bayesopt.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 1. Import functions\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import timeit\n",
    "import subprocess\n",
    "import random\n",
    "import numpy as np\n",
    "import scipy as sp\n",
    "import math\n",
    "import re\n",
    "import chess\n",
    "from bayes_opt import BayesianOptimization\n",
    "from operator import itemgetter\n",
    "from chess import uci\n",
    "from chess import Board\n",
    "from chess import Move\n",
    "from chess import syzygy\n",
    "from numpy import sqrt\n",
    "from scipy.stats import chi2\n",
    "from scipy.stats import norm\n",
    "from statistics import median"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2. Getting Started\n",
    "\n",
    "All we need to get started is to instanciate a `BayesianOptimization` object specifying a function to be optimized `f`, and its parameters with their corresponding bounds, `pbounds`. This is a constrained optimization technique, so you must specify the minimum and maximum values that can be probed for each parameter in order for it to work"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "Engines = [\n",
    "    {'file': 'C:\\\\msys2\\\\home\\\\lanto\\\\safechecks\\\\tune.exe', 'name': 'test'},\n",
    "    {'file': 'C:\\\\msys2\\\\home\\\\lanto\\\\safechecks\\\\tune.exe', 'name': 'base'}\n",
    "    ]\n",
    "\n",
    "Draw = {'movenumber': 40, 'movecount': 8, 'score': 20}\n",
    "Resign = {'movecount': 3, 'score': 400}\n",
    "population_size=40\n",
    "iterations=200\n",
    "dynamic_rate=5\n",
    "Openings = 'C:\\\\Cutechess\\\\2moves.epd'\n",
    "Games = 50\n",
    "UseEngine = False\n",
    "Syzygy = 'C:\\\\Winboard\\\\Syzygy'\n",
    "ParametersFile = 'C:\\\\Rockstar\\\\safechecks.txt'\n",
    "LogFile = 'tuning.txt'\n",
    "DynamicConstraints = True\n",
    "\n",
    "Options = {'Clear Hash': True, 'Hash': 16, 'SyzygyPath': Syzygy, \\\n",
    "          'SyzygyProbeDepth': 10, 'Syzygy50MoveRule': True, 'SyzygyProbeLimit': 5}\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#  Preparatory phase\n",
    "\n",
    " takes parameters from the engine"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "def getPars():\n",
    "  sf = subprocess.Popen(Engines[0]['file'],  stdin=subprocess.PIPE, stdout=subprocess.PIPE, stderr=subprocess.STDOUT, universal_newlines=True, bufsize=1)\n",
    "  sf.stdin.write('isready' + '\\n')\n",
    "  pars = []\n",
    "  outline = []\n",
    "  while outline is not '':\n",
    "    outline = sf.stdout.readline().rstrip()\n",
    "    if not (outline.startswith('Stockfish ') or outline.startswith('Unknown ') or outline == ''):\n",
    "      pars.append(outline.split(','))\n",
    "  sf.terminate()\n",
    "  sf.wait()\n",
    "  return pars"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "takes parameters from file that is copied from engine output"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_pars():\n",
    "  params = []\n",
    "  f = open(ParametersFile)\n",
    "  lines = f.read().split('\\n')\n",
    "  if lines[-1] == '':\n",
    "    lines.remove('')\n",
    "  for p in lines:\n",
    "    params.append(p.split(','))\n",
    "  \n",
    "  return sorted(params)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "if UseEngine:\n",
    "  Pars = getPars()\n",
    "else:\n",
    "  Pars = get_pars()\n",
    "\n",
    "# openings\n",
    "def get_fens():\n",
    "  fens = []\n",
    "  lines = open(Openings).read().splitlines()\n",
    "  for i in range(0, Games, 1):\n",
    "    fen =random.choice(lines)\n",
    "    fens.append(fen)\n",
    "#  print(fens)\n",
    "  return fens\n",
    "\n",
    "def shuffled(x):\n",
    "  y = x[:]\n",
    "  random.shuffle(y)\n",
    "  return y"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "def init_engines(pars):\n",
    "  info_handlers = []\n",
    "  uciEngines = []\n",
    "  for e in Engines:\n",
    "    uciEngines.append(uci.popen_engine(e['file']))\n",
    "\n",
    "  for e,u in enumerate(uciEngines):\n",
    "    u.uci()\n",
    "    u.setoption(Options)\n",
    "    u.setoption(pars[e])\n",
    "    u.isready()\n",
    "\n",
    "  return uciEngines"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "class DifferentialEvolution():\n",
    "\n",
    "  def __init__(self):\n",
    "    self.nameArray = [str(par[0]) for par in Pars]\n",
    "    self.parsArray = [int(par[1]) for par in Pars]\n",
    "    self.bounds = [(int(p[2]), int(p[3])) for p in Pars]\n",
    "    self.n_parameters = len(self.nameArray)\n",
    "    self.current = [int(par[1]) for par in Pars]\n",
    "    self.trial = [int(par[1]) for par in Pars]\n",
    "    self.pbounds = dict(zip(self.nameArray, self.bounds))\n",
    "\n",
    "###  Evaluation\n",
    "  def evaluate(self, variables):\n",
    "    num = 0\n",
    "    fens = get_fens()\n",
    "    trial = dict(zip(self.nameArray, self.trial))\n",
    "    result = []\n",
    "    with syzygy.open_tablebases(Syzygy) as tablebases:\n",
    "      for fen in fens:\n",
    "        result1 = self.trans_result(self.launchSf([variables, trial], fen, tablebases,))\n",
    "        result2 = self.trans_result(self.launchSf([trial, variables], fen, tablebases,))\n",
    "        result.append(result1 + result2)\n",
    "    pentares = self.pentanomial(result)\n",
    "    curr = float(self.calc_los(pentares))\n",
    "    return curr\n",
    "\n",
    "  def trans_result(self, score):\n",
    "    return {'1-0': 2, '1/2-1/2': 1, '0-1': 0}[score]\n",
    "\n",
    "  def pentanomial(self, result):\n",
    "    pentares = []\n",
    "    for i in range(0,5):\n",
    "      pentares.append(result.count(i))\n",
    "    return pentares\n",
    "\n",
    "  def calc_los(self, pentares):\n",
    "    sumi, sumi2 = 0, 0\n",
    "    for i in range(0,5):\n",
    "      res = 0.5 * i\n",
    "      N = sum(pentares)\n",
    "      sumi += pentares[i] * res / N\n",
    "      sumi2 += pentares[i] * res * res / N\n",
    "    sigma = math.sqrt(sumi2 - sumi * sumi)\n",
    "    try:\n",
    "      t = math.sqrt(N) * (sumi - 1) / sigma * 100\n",
    "    except ZeroDivisionError:\n",
    "      t = 0.0\n",
    "#    los = norm.cdf(t) * 100\n",
    "#    return '{0:.2f}'.format(round(t, 2))\n",
    "    return t\n",
    "\n",
    "###  Game playing\n",
    "  def launchSf(self, pars, fen, tablebases,):\n",
    "    try:\n",
    "      board = Board(fen,chess960=False)\n",
    "    except BaseException:\n",
    "      try:\n",
    "        board.set_epd(fen)\n",
    "      except BaseException:\n",
    "        board = Board(chess960=False)\n",
    "    wdl = None\n",
    "    drawPlyCnt, resignPlyCnt = 0, 0\n",
    "    whiteIdx = 1\n",
    "    turnIdx = whiteIdx ^ (board.turn == chess.BLACK)\n",
    "    uciEngines = init_engines(pars)\n",
    "    info_handler = uci.InfoHandler()\n",
    "    for u in uciEngines:\n",
    "      u.info_handlers.append(info_handler)\n",
    "      u.ucinewgame()\n",
    "\n",
    "    try:\n",
    "      while (not board.is_game_over(claim_draw=True)):\n",
    "\n",
    "        if board.castling_rights == 0:\n",
    "\n",
    "#          if len(re.findall(r\"[rnbqkpRNBQKP]\", board.board_fen())) < 6:\n",
    "#            wdl = tablebases.probe_wdl(board)\n",
    "#            if wdl is not None:\n",
    "#              break                       # ~ 1.5 ms\n",
    "\n",
    "          try:\n",
    "            wdl = tablebases.probe_wdl(board)\n",
    "            if wdl is not None:\n",
    "              break\n",
    "          except KeyError:\n",
    "            pass                           # < 1 ms\n",
    "\n",
    "        uciEngines[turnIdx].position(board)\n",
    "        bestmove, score = uciEngines[turnIdx].go(depth=9)\n",
    "        score = info_handler.info[\"score\"][1].cp\n",
    "#        print(score)\n",
    "\n",
    "        if score is not None:\n",
    "            # Resign adjudication\n",
    "            if abs(score) >= Resign['score']:\n",
    "                resignPlyCnt += 1\n",
    "                if resignPlyCnt >= 2 * Resign['movecount']:\n",
    "                    break\n",
    "            else:\n",
    "                resignPlyCnt = 0\n",
    "\n",
    "            # Draw adjudication\n",
    "            if abs(score) <= Draw['score'] and board.halfmove_clock > 0:\n",
    "                drawPlyCnt += 1\n",
    "                if drawPlyCnt >= 2 * Draw['movecount'] \\\n",
    "                        and board.fullmove_number >= Draw['movenumber']:\n",
    "                    break\n",
    "            else:\n",
    "                drawPlyCnt = 0\n",
    "        else:\n",
    "            # Disable adjudication over mate scores\n",
    "            drawPlyCnt, resignPlyCnt = 0, 0\n",
    "\n",
    "        board.push(bestmove)\n",
    "        turnIdx ^= 1\n",
    "\n",
    "      result = board.result(True)\n",
    "      if result == '*':\n",
    "        if resignPlyCnt >= 2 * Resign['movecount']:\n",
    "          if score > 0:\n",
    "            result = '1-0' if board.turn == chess.WHITE else '0-1'\n",
    "          else:\n",
    "            result = '0-1' if board.turn == chess.WHITE else '1-0'\n",
    "        elif wdl is not None:\n",
    "          if wdl <= -1:\n",
    "            result = '1-0' if board.turn == chess.WHITE else '0-1'\n",
    "          elif wdl >= 1:\n",
    "            result = '0-1' if board.turn == chess.WHITE else '1-0'\n",
    "          else:\n",
    "            result = '1/2-1/2'\n",
    "#            print('tb draw')\n",
    "        else:\n",
    "          result = '1/2-1/2'\n",
    "#          print('draw')\n",
    "\n",
    "  #    print(board.fen())\n",
    "  #    print(re.findall(r\"[rnbqkpRNBQKP]\", board.board_fen()))\n",
    "      for u in uciEngines:\n",
    "        u.quit(0)\n",
    "    except (MemoryError, SystemError, KeyboardInterrupt,\n",
    "    OverflowError, OSError, ResourceWarning):\n",
    "      pass\n",
    "#    print(result)\n",
    "    return result\n",
    "    exit(0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "|   iter    |  target   | Bishop... | Knight... | QueenS... | RookSa... |\n",
      "-------------------------------------------------------------------------\n",
      "| \u001b[0m 1       \u001b[0m | \u001b[0m 32.57   \u001b[0m | \u001b[0m 436.0   \u001b[0m | \u001b[0m 794.3   \u001b[0m | \u001b[0m 782.1   \u001b[0m | \u001b[0m 880.9   \u001b[0m |\n",
      "| \u001b[95m 2       \u001b[0m | \u001b[95m 81.36   \u001b[0m | \u001b[95m 433.5   \u001b[0m | \u001b[95m 792.9   \u001b[0m | \u001b[95m 778.8   \u001b[0m | \u001b[95m 887.8   \u001b[0m |\n",
      "| \u001b[95m 3       \u001b[0m | \u001b[95m 125.7   \u001b[0m | \u001b[95m 434.4   \u001b[0m | \u001b[95m 796.2   \u001b[0m | \u001b[95m 780.7   \u001b[0m | \u001b[95m 874.3   \u001b[0m |\n",
      "| \u001b[0m 4       \u001b[0m | \u001b[0m 36.16   \u001b[0m | \u001b[0m 433.5   \u001b[0m | \u001b[0m 782.5   \u001b[0m | \u001b[0m 772.5   \u001b[0m | \u001b[0m 875.9   \u001b[0m |\n",
      "| \u001b[0m 5       \u001b[0m | \u001b[0m 118.8   \u001b[0m | \u001b[0m 435.0   \u001b[0m | \u001b[0m 781.6   \u001b[0m | \u001b[0m 784.3   \u001b[0m | \u001b[0m 884.9   \u001b[0m |\n",
      "| \u001b[0m 6       \u001b[0m | \u001b[0m-91.9    \u001b[0m | \u001b[0m 426.0   \u001b[0m | \u001b[0m 789.9   \u001b[0m | \u001b[0m 790.0   \u001b[0m | \u001b[0m 889.7   \u001b[0m |\n",
      "| \u001b[0m 7       \u001b[0m | \u001b[0m 71.8    \u001b[0m | \u001b[0m 425.8   \u001b[0m | \u001b[0m 787.2   \u001b[0m | \u001b[0m 784.3   \u001b[0m | \u001b[0m 889.3   \u001b[0m |\n",
      "| \u001b[95m 8       \u001b[0m | \u001b[95m 197.0   \u001b[0m | \u001b[95m 425.4   \u001b[0m | \u001b[95m 792.4   \u001b[0m | \u001b[95m 779.7   \u001b[0m | \u001b[95m 886.7   \u001b[0m |\n",
      "| \u001b[0m 9       \u001b[0m | \u001b[0m 49.36   \u001b[0m | \u001b[0m 431.8   \u001b[0m | \u001b[0m 788.6   \u001b[0m | \u001b[0m 777.4   \u001b[0m | \u001b[0m 879.5   \u001b[0m |\n",
      "| \u001b[0m 10      \u001b[0m | \u001b[0m 133.6   \u001b[0m | \u001b[0m 431.6   \u001b[0m | \u001b[0m 786.2   \u001b[0m | \u001b[0m 770.5   \u001b[0m | \u001b[0m 885.0   \u001b[0m |\n",
      "| \u001b[0m 11      \u001b[0m | \u001b[0m 48.62   \u001b[0m | \u001b[0m 427.8   \u001b[0m | \u001b[0m 785.6   \u001b[0m | \u001b[0m 773.1   \u001b[0m | \u001b[0m 887.8   \u001b[0m |\n",
      "| \u001b[0m 12      \u001b[0m | \u001b[0m-50.13   \u001b[0m | \u001b[0m 427.2   \u001b[0m | \u001b[0m 785.8   \u001b[0m | \u001b[0m 789.5   \u001b[0m | \u001b[0m 877.1   \u001b[0m |\n",
      "| \u001b[0m 13      \u001b[0m | \u001b[0m-65.37   \u001b[0m | \u001b[0m 427.6   \u001b[0m | \u001b[0m 787.1   \u001b[0m | \u001b[0m 784.5   \u001b[0m | \u001b[0m 876.7   \u001b[0m |\n",
      "| \u001b[0m 14      \u001b[0m | \u001b[0m 0.0     \u001b[0m | \u001b[0m 425.5   \u001b[0m | \u001b[0m 793.0   \u001b[0m | \u001b[0m 779.9   \u001b[0m | \u001b[0m 887.1   \u001b[0m |\n",
      "| \u001b[0m 15      \u001b[0m | \u001b[0m 36.7    \u001b[0m | \u001b[0m 437.3   \u001b[0m | \u001b[0m 786.1   \u001b[0m | \u001b[0m 789.7   \u001b[0m | \u001b[0m 886.7   \u001b[0m |\n",
      "| \u001b[0m 16      \u001b[0m | \u001b[0m 36.16   \u001b[0m | \u001b[0m 433.3   \u001b[0m | \u001b[0m 799.0   \u001b[0m | \u001b[0m 774.5   \u001b[0m | \u001b[0m 877.1   \u001b[0m |\n",
      "| \u001b[0m 17      \u001b[0m | \u001b[0m 12.6    \u001b[0m | \u001b[0m 444.8   \u001b[0m | \u001b[0m 797.2   \u001b[0m | \u001b[0m 788.6   \u001b[0m | \u001b[0m 880.0   \u001b[0m |\n",
      "| \u001b[0m 18      \u001b[0m | \u001b[0m 116.2   \u001b[0m | \u001b[0m 433.0   \u001b[0m | \u001b[0m 799.5   \u001b[0m | \u001b[0m 770.6   \u001b[0m | \u001b[0m 887.2   \u001b[0m |\n",
      "| \u001b[0m 19      \u001b[0m | \u001b[0m 125.6   \u001b[0m | \u001b[0m 426.4   \u001b[0m | \u001b[0m 781.0   \u001b[0m | \u001b[0m 785.1   \u001b[0m | \u001b[0m 882.0   \u001b[0m |\n",
      "| \u001b[0m 20      \u001b[0m | \u001b[0m-86.15   \u001b[0m | \u001b[0m 425.4   \u001b[0m | \u001b[0m 783.3   \u001b[0m | \u001b[0m 781.3   \u001b[0m | \u001b[0m 885.1   \u001b[0m |\n",
      "| \u001b[0m 21      \u001b[0m | \u001b[0m 123.2   \u001b[0m | \u001b[0m 436.7   \u001b[0m | \u001b[0m 795.8   \u001b[0m | \u001b[0m 770.6   \u001b[0m | \u001b[0m 887.4   \u001b[0m |\n",
      "| \u001b[95m 22      \u001b[0m | \u001b[95m 242.3   \u001b[0m | \u001b[95m 434.1   \u001b[0m | \u001b[95m 784.4   \u001b[0m | \u001b[95m 787.9   \u001b[0m | \u001b[95m 881.3   \u001b[0m |\n",
      "| \u001b[0m 23      \u001b[0m | \u001b[0m 136.9   \u001b[0m | \u001b[0m 439.3   \u001b[0m | \u001b[0m 797.5   \u001b[0m | \u001b[0m 774.5   \u001b[0m | \u001b[0m 879.8   \u001b[0m |\n",
      "| \u001b[0m 24      \u001b[0m | \u001b[0m-86.15   \u001b[0m | \u001b[0m 429.7   \u001b[0m | \u001b[0m 799.2   \u001b[0m | \u001b[0m 779.8   \u001b[0m | \u001b[0m 888.7   \u001b[0m |\n",
      "| \u001b[0m 25      \u001b[0m | \u001b[0m 65.37   \u001b[0m | \u001b[0m 441.3   \u001b[0m | \u001b[0m 795.3   \u001b[0m | \u001b[0m 770.7   \u001b[0m | \u001b[0m 874.8   \u001b[0m |\n",
      "| \u001b[0m 26      \u001b[0m | \u001b[0m 165.6   \u001b[0m | \u001b[0m 438.9   \u001b[0m | \u001b[0m 786.7   \u001b[0m | \u001b[0m 771.7   \u001b[0m | \u001b[0m 871.0   \u001b[0m |\n",
      "| \u001b[0m 27      \u001b[0m | \u001b[0m 65.37   \u001b[0m | \u001b[0m 429.7   \u001b[0m | \u001b[0m 798.7   \u001b[0m | \u001b[0m 775.9   \u001b[0m | \u001b[0m 883.6   \u001b[0m |\n",
      "| \u001b[95m 28      \u001b[0m | \u001b[95m 266.3   \u001b[0m | \u001b[95m 426.5   \u001b[0m | \u001b[95m 784.4   \u001b[0m | \u001b[95m 781.8   \u001b[0m | \u001b[95m 874.2   \u001b[0m |\n",
      "| \u001b[0m 29      \u001b[0m | \u001b[0m 59.55   \u001b[0m | \u001b[0m 438.5   \u001b[0m | \u001b[0m 795.2   \u001b[0m | \u001b[0m 788.6   \u001b[0m | \u001b[0m 878.2   \u001b[0m |\n",
      "| \u001b[0m 30      \u001b[0m | \u001b[0m-36.7    \u001b[0m | \u001b[0m 432.6   \u001b[0m | \u001b[0m 797.2   \u001b[0m | \u001b[0m 784.3   \u001b[0m | \u001b[0m 887.9   \u001b[0m |\n",
      "| \u001b[0m 31      \u001b[0m | \u001b[0m-35.16   \u001b[0m | \u001b[0m 435.7   \u001b[0m | \u001b[0m 786.9   \u001b[0m | \u001b[0m 787.1   \u001b[0m | \u001b[0m 889.9   \u001b[0m |\n",
      "| \u001b[0m 32      \u001b[0m | \u001b[0m 11.87   \u001b[0m | \u001b[0m 438.9   \u001b[0m | \u001b[0m 797.0   \u001b[0m | \u001b[0m 782.3   \u001b[0m | \u001b[0m 884.0   \u001b[0m |\n",
      "=========================================================================\n",
      "{'target': 266.31396209587814, 'params': {'BishopSafeCheck': 426.5468446744946, 'KnightSafeCheck': 784.4251834641562, 'QueenSafeCheck': 781.8090358145664, 'RookSafeCheck': 874.2404789009539}}\n"
     ]
    }
   ],
   "source": [
    "if __name__ == '__main__':\n",
    "  de = DifferentialEvolution()\n",
    "\n",
    "  variables = dict(zip(de.nameArray, de.current))\n",
    "  def black_box_function(**variables):\n",
    "    f = de.evaluate(variables)\n",
    "    return f\n",
    "\n",
    "  pbounds = dict(zip(de.nameArray, de.bounds))\n",
    "  \n",
    "  optimizer = BayesianOptimization(\n",
    "    f=black_box_function,\n",
    "    pbounds=pbounds,\n",
    "    verbose=2, # verbose = 1 prints only when a maximum is observed, verbose = 0 is silent\n",
    "    random_state=0,\n",
    "  )\n",
    "\n",
    "  optimizer.maximize(\n",
    "    init_points=2,\n",
    "    n_iter=30,\n",
    "    acq='poi', \n",
    "  )\n",
    "  print(optimizer.max)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}