自己制作适合6寸阅读器的棋书

网上有许多棋谱，例如 棋谱 - 象棋巫师棋谱仓库。在 象棋资源 - 象棋百科全书 下载象棋巫师，选项—微博选项，图片类型选印刷，魔法—生成图片棋盘：

   

或者 印刷用棋盘工具带源码 ：(棋子上的字是DrawText出来的，可以方便地更换)

在WPS Office里输入(比如)： 

1. 炮二平五 马8进7 2. 马二进三 车9平8
3. 兵七进一 卒7进1 4. 马八进七 炮2平5

按Ctrl+Enter换页，然后粘贴棋盘图片。如此重复直至结束。然后文件—输出为PDF。[ 成品例子 ]

阅读方法：在脑海里走棋，然后翻到下页印证下，如此重复。

象棋巫师—文件—查看文本棋谱并从"1. 炮二平五"复制到结束，再运行下面的Python程序：

# -*- coding: gbk -*-
# pip install pywin32
import win32clipboard as w
import win32con as C
def get_text():
    w.OpenClipboard(); s = w.GetClipboardData(C.CF_UNICODETEXT); w.CloseClipboard()
    return s
def set_text(s):
    w.OpenClipboard(); w.EmptyClipboard();
    w.SetClipboardData(C.CF_UNICODETEXT, s); w.CloseClipboard()
out = ''; i = 1
for s in get_text().splitlines():
    out += s.strip() + ('\n' if i % 2 == 0  else '　')
    i += 1
print(out); set_text(out)

该程序读取剪贴板，把两行合并成一行，然后再复制到剪贴板。

下载象棋巫师后，可先用你信任的各种杀毒软件检查下，然后在Windows安全中心里把它所在的目录加入排除项。这样它自带的引擎ELEEYE.exe的启动速度会快很多。Windows安全中心搞歧视，我用VC6和tdm-gcc编译出来的程序，每次运行都要被扫一扫。python.exe好像也被扫。VC6工程里加入VERSION也没用。

舍得花几百元买皮肤，舍不得花90元买象棋旋风？ELO等级分计算公式

• 象棋旋风基础版(单核版)售价90元，在低档电脑上具有相当于特级大师的对弈水平
• 象棋旋风专业版(双核版)售价180元，在普通电脑上具有超越特级大师的对弈水平

讲道理，付款是有点麻烦——我没有装支付宝。

365行的象棋程序 | 666行的象棋程序下得不错啊

An efficiently updatable neural network (NNUE, a Japanese wordplay on Nue, sometimes stylised as ƎUИИ) is a neural network -based evaluation function whose inputs are piece-square tables, or variants thereof like the king-piece-square table. NNUE is used primarily for the leaf nodes of the alpha–beta tree.

红黑双方各有7种棋子，合计14种，因此可把局面看作一张9x10的图片，每个像素16种颜色——对于人脸识别来说好像算微数据啊。类似的技术应该可以对于任意一个局面，给出它的分数？

评价函数 | http://www.ics.uci.edu/~eppstein/180a/970415.html (Lecture notes for April 15, 1997)

Intel CPU 发展简史：1997年5月7日，英特尔发布Pentium II 233MHz、Pentium II 266MHz、Pentium II 300MHz三款处理器，采用了0.35微米工艺技术，核心提升到750万个晶体管。0.3GHz主频，350nm工艺，0.075亿个晶体管，0.5MB cache。1998年6月25日，Windows 98正式发布，要求英特尔0.066GHz或更高的处理器；内存最低为0.000008TB，建议64MB(小于64MB会降低性能)，最高只能支持256MB。和下面这个比：

1P=1000T=100万G，11.5 petaflops大约是0.3G的38,333,333倍。AlphaZero的训练9小时的那么多倍是39,383年，我咋觉得还退步了涅？:-)

In 2017, DeepMind announced a new engine called AlphaZero, which significantly improved on the prior AlphaGo. While AlphaGo only was able to play Go, and was trained using human games as an input, AlphaZero was trained from scratch, and could play Chess, Shogi, and Go - all at state of the art levels.

engines - Which is better-Stockfish 10 or AlphaZero? - Chess Stack Exchange

AlphaZero beat Stockfish 8 by +155 = 839 -6, which is an elo difference of about 50. The latest versions of Stockfish are capable of beating Stockfish 8 by about 150 elo. 现在Stockfish 16有beta版了，AlphaZero呢？

不提识别率光吹算力，是怕别人不知道算法其实进步不大么？:-) QuickSort也就几十行，ShellSort的gap都很有学问：Shell sort

DeepMind was acquired by Google in 2014. [买来的啊] Alpha Zero works slightly differently. The key to AZ is the Monte Carlo Tree Search (MCTS) algorithm. MCTS is a four-step process to building out a tree. First, AZ will select a new node in the tree. This selection process weights both exploration and exploitation using the PUCT algorithm, meaning that more promising nodes (the nodes which seem better so far) tend to be explored more often, but nodes that haven't been explored much so far also get a chance for expansion.

Monte Carlo Tree Search was introduced by Rémi Coulom in 2006 as a building block of Crazy Stone – Go playing engine with an impressive performance. Rémi Coulom, a French computer scientist, freelance programmer in the field of artificial intelligence in games, and former associate professor (Maître de Conférences) of computer science at the Charles de Gaulle University. His generic AlphaZero implementation dubbed CrazyZero has networks trained to play Go, Shogi, Gomoku, Othello, Renju, and Chess. 法国计算机科学家，游戏中的人工智能领域的自由程序员，戴高乐大学计算机科学前副教授。他的通用AlphaZero实现被称为CrazyZero，其网络经过训练可以玩围棋，将棋，五子棋，奥赛罗，连珠和国际象棋。In the French system and other francophone systems, tenured faculty is divided in two categories: maitre de conferences, equivalent to associate professors if I understand correctly, and professeurs, equivalent to full professors [link].

Artificial Intelligence - Chessprogramming wiki: Poker, the next Challenge.

没有提推箱子，比如Something Random - Sokoban Online. 国际象棋，人类冠军能想14步，程序想10步，下得就有模有样了，看起来很智能。首先推箱子不能推14步然后宣布能推出来——象棋对方走一步你沉思半天宣称自己必胜？(武侠小说)。其次就算能推出来，听起来也远不如8*8的国际象棋智能。围棋19*19，可是棋子种类少很多。