ifchess

Piece Set

Army Matchup

White

Bot level — 1500 Elo

Think time (only affects "Max") ms

Black

Bot level — 1500 Elo

Think time (only affects "Max") ms

Starting position

Create custom pieces here. They're saved to your library and become available when you build a set.

Name Key (FEN letter)?The single letter used for this piece in FEN/notation. Must not clash with the standard pieces p n b r q k. Uppercase = White, lowercase = Black. Movement (Betza)?Betza notation — how the piece moves.
Leapers: W 1-step orthogonal, F 1-step diagonal, N knight, D (2,0), A (2,2), H (3,0).
Sliders: R rook, B bishop, Q queen. A number limits range — Q2 = queen up to 2 squares.
Modifiers: m move-only, c capture-only, p hop over a screen (cannon). Combine atoms, e.g. BN = bishop+knight (Archbishop).

Values ?Engine values (MG / EG) in centipawns (pawn ≈ 126). MG = middlegame, EG = endgame. The small Value field is just a rough 1–9 display scale.

Piece	MG	EG
Pawn	126	208
Knight	781	854
Bishop	825	915
Rook	1276	1380
Queen	2538	2682
Calibrated fairy pieces (this project)
Princess `Q2`	1250	1420
Cross `DH`	810	890
Cannon	850	730
Duchess `BNW`	2540	2630
Archbishop `BN`	2100	2200
Chancellor `RN`	2250	2400
Nightrider `NN`	1400	1300
Amazon `QN`	3300	3500

Value Value MG Value EG

Description

Avatar ?Pick a medieval icon for this piece, or “Letter” to show its key. Each icon has a light (White) and dark (Black) version drawn to match the standard pieces.

Your pieces

Ready to benchmark 0/1 games active Elapsed 00:00:00

?Two modes
Strength — plays your army vs a traditional army at the piece's current values; reports win rate + Elo. Answers "is my army stronger?"
Calibrate — sweeps MG/EG values for one piece to find the value that makes it balanced (≈50%) against the piece it replaces.

Set to test

Plays a custom army vs a traditional army and reports how the replacement scores (win rate + Elo).

Engine Matchups?Each custom piece alone drops one custom piece into an otherwise-traditional army and tests it in isolation, then also runs the full army — so you see each piece's own contribution. Full only tests just the whole custom army (faster).

Games in parallel

Fairy-Stockfish nodes / move Max plies Openings?Opening book
So games don't all start identically, each begins from a short balanced opening. Openings = N uses the first N from a built-in book, and each is played by both sides (colours reversed) per repetition — so N openings = N×2 games per matchup/cell per rep.

Book order: start, 1.e4 e5, 1.d4 d5, 1.c4 c5, 1.f4 f5, 1.g3 g6, 1.b3 b6, 1.e3 e6, 1.d3 d6, 1.c3 c6, 1.e4 c5, 1.e4 e6, 1.e4 c6, 1.e4 d6, 1.e4 g6, 1.d4 g6, 1.d4 f5, 1.d4 c5, 1.c4 e5, 1.c4 g6, … (30 total). Any Extra lines you add below are appended to these. Repetitions

Ignore ply-cap draws in the score ?Ply cap = the maximum number of half-moves (plies — one move by one side) a game may reach before it's cut off and scored as a draw, set by “Max plies”. It stops a single stuck game from stalling the whole run. These forced draws aren't real chess outcomes, so ticking this leaves them out of the score (they're still counted and shown as “Cap draws”). Toggle it any time — results re-score instantly, no re-run. Extra opening lines (one per line, UCI moves e.g. e2e4 e7e5)

Results?Reading results
Score = wins + ½·draws, from the custom side's view: 50% = balanced, above 50% = custom favoured.
Elo converts score to a rating gap (positive favours custom).
CI = 95% bootstrap confidence range.
In calibration, each heatmap cell is one MG/EG pair — aim for a stable green region near 50%, not one lucky cell.

Game log Run details, retries & errors

New Game

White

Black

Position Editor

Add / edit pieces

Build a set

Piece Benchmark