Understand the failure.
Verify the repair.
Small, reproducible software failures. The broken implementation, the fix that didn’t work, and the one that passed—preserved together.
Explore the cases ↓How results are verified ↗WHAT THE ARCHIVE CONTAINS
100840 executable cases. 20168 are open.
Every case records the implementation that fails, the fix that did not work, and the repair that passed its checks—with recorded outputs and source hashes. This release adds 100840 cases across 20168 failure mechanisms and 254 domains.
The open tier gives you the failure and the unsuccessful fix for one case in every mechanism. The remaining 80672 cases, 5 variants per mechanism, are member-only: the verified repair, its recorded checks, and the full fixture suite are held in the member archive. Read the methodology ↗
A RECORD OF WHAT WENT WRONG
Browse the archive / 100840
Connector bonus granted to high cards · case 01
Queen-jack and ace-king receive the low-connector bonus point.
Connector bonus granted to high cards · case 02
Queen-jack and ace-king receive the low-connector bonus point.
Connector bonus granted to high cards · case 03
Queen-jack and ace-king receive the low-connector bonus point.
Connector bonus granted to high cards · case 04
Queen-jack and ace-king receive the low-connector bonus point.
Connector bonus granted to high cards · case 05
Queen-jack and ace-king receive the low-connector bonus point.
Chen score rounded to even instead of up · case 01
Nine-seven offsuit scores 2 instead of 3.
Chen score rounded to even instead of up · case 02
Nine-seven offsuit scores 2 instead of 3.
Chen score rounded to even instead of up · case 03
Nine-seven offsuit scores 2 instead of 3.
Chen score rounded to even instead of up · case 04
Nine-seven offsuit scores 2 instead of 3.
Chen score rounded to even instead of up · case 05
Nine-seven offsuit scores 2 instead of 3.
Paired badugi counted as four cards · case 01
A-A-3-4 in four suits beats a genuine king-high badugi.
Paired badugi counted as four cards · case 02
A-A-3-4 in four suits beats a genuine king-high badugi.
Paired badugi counted as four cards · case 03
A-A-3-4 in four suits beats a genuine king-high badugi.
Paired badugi counted as four cards · case 04
A-A-3-4 in four suits beats a genuine king-high badugi.
Paired badugi counted as four cards · case 05
A-A-3-4 in four suits beats a genuine king-high badugi.
Worst qualifying subset chosen for a badugi · case 01
A three-card hand is valued by its highest valid subset rather than its lowest.
Worst qualifying subset chosen for a badugi · case 02
A three-card hand is valued by its highest valid subset rather than its lowest.
Worst qualifying subset chosen for a badugi · case 03
A three-card hand is valued by its highest valid subset rather than its lowest.
Worst qualifying subset chosen for a badugi · case 04
A three-card hand is valued by its highest valid subset rather than its lowest.
Worst qualifying subset chosen for a badugi · case 05
A three-card hand is valued by its highest valid subset rather than its lowest.
Higher badugi wins the pot · case 01
A-2-3-8 beats A-2-4-7 although the eight-high badugi is the worse hand.
Higher badugi wins the pot · case 02
A-2-3-8 beats A-2-4-7 although the eight-high badugi is the worse hand.
Higher badugi wins the pot · case 03
A-2-3-8 beats A-2-4-7 although the eight-high badugi is the worse hand.
Higher badugi wins the pot · case 04
A-2-3-8 beats A-2-4-7 although the eight-high badugi is the worse hand.
Higher badugi wins the pot · case 05
A-2-3-8 beats A-2-4-7 although the eight-high badugi is the worse hand.
Badugi search settles for one-card subsets · case 01
Every hand evaluates as a one-card hand, so comparisons depend only on the lowest card.
Badugi search settles for one-card subsets · case 02
Every hand evaluates as a one-card hand, so comparisons depend only on the lowest card.
Badugi search settles for one-card subsets · case 03
Every hand evaluates as a one-card hand, so comparisons depend only on the lowest card.
Badugi search settles for one-card subsets · case 04
Every hand evaluates as a one-card hand, so comparisons depend only on the lowest card.
Badugi search settles for one-card subsets · case 05
Every hand evaluates as a one-card hand, so comparisons depend only on the lowest card.
Reports the most hole cards that can play · case 01
When the board plays, the hand is annotated as using both hole cards.
Reports the most hole cards that can play · case 02
When the board plays, the hand is annotated as using both hole cards.
Reports the most hole cards that can play · case 03
When the board plays, the hand is annotated as using both hole cards.
Reports the most hole cards that can play · case 04
When the board plays, the hand is annotated as using both hole cards.
Reports the most hole cards that can play · case 05
When the board plays, the hand is annotated as using both hole cards.
Board declared to play when only the category matches · case 01
A player whose ace kicker matters is told the board plays.
Board declared to play when only the category matches · case 02
A player whose ace kicker matters is told the board plays.
Board declared to play when only the category matches · case 03
A player whose ace kicker matters is told the board plays.
Board declared to play when only the category matches · case 04
A player whose ace kicker matters is told the board plays.
Board declared to play when only the category matches · case 05
A player whose ace kicker matters is told the board plays.
Knight on the h-file jumps to the a-file · case 01
A knight on h3 is offered a2 and a4, which wrap around the board edge.
Knight on the h-file jumps to the a-file · case 02
A knight on h3 is offered a2 and a4, which wrap around the board edge.
Knight on the h-file jumps to the a-file · case 03
A knight on h3 is offered a2 and a4, which wrap around the board edge.
Knight on the h-file jumps to the a-file · case 04
A knight on h3 is offered a2 and a4, which wrap around the board edge.
Knight on the h-file jumps to the a-file · case 05
A knight on h3 is offered a2 and a4, which wrap around the board edge.
Knight target off the top edge accepted · case 01
A knight on c7 is offered a ninth-rank square named "a9".
Knight target off the top edge accepted · case 02
A knight on c7 is offered a ninth-rank square named "a9".
Knight target off the top edge accepted · case 03
A knight on c7 is offered a ninth-rank square named "a9".
Knight target off the top edge accepted · case 04
A knight on c7 is offered a ninth-rank square named "a9".
Knight target off the top edge accepted · case 05
A knight on c7 is offered a ninth-rank square named "a9".
Knight captures its own side but not the enemy · case 01
The knight may land on a friendly bishop but not capture an enemy pawn.
Knight captures its own side but not the enemy · case 02
The knight may land on a friendly bishop but not capture an enemy pawn.
Knight captures its own side but not the enemy · case 03
The knight may land on a friendly bishop but not capture an enemy pawn.
Knight captures its own side but not the enemy · case 04
The knight may land on a friendly bishop but not capture an enemy pawn.
Knight captures its own side but not the enemy · case 05
The knight may land on a friendly bishop but not capture an enemy pawn.
Mistyped knight offset produces a pawn-like step · case 01
A knight is offered the square two ranks straight below it.
Mistyped knight offset produces a pawn-like step · case 02
A knight is offered the square two ranks straight below it.
Mistyped knight offset produces a pawn-like step · case 03
A knight is offered the square two ranks straight below it.
Mistyped knight offset produces a pawn-like step · case 04
A knight is offered the square two ranks straight below it.
Mistyped knight offset produces a pawn-like step · case 05
A knight is offered the square two ranks straight below it.
Rank digit used as a zero-based row · case 01
Every knight target is reported one rank too high.
Rank digit used as a zero-based row · case 02
Every knight target is reported one rank too high.
Rank digit used as a zero-based row · case 03
Every knight target is reported one rank too high.
Rank digit used as a zero-based row · case 04
Every knight target is reported one rank too high.
Rank digit used as a zero-based row · case 05
Every knight target is reported one rank too high.
Slider x-rays through an enemy piece · case 01
A rook facing an enemy pawn is offered the squares behind it, including the enemy king square.
Slider x-rays through an enemy piece · case 02
A rook facing an enemy pawn is offered the squares behind it, including the enemy king square.
Slider x-rays through an enemy piece · case 03
A rook facing an enemy pawn is offered the squares behind it, including the enemy king square.
Slider x-rays through an enemy piece · case 04
A rook facing an enemy pawn is offered the squares behind it, including the enemy king square.
Slider x-rays through an enemy piece · case 05
A rook facing an enemy pawn is offered the squares behind it, including the enemy king square.
Rays run onto a ninth file and rank · case 01
A rook on a1 is offered "i1" and "a9".
Rays run onto a ninth file and rank · case 02
A rook on a1 is offered "i1" and "a9".
Rays run onto a ninth file and rank · case 03
A rook on a1 is offered "i1" and "a9".
Rays run onto a ninth file and rank · case 04
A rook on a1 is offered "i1" and "a9".
Rays run onto a ninth file and rank · case 05
A rook on a1 is offered "i1" and "a9".
Queen moves only like a rook · case 01
A queen is never offered diagonal moves.
Queen moves only like a rook · case 02
A queen is never offered diagonal moves.
Queen moves only like a rook · case 03
A queen is never offered diagonal moves.
Queen moves only like a rook · case 04
A queen is never offered diagonal moves.
Queen moves only like a rook · case 05
A queen is never offered diagonal moves.
Bishop never moves toward a1 · case 01
A bishop on f5 is not offered e4, d3 or the capture on c2, and some targets repeat.
Bishop never moves toward a1 · case 02
A bishop on f5 is not offered e4, d3 or the capture on c2, and some targets repeat.
Bishop never moves toward a1 · case 03
A bishop on f5 is not offered e4, d3 or the capture on c2, and some targets repeat.
Bishop never moves toward a1 · case 04
A bishop on f5 is not offered e4, d3 or the capture on c2, and some targets repeat.
Bishop never moves toward a1 · case 05
A bishop on f5 is not offered e4, d3 or the capture on c2, and some targets repeat.
Pawn double step jumps over a blocker · case 01
A pawn on e2 with a knight on e3 is offered e4.
Pawn double step jumps over a blocker · case 02
A pawn on e2 with a knight on e3 is offered e4.
Pawn double step jumps over a blocker · case 03
A pawn on e2 with a knight on e3 is offered e4.
Pawn double step jumps over a blocker · case 04
A pawn on e2 with a knight on e3 is offered e4.
Pawn double step jumps over a blocker · case 05
A pawn on e2 with a knight on e3 is offered e4.
Black pawns cannot make their initial double step · case 01
A black pawn on d7 is offered only d6.
Black pawns cannot make their initial double step · case 02
A black pawn on d7 is offered only d6.
Black pawns cannot make their initial double step · case 03
A black pawn on d7 is offered only d6.
Black pawns cannot make their initial double step · case 04
A black pawn on d7 is offered only d6.
Black pawns cannot make their initial double step · case 05
A black pawn on d7 is offered only d6.
Pawn captures its own pieces diagonally · case 01
A white pawn on d4 may capture the white bishop on e5.
Pawn captures its own pieces diagonally · case 02
A white pawn on d4 may capture the white bishop on e5.
Pawn captures its own pieces diagonally · case 03
A white pawn on d4 may capture the white bishop on e5.
Pawn captures its own pieces diagonally · case 04
A white pawn on d4 may capture the white bishop on e5.
Pawn captures its own pieces diagonally · case 05
A white pawn on d4 may capture the white bishop on e5.
INSPECTABLE BY DESIGN
Every result has a runnable source.
Runnable implementations with recorded outputs, source hashes, and explicit contracts. Related variants share a failure mechanism and belong together in evaluation splits.
Read the methodology ↗