What Does That Little Bar on the Left Actually Mean?
When you analyze a game on ChessPadi, the first thing you see alongside the chessboard is the Evaluation Bar and a numeric indicator like +1.45 or -2.80.
Most players understand the basics: positive numbers favor White, negative numbers favor Black, and #3 means checkmate in 3 moves. But how does an engine calculate these values, what is Average Centipawn Loss (ACPL), and how should a human player interpret them to actually gain rating points?
The Centipawn: The Currency of Chess Engines
In computer chess, the fundamental unit of measurement is the centipawn (cp).
100 Centipawns = 1 Pawn of Material Advantage
1 Centipawn = 0.01 PawnsHere is a practical conversion guide for human comprehension:
| Evaluation Score | What It Means in Practice | Realistic Human Win Rate |
|---|---|---|
| 0.00 to ±0.30 | Dead equal or balanced dynamic tension | ~50% Draw / Equal Chances |
| +0.50 to +0.90 | Slight pull for White (natural opening advantage) | 58% White Win / 35% Draw |
| +1.00 to +1.80 | Clear advantage (healthy extra pawn or monster knight) | 75% Win Rate |
| +2.00 to +3.50 | Winning advantage (extra minor piece or crushing attack) | 90%+ Win Rate |
| +4.00 and above | Completely decisive (opponent can resign with dignity) | 99% Win Rate |
| #N (e.g. #4) | Forced checkmate in N moves under best play | 100% Win Rate |
What Is Average Centipawn Loss (ACPL) in Chess?
Average Centipawn Loss (ACPL) is a mathematical metric that measures the quality and accuracy of a player's chess moves throughout a game. It calculates the average difference, measured in centipawns, between the optimal move recommended by the chess engine and the move the player actually executed on the board.
In simple terms:
- A lower ACPL means you played closer to engine perfection.
- A higher ACPL means your moves leaked tactical or positional advantages.
How Centipawn Loss Is Calculated per Move
On every single turn, Stockfish determines the best possible move in the position and calculates its evaluation score. When you play your move, the engine compares your move's evaluation to the best move:
Centipawn Loss per Move = Eval(Engine Best Move) - Eval(Move Played)Concrete Example:
- 1The position is balanced: Stockfish finds an optimal move evaluating at +1.20 (+120 centipawns).
- 2You play a natural-looking move that drops the evaluation to +0.70 (+70 centipawns).
- 3Your Centipawn Loss for that turn is:
120 - 70 = 50 centipawns (half a pawn lost)- 1If you play the engine's #1 top move, your Centipawn Loss for that turn is 0.
The Average Centipawn Loss (ACPL) Formula
To calculate your Average Centipawn Loss across the entire match, the engine adds up all your centipawn losses and divides by the total number of moves you played:
ACPL = (Total Centipawn Loss Across All Moves) ÷ (Total Number of Moves Played)For example, if you played 40 moves and accumulated a total centipawn loss of 1,200 centipawns:
ACPL = 1,200 ÷ 40 = 30 ACPLAverage Centipawn Loss Benchmarks by Elo Rating
Rather than focusing merely on whether you won or lost, examining your ACPL across 20 to 30 games reveals your true playing caliber:
| Rating Tier | Typical Elo (FIDE / Online) | Expected ACPL Range | Playing Characteristics |
|---|---|---|---|
| Super-GM / World Champion | 2700+ FIDE | 8 – 18 ACPL | Near-machine precision; virtually zero tactical oversights |
| Grandmaster (GM) | 2500 – 2700 FIDE | 15 – 25 ACPL | Flawless tactical basics; minor positional inaccuracies only |
| Master / Expert | 2000 – 2400 Elo | 25 – 40 ACPL | Rare one-move blunders; strong opening and endgame execution |
| Club Advanced | 1600 – 1900 Elo | 40 – 60 ACPL | Solid plans; 1-2 tactical slips or missed defensive refutations |
| Club Intermediate | 1200 – 1600 Elo | 60 – 85 ACPL | Frequent tactical inaccuracies, hanging pawns, missed forks |
| Beginner / Novice | Under 1200 Elo | 85 – 140+ ACPL | Multiple game-ending blunders per match |
ACPL vs. Accuracy Percentage: Which Metric Is Better?
Most players are familiar with Chess.com's "Accuracy %" (e.g., 88.4%). How does that relate to ACPL?
- ACPL is linear and raw: Losing 100 centipawns is always 100 centipawns, regardless of whether the position is +8.0 or 0.0.
- Accuracy % is non-linear: Online platforms pass your ACPL through a sigmoid win-probability curve. In a completely winning position (+7.0), playing a sloppy move that drops the eval to +5.0 barely lowers your Accuracy % because your win probability remains near 99%. However, your ACPL still records the 200 centipawn drop!
Why serious players track ACPL: ACPL does not lie. It measures absolute calculation precision. If your goal is to reach 2000 Elo, aim to bring your median ACPL below 35, regardless of what the accuracy percentage says.
Why Equal Material Often Evaluates at +3.00
Beginners often get confused when looking at a board where both players have 16 pawns and pieces, yet Stockfish displays +3.20.
How can White be "up 3 pawns" when no material has left the board?
Modern engines—especially Stockfish 18 powered by NNUE—evaluate three critical non-material dimensions:
- 1King Safety: An uncastled king stuck in the center facing open files is often penalized -2.0 to -4.0 pawns because of incoming mating nets.
- 2Piece Mobility & Outposts: A knight cemented on
d5controlling 8 critical squares can be worth more than an enemy rook trapped behind its own pawns. - 3Pawn Structure Health: Doubled, isolated, and backward pawns represent long-term structural liabilities that engines ruthlessly exploit in endgames.
Why a Single Blunder Distorts Your ACPL
Be aware of how blunders affect the math.
If you play 39 moves at a Grandmaster level (averaging 12 centipawns lost per move), but on move 40 you accidentally hang your queen (losing 900 centipawns), your game ACPL jumps from 12 to:
(39 × 12 + 900) ÷ 40 = 34.2 ACPLA single tactical blackout doubled your entire game's ACPL! This is why modern reviews must separate tactical blunders from steady positional inaccuracies.
How ChessPadi Makes Engine Data Useful
Raw numbers on a screen do not teach you chess. ChessPadi bridges the gap between machine metrics and human intuition:
- Move-by-Move Centipawn Tracking: See exactly which moves leaked 20, 50, or 200+ centipawns.
- GM-Level Coaching Commentary: Understand the strategic idea behind why a move lost centipawns.
- Full Spoken Voiceover: Replay the game with synchronized grandmaster narration.
- Practice Studio: Step directly onto the board from your biggest blunder and test whether you can find the winning continuation against Stockfish.
Check Your Average Centipawn Loss on ChessPadi
Ready to see your real calculation precision and find out which moves are holding back your rating?
Import your PGN from Chess.com, Lichess, or an over-the-board tournament into ChessPadi for complete game analysis, ACPL metrics, and interactive voice coaching.
Mastery Action Checklist
Follow these essential steps for consistent improvement
Understand the Centipawn Scale
In chess engines, 100 centipawns equal 1 pawn of material advantage. An evaluation of +1.50 means White has an advantage equivalent to one and a half pawns.
Calculate Centipawn Loss per Move
On every ply, subtract the evaluation of your played move from the evaluation of the engine's best move. If the top move is +1.80 and you play +1.20, you lost 60 centipawns on that turn.
Track Your Game Average Centipawn Loss (ACPL)
Sum all centipawn losses across your moves and divide by the total number of moves played. Focus on lowering your long-term ACPL across 20-30 games rather than obsessing over single win/loss results.
Recognize Positional vs Material Value
Do not confuse evaluation with piece count. A +3.00 evaluation with equal material means White possesses commanding king safety, piece activity, or passed pawn advantages that equate to a minor piece.
Spar Alternative Lines in Practice Studio
When you identify a move with high centipawn loss, step into ChessPadi's Practice Studio to play out the engine's refutation until the defensive or attacking habit becomes second nature.
Frequently Asked Questions
What is average centipawn loss (ACPL) in chess?
Average Centipawn Loss (ACPL) is a mathematical metric that measures the quality and accuracy of a player's chess moves. It calculates the average difference, in hundredths of a pawn (centipawns), between the engine's best move evaluation and the move actually played. A lower ACPL indicates higher precision, with grandmasters typically averaging under 25 ACPL and club players scoring between 45 and 75 ACPL.
What is considered a good average centipawn loss?
For club and intermediate players (1200-1800 Elo), an ACPL between 40 and 65 is considered solid. Advanced players (1800-2200 Elo) typically average between 25 and 40 ACPL, while titled Grandmasters consistently achieve an ACPL under 20 in classical tournament games.
What is the difference between ACPL and Accuracy percentage?
ACPL is raw and linear: it directly measures the exact centipawns lost per turn. Accuracy percentage (such as CAPS formulas) transforms ACPL into a non-linear 0-100% score based on win probability curves. Serious chess analysts prefer ACPL because it does not artificially inflate or compress errors.
Why did I get a high ACPL even though I won the game?
ACPL measures every move of the game, not just the final outcome. If you played an inaccurate opening or blundered a piece before your opponent blundered their queen back, your high centipawn losses from the early mistakes remain in your game average.
What does an evaluation of 0.00 mean?
A 0.00 evaluation indicates dynamic or dead equality under perfect play. However, in sharp positions with both kings exposed, a 0.00 evaluation can be extremely difficult for humans to navigate without making a game-ending mistake.
What is NNUE in Stockfish?
NNUE stands for Efficiently Updatable Neural Network. Unlike older engines that used hand-crafted heuristic rules, NNUE uses a neural network directly on the CPU to evaluate piece coordination and king vulnerability with superhuman nuance.