Perft function example

[tissatussa]

I expected some basic Perft implementation in the README, but it's not there.
Can you provide such code ? How to traverse the variation tree ?

[tissatussa]

I created this Perft script :

package main

import (
	"fmt"
	"github.com/corentings/chess/v2"
	"log"
	"os"
	"encoding/json"
	"time"
)

type FenConfig struct {
	Desc   string `json:"desc"`
	Fen    string `json:"fen"`
	Active bool   `json:"active"`
}

// Perft recursively counts all leaf nodes up to the given depth.
func Perft(pos *chess.Position, depth int) int {
	if depth == 0 {
		return 1
	}
	moves := pos.ValidMoves()
	if depth == 1 {
		return len(moves)
	}
	nodes := 0
	for _, move := range moves {

		// Clone position to avoid mutating original
		next := pos.Update(&move)
		nodes += Perft(next, depth-1)
	}
	return nodes
}

func main() {
	// Read JSON config
	data, err := os.ReadFile("config.json")
	if err != nil {
		log.Fatalf("Error reading config: %v", err)
	}

	var configs []FenConfig
	if err := json.Unmarshal(data, &configs); err != nil {
		log.Fatalf("Error parsing JSON: %v", err)
	}

	// Find the active FEN
	var activeFen string
	for _, cfg := range configs {
		if cfg.Active {
			activeFen = cfg.Fen
			fmt.Printf("Using FEN: %s\nDescription: %s\n", cfg.Fen, cfg.Desc)
			break
		}
	}
	if activeFen == "" {
		log.Fatal("No active FEN found in config.")
	}

	// Set up chess position
	fen, err := chess.FEN(activeFen)
	if err != nil {
		log.Fatalf("Invalid FEN: %v", err)
	}
	game := chess.NewGame(fen)

	fmt.Println(game.Position().Board().Draw())

	// depth := 1
	// depth := 2
	// depth := 3
	// depth := 4
	depth := 5

	tmStart := time.Now()

	nodes := Perft(game.Position(), depth)

	fmt.Printf("Perft(%d): %d\n", depth, nodes)
	tm := time.Since(tmStart)
	fmt.Printf("Elapsed time: %s\n", tm)
}

It works, but the famous Kiwipete position fails at perft 5 : the amount of nodes should be 193690690 but my script gives 193686252. I tested other FENs and let stockfish (which correctly shows Perft 5 for Kiwipete) calculate perft values upto depth 5.

It seems my script correctly handles ep captures, castling and (minor) promotions, it shows correct Perft values for all other positions mentioned in https://www.chessprogramming.org/Perft_Results

So why does it fail Perft 5 for Kiwipete ?
Does my script or your library contain a bug ?

[tissatussa]

Here's go.mod :

module urmom

go 1.22.0

toolchain go1.22.2

require github.com/corentings/chess/v2 v2.0.8
require golang.org/x/exp v0.0.0-20250128182459-e0ece0dbea4c // indirect

Here's go.sum :

github.com/corentings/chess/v2 v2.0.8 h1:QWsd8AD+7aUwCWPEVLZ5K/nQrztkCQnYaBqSLDHLgzA=
github.com/corentings/chess/v2 v2.0.8/go.mod h1:JhWYDbjY81/7NECXrLzz4g2r9taaMEXvyqS4gYZciVE=
golang.org/x/exp v0.0.0-20250128182459-e0ece0dbea4c h1:KL/ZBHXgKGVmuZBZ01Lt57yE5ws8ZPSkkihmEyq7FXc=
golang.org/x/exp v0.0.0-20250128182459-e0ece0dbea4c/go.mod h1:tujkw807nyEEAamNbDrEGzRav+ilXA7PCRAd6xsmwiU=

[tissatussa]

Here's config.json (to use a FEN, set "active" true) :

[
  {
    "desc": "Standard Start position @@@ 1:20 @@@ 2:400 @@@ 3:8902 @@@ 4:197281",
    "fen": "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq - 0 1",
    "active": false
  },
  {
    "desc": "Kiwipete @@@ 1:48 @@@ 2:2039 @@@ 3:97862 @@@ 4:4085603 @@@ 5:193690690",
    "fen": "r3k2r/p1ppqpb1/bn2pnp1/3PN3/1p2P3/2N2Q1p/PPPBBPPP/R3K2R w KQkq - 0 1",
    "active": true
  },
  {
    "desc": "@@@ 1:14 @@@ 2:191 @@@ 3:2812 @@@ 4:43238 @@@ 5:674624",
    "fen": "8/2p5/3p4/KP5r/1R3p1k/8/4P1P1/8 w - - 0 1 ",
    "active": false
  },
  {
    "desc": "@@@ 1:6 @@@ 2:264 @@@ 3:9467 @@@ 4:422333 @@@ 5:15833292",
    "fen": "r3k2r/Pppp1ppp/1b3nbN/nP6/BBP1P3/q4N2/Pp1P2PP/R2Q1RK1 w kq - 0 1",
    "active": false
  },
  {
    "desc": "@@@ 1:44 @@@ 2:1486 @@@ 3:62379 @@@ 4:2103487 @@@ 5:89941194",
    "fen": "rnbq1k1r/pp1Pbppp/2p5/8/2B5/8/PPP1NnPP/RNBQK2R w KQ - 1 8",
    "active": false
  },
  {
    "desc": "White pawn promotes @@@ 1:6 @@@ 2:66 @@@ 3:460 @@@ 4:5849 @@@ 5:36616",
    "fen": "8/4P3/8/4K3/2k5/8/8/8 b - - 0 1",
    "active": false
  },
  {
    "desc": "Black can capture ep at first move @@@ 1:8 @@@ 2:54 @@@ 3:438 @@@ 4:3350 @@@ 5:26670",
    "fen": "8/8/8/4K3/2k2Pp1/8/8/8 b - f3 0 1",
    "active": false
  },
  {
    "desc": "Black can capture ep in future @@@ 1:7 @@@ 2:63 @@@ 3:525 @@@ 4:4690 @@@ 5:38237",
    "fen": "8/8/8/4K3/2k3p1/8/5P2/8 b - - 0 1",
    "active": false
  },
  {
    "desc": "Both sides can castle @@@ 1:27 @@@ 2:629 @@@ 3:15718 @@@ 4:360451 @@@ 5:8925435",
    "fen": "r3k2r/8/8/8/6p1/8/5P2/R3K2R b KQkq - 0 1",
    "active": false
  },
  {
    "desc": "Include minor promotions @@@ 1:4",
    "fen": "8/2P5/8/8/8/8/1p6/K6k w - - 0 1",
    "active": false
  }
]

[tissatussa]

I think it's all about the function ValidMoves() : if you provide a decent Perft, this will show and prove that function has no bug. My script is simple, i see no fault ..

[CorentinGS]

Hello,

I'll have a look at it during the coming week. If you have time to explore further and identify areas for improvement in the library, feel free to describe them here or open a pull request (PR).
Thank you for raising the issue. I'll do my best as soon as possible to work on this.

[tissatussa]

@CorentinGS hi, in the meantime i'm using dragontoothmg https://github.com/dylhunn/dragontoothmg with Go, it has a fast Perft function, the Kiwipete FEN node count is accurate, i also build an own perft, counting captures, checks etc. with its move generator. It's not on GitHub, i'm not familiar with git commands yet .. did you ever create a chess engine ?