · 0 mistakes
Mar 20, 2026 · vs ╰○♦︎╰●○ (1683) · won by abandonment · 94% accuracy (vs 69%) · 0 mistakes
Mar 20, 2026 · vs ▚▐╲●╲▌ (1668) · won by resignation · 90% accuracy (vs 75%) · 1 mistake
Mar 20, 2026 · vs ▟◥●▀▝╰ (1691) · lost by resignation · 62% accuracy (vs 77%) · 7 mistakes
Mar 19, 2026 · vs ╲◥♠︎▄○▄ (1710) · drew by repetition · 92% accuracy (vs 91%) · 3 mistakes
Mar 19, 2026 · vs ╰╳◤▟♣︎▛ (1675) · won by resignation · 96% accuracy (vs 84%) · 0 mistakes
Mar 19, 2026 · vs ♥︎▟╲♦︎◣▄ (1675) · won by resignation · 82% accuracy (vs 77%) · 6 mistakes
Mar 19, 2026 · vs ◣♠︎▘▛◤╭ (1686) · lost by resignation · 70% accuracy (vs 87%) · 3 mistakes
Mar 19, 2026 · vs ╲♣︎▚▐♦︎◢ (1672) · won by checkmate · 86% accuracy (vs 80%) · 4 mistakes
Mar 18, 2026 · vs ●▖╭▚♦︎▖ (1666) · won by checkmate · 85% accuracy (vs 73%) · 2 mistakes
Mar 18, 2026 · vs ♥︎▒╱▒╳▐ (1672) · lost by checkmate · 24% accuracy (vs 85%) · 1 mistake
Mar 18, 2026 · vs ▟╭π▗╭▌ (1667) · won on time · 81% accuracy (vs 91%) · 3 mistakes
Mar 18, 2026 · vs ▐▛╯▜▄◥ (1719) · lost by resignation · 69% accuracy (vs 96%) · 2 mistakes
Mar 18, 2026 · vs ╭▟◢╯╱♠︎ (1633) · won by checkmate · 77% accuracy (vs 68%) · 5 mistakes
Mar 18, 2026 · vs ╲▒▗○▐╱ (1630) · won by checkmate · 89% accuracy (vs 73%) · 2 mistakes
Mar 18, 2026 · vs π▚▖●▚╱ (1604) · won by resignation · 92% accuracy (vs 76%) · 2 mistakes
Mar 17, 2026 · vs ♠︎╲◥π╮π (1622) · won by checkmate · 46% accuracy (vs 40%) · 3 mistakes
Mar 17, 2026 · vs ▀╯◢▟▞♥︎ (1632) · drew by repetition · 84% accuracy (vs 88%) · 4 mistakes
Mar 17, 2026 · vs π●▚╳▘◢ (1666) · lost by resignation · 67% accuracy (vs 91%) · 1 mistake
Mar 17, 2026 · vs ▝◤▞▛╯π (1683) · lost by resignation · 76% accuracy (vs 86%) · 3 mistakes
Mar 17, 2026 · vs ◣▐▄♦︎◤▚ (1660) · lost by checkmate · 70% accuracy (vs 91%) · 1 mistake
Mar 17, 2026 · vs ◤▘▀◤╱▒ (1665) · won by resignation · 92% accuracy (vs 90%) · 1 mistake
Mar 17, 2026 · vs ▌●▝▘○▜ (1652) · won by checkmate · 85% accuracy (vs 66%) · 1 mistake
Mar 16, 2026 · vs ▘♠︎◥▞▙◢ (1655) · lost by resignation · 59% accuracy (vs 68%) · 6 mistakes
Mar 16, 2026 · vs ▝◥▒◢◣╰ (1650) · won by checkmate · 76% accuracy (vs 67%) · 3 mistakes
Mar 16, 2026 · vs ▟▖♠︎╳♣︎╲ (1661) · lost by checkmate · 60% accuracy (vs 75%) · 5 mistakes
Mar 16, 2026 · vs ╲▖▝▟▀▒ (1721) · won on time · 91% accuracy (vs 85%) · 3 mistakesaccount closed for fair play
Mar 16, 2026 · vs ♦︎▌╲╭▜♣︎ (1652) · won by resignation · 79% accuracy (vs 70%) · 2 mistakes
Mar 16, 2026 · vs ▜◣▌╰╰○ (1705) · lost by resignation · 57% accuracy (vs 76%) · 4 mistakes
Mar 16, 2026 · vs ╰▛◥●▌╳ (1620) · won by resignation · 88% accuracy (vs 76%) · 2 mistakes
Mar 16, 2026 · vs ╮╭▘╰▘◣ (1637) · won by abandonment · 78% accuracy (vs 71%) · 4 mistakes
How this was worked out
Every position was searched by Stockfish (150,000 nodes per position, about depth 15-17). Evaluations are converted to win chance on a curve fitted to how often players win from each one. A move that drops the mover's win chance by 10 points or more counts as a mistake, 20 or more as a blunder. Threats are found by letting the opponent move twice (a null move). If that wins material or mates, there was a threat on the board. How hard a tactic was is the number of opponent moves in the engine's line before the side that erred is two pawns or more down for good, or the length of the mate.
Engines are much better at telling that a move was bad than why. The pattern labels (fork, pin, loose piece...) come from rules applied to the engine's line and are usually, but not always, the whole story.
Sharp positions: before each of ▚▚▙♥︎╱╯'s moves after the opening with the game still open, Stockfish searched every legal move at once (200,000 nodes between them). A move within 5 points of win chance of the best was safe; a position with fewer than a quarter of its moves safe was sharp.
The human side comes from Maia-2 (Tang, Jiao, McIlroy-Young, Kleinberg, Sen and Anderson, NeurIPS 2024; CSSLab, University of Toronto; MIT licence), a model trained on lichess games that gives, for players of a rating, the chance of each move and the expected score. Its ratings are lichess's, and are put on the site's scale by fitting the opponents' moves, whose ratings are known. A mistake's cost in practice is the player's expected score after the best move less that after the move played, each averaged over the replies an opponent of their level would likely make. The level a side's moves are most like is the one whose players would most likely have made them.
Against players of the same level: in each position searched move by move, the chance that a player of ▚▚▙♥︎╱╯'s level plays one of the moves that were mistakes there (by Maia-2) is the mistakes expected of the level. They're scaled to match the count of mistakes over all those positions together, so a ratio says where the player differs from one kind of position to another, not overall.
A pattern is called clear when it's bigger than game-to-game luck would explain: a game's moves go well or badly together, so games are counted as the samples, not moves. Where a section compares several things at once, each has to clear a stricter bar, so that one doesn't stand out by chance alone.
Made 18 Nov 2026 · runner 0.1.0-5e42a45 · engine 0.1.0-b15ca3b