[{"data":1,"prerenderedAt":1124},["ShallowReactive",2],{"\u002Fzh\u002Fglossary\u002Fwhat-is-a-z-score-in-trading":3,"\u002Fzh\u002Fglossary\u002Fwhat-is-a-z-score-in-trading-related":242,"\u002Fzh\u002Fglossary\u002Fwhat-is-a-z-score-in-trading-langs":1120},{"id":4,"title":5,"body":6,"category":224,"date":225,"description":226,"draft":227,"extension":228,"meta":229,"navigation":230,"path":231,"related":232,"seo":237,"stem":238,"term":239,"updated":240,"__hash__":241},"glossary\u002Fzh\u002F5.glossary\u002Fwhat-is-a-z-score-in-trading.md","交易中的 Z 值是什么？连胜连亏还是偶然",{"type":7,"value":8,"toc":216},"minimark",[9,13,17,24,35,42,45,67,74,77,86,94,102,105,108,132,148,152,172,179,182,209],[10,11,12],"p",{},"交易中的 Z 值，看的是盈利和亏损是扎堆出现还是交替出现：先数出实际序列里有多少个游程，再和同样数量的盈亏随机排列时应有的游程数比较。它不说明账户赚没赚钱，只说明上一笔的结果对下一笔有没有提示作用。",[14,15,16],"h2",{"id":16},"工作原理",[10,18,19,23],{},[20,21,22],"strong",{},"游程","指一段连续的同号结果。序列 W W L W L L 有四个游程。所用的统计量是 Wald–Wolfowitz 游程检验：",[25,26,31],"pre",{"className":27,"code":29,"language":30},[28],"language-text","R  = 2WL \u002F N + 1                    （期望游程数）\nσ² = 2WL(2WL − N) \u002F ( N² (N − 1) )  （方差）\nZ  = (游程数 − R) \u002F σ\n\nW = 盈利交易数，L = 亏损交易数，N = W + L\n","text",[32,33,29],"code",{"__ignoreMap":34},"",[10,36,37,38,41],{},"交易按平仓时间排序，盈亏按",[20,39,40],{},"净","结果判定，佣金和库存费都已计入。一笔仓位毛利为正、扣掉成本后净亏，在这里就算亏损，并且会打断一段连胜的游程。",[10,43,44],{},"看符号就知道怎么读：",[46,47,48,55,61],"ul",{},[49,50,51,54],"li",{},[20,52,53],{},"Z 为负","：游程比随机情况少。结果扎堆，赢了接着赢，亏了接着亏。",[49,56,57,60],{},[20,58,59],{},"Z 为正","：游程比随机情况多。盈亏交替得比抛硬币还频繁。",[49,62,63,66],{},[20,64,65],{},"接近零","：和相互独立的交易没有区别。",[10,68,69,70,73],{},"账户页面上，Z 值后面的括号里还有一个百分比：把同一个值压缩到 0–100 的刻度上，",[20,71,72],{},"50% 表示和抛硬币一样随机","。低于 50% 说明比随机更容易连续，高于 50% 说明更容易交替；Z 为 −2.41 时显示 8.24%，Z 为 +1.00 时显示 73.11%。这只是换个更好读的说法，不是 p 值，所以 8.24% 不是显著性水平。",[14,75,76],{"id":76},"为什么重要",[10,78,79,80,85],{},"策略统计几乎都默认交易之间相互独立。Z 值一旦显示并非如此，页面上其他数字的含义就会悄悄变样。亏损扎堆出现时，",[81,82,84],"a",{"href":83},"\u002Fglossary\u002Fwhat-is-drawdown","回撤","会比同样胜率下预想的更深，因为回撤不光看亏了几笔，还看这些亏损是怎么排列的。",[10,87,88,89,93],{},"Z 值明显为负，原因往往并不陌生。网格、逆势加仓摊平和篮子系统会围绕同一个判断开出很多仓位，再一起平掉，于是一个决定就产生一长串同号结果。凡是在亏损仓位上逐步加仓的做法都是如此，这类账户通常在净值回撤和",[81,90,92],{"href":91},"\u002Fglossary\u002Fwhat-is-drawdown-on-balance","余额回撤","之间也有很大差距。",[10,95,96,97,101],{},"Z 值明显为正，通常说明账户里跑的根本不是一个序列。对冲仓位会在同一时刻各记下一笔盈利和一笔亏损；两个互不相关的策略在同一个登录号上交错运行，会制造出一种不属于任何一方的人为交替。交易如果带有",[81,98,100],{"href":99},"\u002Fglossary\u002Fwhat-is-a-magic-number","魔术号","，只有按交易流分开看，才能得到一个可信的数值。",[10,103,104],{},"它的实际用途不宽，但很实在：Z 值为负，是给连续亏损设规则的统计依据；接近零，则说明这类规则没有必要。",[14,106,107],{"id":107},"数据显示了什么",[10,109,110,111,114,115,119,120,123,124,127,128,131],{},"以下数字只描述在 ShowMyTrades 上发布的账户，不代表全体交易者。中位数的已发布账户有 ",[20,112,113],{},"171 笔已平仓交易","，",[81,116,118],{"href":117},"\u002Fguides\u002Fwin-rate-is-not-an-edge","胜率","为 ",[20,121,122],{},"68.8%","，大约是 118 笔盈利、53 笔亏损。按这个比例，随机情况下预计约有 ",[20,125,126],{},"74 个游程","，一个标准差约 ",[20,129,130],{},"5.6 个游程","。",[10,133,134,135,138,139,142,143,147],{},"也就是说，在中位数账户上，游程数要缩到 63 个左右或超过 85 个，Z 值才会到 ±2。落在这两个界限之间，都和抛硬币没有区别。这实实在在限制了这个数字的分量，而历史越长，界限收得越紧：同样的规律出现在 1,700 笔交易上算是发现，出现在 171 笔上只算线索。还有一点：这些账户中 ",[20,136,137],{},"53.9%"," 的自动化程度超过 90%，",[20,140,141],{},"42.2%"," 低于 10%。扎堆现象在前一组里要常见得多，因为",[81,144,146],{"href":145},"\u002Fglossary\u002Fwhat-is-an-expert-advisor","机器人","平掉一篮子订单时，会一次往序列里写进好几个相同的结果。",[14,149,151],{"id":150},"在-showmytrades-上的位置","在 ShowMyTrades 上的位置",[10,153,154,155,158,159,162,163,166,167,171],{},"这项指标在每个已发布账户页面的",[20,156,157],{},"高级统计","模块里，名为 ",[20,160,161],{},"Z 值（概率）","，显示两位小数的数值，旁边是换算后的百分比，位于",[20,164,165],{},"表现指标","一栏，和盈利因子、标准差、夏普比率、期望值、",[81,168,170],{"href":169},"\u002Fglossary\u002Fwhat-is-ahpr-and-ghpr","AHPR 和 GHPR"," 排在一起。数值由我们的统计服务在账户同步时算好，不是在您的浏览器里计算。",[10,173,174,175,178],{},"同一页面的 ",[20,176,177],{},"Smart Insights"," 面板也会据此给出一条结论：账户至少有 30 笔已平仓交易、且数值在 ±2 以内时，会提示交易模式稳定。",[14,180,181],{"id":181},"常见误解",[46,183,184,191,197,203],{},[49,185,186,187,190],{},"“",[20,188,189],{},"Z 值为负是坏事。","”它不衡量好坏，只描述交易之间是否相互关联，和赚不赚钱无关。很多优秀的趋势系统，Z 值都是负的。",[49,192,186,193,196],{},[20,194,195],{},"它能预测下一笔交易。","”它总结的是已经发生的序列。游程检验里没有任何预测的成分。",[49,198,186,199,202],{},[20,200,201],{},"这个百分比就是系统有效的概率。","”它只是同一个值换了个刻度，仅此而已。百分之五十是随机的中点，不是及格线。",[49,204,186,205,208],{},[20,206,207],{},"顺序无所谓。","”顺序恰恰是唯一要紧的。同一秒内平仓的几笔交易，彼此之间没有真正意义上的先后，所以批量平掉篮子订单的系统，游程数可能因为和策略无关的原因而变化。",[10,210,211,212,131],{},"这类统计量需要多少笔交易才有意义，请参阅",[81,213,215],{"href":214},"\u002Fguides\u002Fhow-much-history-a-track-record-needs","一份业绩记录需要多长的历史",{"title":34,"searchDepth":217,"depth":217,"links":218},2,[219,220,221,222,223],{"id":16,"depth":217,"text":16},{"id":76,"depth":217,"text":76},{"id":107,"depth":217,"text":107},{"id":150,"depth":217,"text":151},{"id":181,"depth":217,"text":181},"指标","2026-08-19T00:00:00.000Z","交易中的 Z 值把盈亏的连续程度和随机情况做比较。本文介绍我们的 Z 值（概率）怎么算，以及为什么 171 笔交易的样本还很单薄。",false,"md",{},true,"\u002Fzh\u002Fglossary\u002Fwhat-is-a-z-score-in-trading",[233,234,235,236],"what-is-ahpr-and-ghpr","what-is-drawdown-on-balance","what-is-a-magic-number","what-is-expectancy",{"title":5,"description":226},"zh\u002F5.glossary\u002Fwhat-is-a-z-score-in-trading","Z 值 (Z-Score)","2026-09-17T00:00:00.000Z","EaJTDRCihu0CO4ef0bSKbKcJ2LSNVoL-2S08WChZfBg",[243,464,658,885],{"id":244,"title":245,"body":246,"category":450,"date":225,"description":451,"draft":227,"extension":228,"meta":452,"navigation":230,"path":453,"related":454,"seo":459,"stem":460,"term":461,"updated":462,"__hash__":463},"glossary\u002Fzh\u002F5.glossary\u002Fwhat-is-a-magic-number.md","什么是 MT4 和 MT5 中的魔术号？",{"type":7,"value":247,"toc":443},[248,260,262,265,271,274,294,297,299,307,310,312,322,343,345,374,408,410,436],[10,249,250,251,254,255,259],{},"魔术号是 ",[81,252,253],{"href":145},"EA（智能交易系统）","写进每笔订单的一个整数，",[81,256,258],{"href":257},"\u002Fglossary\u002Fwhat-is-metatrader","MetaTrader"," 会在这笔交易的整个生命周期里一直保留它。有了它，多个策略可以共用一个账户，事后照样分得清。手动下的单，魔术号是零。",[14,261,16],{"id":16},[10,263,264],{},"这个数字是下单函数的一个参数，由代码设定，经纪商和平台都不管。",[25,266,269],{"className":267,"code":268,"language":30},[28],"MT4:  OrderSend(symbol, cmd, lots, price, slip, sl, tp, comment, magic, ...)\nMT5:  request.magic = 55566;   → 带到仓位和每一笔成交上\n手动点击:  magic = 0\n",[32,270,268],{"__ignoreMap":34},[10,272,273],{},"由此带来三个特点，每一个都容易让人踩坑。",[46,275,276,282,288],{},[49,277,278,281],{},[20,279,280],{},"没有人校验它","。没有任何环节检查这个值是否唯一、有没有意义。两个 EA 设了同一个整数，在记录里就永远混在一起，而且各自都会把对方的仓位当成自己的来管理。",[49,283,284,287],{},[20,285,286],{},"零是一个实实在在的值，不是空值","。手动点击下单带的是零，魔术号参数一直没改默认值的 EA 带的也是零，两者从此就分不开了。",[49,289,290,293],{},[20,291,292],{},"事后补不上","。请求离开终端的那一刻，这个值就定死了，并随交易一起保存；当初是零的历史，永远都是零。",[10,295,296],{},"其他平台用别的字段表达同样的意思：cTrader 上的 cBot 写的是文本标签，TradeLocker 在订单上附带策略 ID，其中数字形式的 ID 在本站按魔术号读取。实际上，按策略拆分主要还是 MT4 和 MT5 的事。",[14,298,76],{"id":76},[10,300,301,302,306],{},"一个账户跑三个策略，页面上每个核心数字都是三者混在一起的结果。",[81,303,305],{"href":304},"\u002Fglossary\u002Fwhat-is-a-profit-factor","盈利因子","、胜率、回撤和净值曲线都按合并后的数据计算，所以一个稳定亏钱的策略可以在盈利的总账里藏上好几个月，任何图表上都看不到它单独的那条线。账户看起来很健康，因为从总体上看，它确实健康。",[10,308,309],{},"合并之后还能把策略区分开的，只有魔术号这一个字段。设得好，历史一进来就已经按策略分成了一份份独立的业绩记录，可以互相比较，分别调整仓位，或者一个个停掉。设得马虎，事后再怎么分析也还原不出来。",[14,311,107],{"id":107},[10,313,314,315,114,318,321],{},"这个字段在本站的分量，比大多数账户所有者想象的更重，因为每个账户页面上的自动化比例就是靠它算出来的：一笔交易只要带着非零魔术号，或者在用策略标签代替魔术号的平台上带着策略标签，就算作自动交易。在 ShowMyTrades 上有交易历史的公开账户中（2026 年 8 月），",[20,316,317],{},"自动交易占比中位数为 99%",[20,319,320],{},"53.9% 的账户自动化程度超过 90%","。换句话说，这些账户里几乎每一笔订单进来时都已经带好了标记。",[10,323,324,325,328,329,331,332,336,337,114,340,131],{},"之所以要尽早打标记，是因为样本量。",[20,326,327],{},"账户的已平仓交易数中位数为 171 笔","。分到三个策略上，每个只有五十七笔左右，还不足以让",[81,330,118],{"href":117},"或盈利因子说明什么。从第一笔订单就打标记，每个策略才能在您必须对它做决定之前，攒够一个看得懂的样本。另一种做法是",[81,333,335],{"href":334},"\u002Fguides\u002Ftracking-multiple-accounts-portfolio","一个策略一个账户","，在本站也很常见：",[20,338,339],{},"699 位用户有不止一个账户",[20,341,342],{},"429 位在不止一家经纪商交易",[14,344,151],{"id":150},[10,346,347,350,351,354,355,358,359,362,363,366,367,369,370,373],{},[20,348,349],{},"分类统计","模块（账户设置里叫",[20,352,353],{},"货币对统计","）在",[20,356,357],{},"按品种","旁边有一个",[20,360,361],{},"按魔术号","标签页，只要账户有可以拆分的已平仓交易就会出现。每一行显示 ",[20,364,365],{},"EA \u002F 策略","、原始",[20,368,100],{},"、交易数、盈利，以及盈利单和亏损单的数量与百分比。魔术号为 0 的那一行标为",[20,371,372],{},"手动","，也是唯一不能改名的一行：它不是一个策略，而是所有没带标记的交易。",[10,375,376,378,379,382,383,386,387,389,390,396,397,399,400,403,404,407],{},[20,377,365],{},"单元格可以直接编辑，55566 就这样变成一个半年后您还认得出的名字。这些名字之后会跟着数据走：",[20,380,381],{},"交易历史与余额变化","下的",[20,384,385],{},"已平仓交易","表格有一个",[20,388,100],{},"列，逐笔显示原始整数；",[81,391,393],{"href":392},"\u002Fguides\u002Fcustom-analysis-and-filter-presets",[20,394,395],{},"自定义分析","侧边面板有一个",[20,398,100],{},"筛选器，标签显示为",[20,401,402],{},"手动 (0)"," 或",[20,405,406],{},"标签 - 55566","。选中其中一个，页面上的所有统计都会只按这个策略重新计算。",[14,409,181],{"id":181},[46,411,412,418,424,430],{},[49,413,186,414,417],{},[20,415,416],{},"魔术号以后再设。","”没有以后。这个字段在下单时写入，过去的交易没法重新打标记。",[49,419,186,420,423],{},[20,421,422],{},"我的 EA 会自动搞定。","”前提是参数改过。默认值为 0 的情况很常见，会悄无声息地把 EA 和您的手动交易混在一起。",[49,425,186,426,429],{},[20,427,428],{},"标为手动，就是我手动下的单。","”手动这一行装的是所有没带标记的交易：手动下的单，加上所有魔术号停在零的 EA。",[49,431,186,432,435],{},[20,433,434],{},"魔术号能告诉您这是哪个机器人。","”什么也说明不了。55566 不过是作者填进去的数字。",[10,437,438,439,131],{},"从给策略命名，到揪出那个悄悄亏钱的策略，完整流程见",[81,440,442],{"href":441},"\u002Fguides\u002Fmagic-numbers-and-strategy-labels","魔术号与策略标签",{"title":34,"searchDepth":217,"depth":217,"links":444},[445,446,447,448,449],{"id":16,"depth":217,"text":16},{"id":76,"depth":217,"text":76},{"id":107,"depth":217,"text":107},{"id":150,"depth":217,"text":151},{"id":181,"depth":217,"text":181},"自动化","魔术号是 EA 给自己开的每笔订单打上的整数编号。有了它，一个账户里跑多个策略，也能逐个分开来看。",{},"\u002Fzh\u002Fglossary\u002Fwhat-is-a-magic-number",[455,456,457,458],"what-is-an-expert-advisor","what-is-metatrader","what-is-a-profit-factor","what-is-backtesting",{"title":245,"description":451},"zh\u002F5.glossary\u002Fwhat-is-a-magic-number","魔术号 (Magic Number)","2026-09-04T00:00:00.000Z","6MCBx5dRlfBGfCWb6cUQEPJaeMGLLDoA4TP0bWC1lng",{"id":465,"title":466,"body":467,"category":224,"date":225,"description":647,"draft":227,"extension":228,"meta":648,"navigation":230,"path":649,"related":650,"seo":654,"stem":655,"term":656,"updated":240,"__hash__":657},"glossary\u002Fzh\u002F5.glossary\u002Fwhat-is-ahpr-and-ghpr.md","什么是 AHPR 和 GHPR？公式与两者的差距",{"type":7,"value":468,"toc":640},[469,472,474,477,483,491,498,501,507,518,520,523,536,544,546,559,575,577,594,597,599,633],[10,470,471],{},"AHPR（平均持仓期收益率）和 GHPR（几何持仓期收益率）都是单笔交易的平均收益，以该笔交易之前的资金为基数，用百分比表示：AHPR 取算术平均，GHPR 取几何平均。两者衡量的是同一批交易，回答的却是不同的问题；两者之间的差距，反映了结果的波动有多大。",[14,473,16],{"id":16},[10,475,476],{},"每笔已平仓交易都有一个持仓期收益率，即该笔净结果除以交易前的资金：",[25,478,481],{"className":479,"code":480,"language":30},[28],"HPRᵢ = (盈利 + 库存费 + 佣金)ᵢ \u002F 第 i 笔交易之前的资金\n\nAHPR = ( Σ HPRᵢ \u002F N ) × 100\nGHPR = ( [ Π (1 + HPRᵢ) ] ^ (1\u002FN) − 1 ) × 100\n",[32,482,480],{"__ignoreMap":34},[10,484,485,486,490],{},"成本都在分子里：佣金和",[81,487,489],{"href":488},"\u002Fglossary\u002Fwhat-is-a-swap-in-forex","库存费","先计入每笔交易，再算 HPR，所以两个数字都是净值。",[10,492,493,494,497],{},"基数有个细节要注意。我们的实现以账户的",[20,495,496],{},"历史累计入金总额","作为滚动资金的起点，再逐笔加上每笔交易的净结果。分期注资的账户，早期交易是拿还没到账的钱来衡量的，两个数字都会被拉向零；只入金一次的账户，基数就和公式字面上的意思完全一致。",[10,499,500],{},"GHPR 永远不会高于 AHPR。这是算术，不是交易本身的规律：正数的几何平均数最多等于算术平均数，只有所有数字都相同时两者才相等。差额大约是收益方差的一半：",[25,502,505],{"className":503,"code":504,"language":30},[28],"GHPR ≈ AHPR − σ² \u002F 2\n",[32,506,504],{"__ignoreMap":34},[10,508,509,510,513,514,517],{},"两笔交易就能说明问题。先 +50% 再 −50%，AHPR 正好是 ",[20,511,512],{},"0%","，账户却只剩起点的 0.75。GHPR 是 √(1.50 × 0.50) − 1 = ",[20,515,516],{},"−13.4%","，复利两次正好落在 0.75。算术平均描述的是谁都没拿到的结果，几何平均才还原了真实的余额。",[14,519,76],{"id":76},[10,521,522],{},"算术平均数是花不出去的。只有几何平均值按交易笔数乘方之后，才能还原账户实际的表现，所以该引用的是 GHPR，AHPR 是拿来对照的标尺。两者差距大，就是所谓的波动率税：方差在复利落到余额之前先拿走的那一部分。",[10,524,525,526,530,531,535],{},"到了尾部，这种不对称会变得非常残酷。只要有一笔交易把资金全部亏光，乘积里就多了一个零，乘积也就是零：之前一千笔交易做得再好，GHPR 也显示 −100%。这就是",[81,527,529],{"href":528},"\u002Fglossary\u002Fwhat-is-position-sizing","仓位管理","比胜率更重要的道理，只是换成乘法来讲：一种方法即使 AHPR 不错，只要",[81,532,534],{"href":533},"\u002Fglossary\u002Fwhat-is-standard-deviation-in-trading","标准差","大到偶尔会押上全部资金，几何期望照样是负的。",[10,537,538,539,543],{},"AHPR 也不是摆设：它就是每笔交易的",[81,540,542],{"href":541},"\u002Fglossary\u002Fwhat-is-expectancy","期望值","，只是用资金的百分比来表示，而不是用点数或货币，所以能在不同规模的账户之间比较。两者放在一起看，能告诉您平均每笔交易赚了多少，以及扣掉方差的代价之后还剩多少。",[14,545,107],{"id":107},[10,547,110,548,114,550,554,555,558],{},[20,549,113],{},[81,551,553],{"href":552},"\u002Fglossary\u002Fwhat-is-time-weighted-return","时间加权收益率","中位数为 ",[20,556,557],{},"+3.2%","。把这个收益平摊到这么多笔交易上，每笔的几何收益约为 +0.018%，面板四舍五入后显示 +0.02%。这只是示意，不是实测的中位数，但它如实反映了这些数字的量级，也说明了为什么它们在几乎所有真实账户上都显得不起眼。",[10,560,561,562,554,564,567,568,570,571,574],{},"同一批账户的",[81,563,305],{"href":304},[20,565,566],{},"1.28","（已扣除库存费和佣金），胜率中位数为 ",[20,569,122],{},"。所以一个页面上",[81,572,573],{"href":117},"胜率看起来很高","、GHPR 却几乎为零，两者并不矛盾。要看的是正负号、和 AHPR 的差距，以及背后有多少笔交易。",[14,576,151],{"id":150},[10,578,579,580,582,583,586,587,590,591,593],{},"两者都在每个已发布账户页面的",[20,581,157],{},"模块里，标签就是 ",[20,584,585],{},"AHPR"," 和 ",[20,588,589],{},"GHPR","，位于",[20,592,165],{},"一栏底部，排在盈利因子、标准差、夏普比率、Z 值（概率）和期望值之后。两者都显示为带正负号、保留两位小数的百分比，大于等于零为绿色，小于零为红色。",[10,595,596],{},"这个模块有四个视图：表格、星期、小时和持续时间。AHPR 和 GHPR 在表格视图里；另外三个视图能看出是哪些交易造成了拉开两者差距的方差。两个数值都由我们的统计服务在账户同步时算好，不是在您的浏览器里计算。",[14,598,181],{"id":181},[46,600,601,607,621,627],{},[49,602,186,603,606],{},[20,604,605],{},"AHPR 是平均值，所以它才老实。","”这是一个谁都赚不到的平均值。只要结果有波动，它就会高估复利增长，高出的幅度约为方差的一半。",[49,608,186,609,612,613,616,617,620],{},[20,610,611],{},"它们是月收益或年收益。","”它们是按笔计算的。同样的 +0.02%，放在 20 笔交易和 2,000 笔交易上意义完全不同，所以要和同一模块",[20,614,615],{},"交易","一栏的",[20,618,619],{},"总交易数","对照着看。",[49,622,186,623,626],{},[20,624,625],{},"GHPR 接近零，说明系统不行。","”在正常的单笔风险下，两个数字本来就很小，面板又只保留两位小数。一个账户在几百笔交易上每笔显示 +0.02%，累计涨幅已是两位数。",[49,628,186,629,632],{},[20,630,631],{},"AHPR 为正、GHPR 为负，是 bug。","”这正是大幅波动的特征：收益平均下来高于零，复利结果却依然向下。这也是这对指标能告诉您的最有用的信息。",[10,634,635,636,131],{},"单笔交易指标如何放进一份要给别人看的业绩记录里，请参阅",[81,637,639],{"href":638},"\u002Fguides\u002Ftracking-trading-performance","交易表现追踪",{"title":34,"searchDepth":217,"depth":217,"links":641},[642,643,644,645,646],{"id":16,"depth":217,"text":16},{"id":76,"depth":217,"text":76},{"id":107,"depth":217,"text":107},{"id":150,"depth":217,"text":151},{"id":181,"depth":217,"text":181},"AHPR 和 GHPR 是每笔交易的算术平均收益和几何平均收益。本文介绍公式、几何平均为什么总是更低，以及数千个账户的实测结果。",{},"\u002Fzh\u002Fglossary\u002Fwhat-is-ahpr-and-ghpr",[236,651,652,653],"what-is-standard-deviation-in-trading","what-is-position-sizing","what-is-a-z-score-in-trading",{"title":466,"description":647},"zh\u002F5.glossary\u002Fwhat-is-ahpr-and-ghpr","AHPR 与 GHPR","-kX8ydWcwYp-iVY5E0IA7RC9EMTk0Vt5UJs8vlpjUJ4",{"id":659,"title":660,"body":661,"category":872,"date":225,"description":873,"draft":227,"extension":228,"meta":874,"navigation":230,"path":875,"related":876,"seo":881,"stem":882,"term":883,"updated":240,"__hash__":884},"glossary\u002Fzh\u002F5.glossary\u002Fwhat-is-drawdown-on-balance.md","什么是基于余额的回撤？余额回撤（DD on Balance）",{"type":7,"value":662,"toc":865},[663,681,683,686,692,695,698,701,727,729,732,739,742,755,757,783,797,799,816,830,832,858],[10,664,665,666,669,670,674,675,677,678,680],{},"基于余额的回撤，是账户",[20,667,668],{},"已平仓","余额有史以来从峰值到谷底的最深跌幅，持仓的未实现盈亏不算在内。亏损只有兑现了才计入，所以在",[81,671,673],{"href":672},"\u002Fglossary\u002Fwhat-is-a-track-record","业绩记录","的两个回撤数字里，它是更宽容的那一个。ShowMyTrades 把它标作",[20,676,92],{},"，并且总是和基于净值的",[20,679,84],{},"并排显示。",[14,682,16],{"id":16},[10,684,685],{},"两者都是历史最差值，参照的是同一条按余额画出的高水位线。区别在于拿什么去和它比，以及多久比一次：",[25,687,690],{"className":688,"code":689,"language":30},[28],"峰值 = 已平仓余额的高水位线，随每一笔资金流动按比例调整\n\n余额回撤 %  = 最差的 (余额 − 峰值) \u002F 峰值 × 100　　              每笔交易时\n回撤　　 %  = 最差的 (min(余额, 净值) − 峰值) \u002F 峰值 × 100       每天一次\n　　　　      其中 净值 = 余额 + 持仓的浮动盈亏\n",[32,691,689],{"__ignoreMap":34},[10,693,694],{},"余额按账户自身的历史逐笔推算，每笔已平仓交易都扣掉佣金和库存费后计入。峰值随交易盈利往上走，自己永远不会回落。两者记录的都是历史最差值，不是当前状况，所以只会变差，不会好转。",[10,696,697],{},"资金进出有专门的处理：入金或出金让余额按什么比例变化，峰值就按同样的比例调整，所以存钱取钱既不会制造回撤，也不会修复回撤。两者的下限都是 −100%，余额被强制平仓（stop-out）打到零以下时，显示为全部亏光。",[10,699,700],{},"既然两者共用一个峰值，这两个数字就可以合起来当作一次测量来看：",[46,702,703,709,715,721],{},[49,704,705,708],{},[20,706,707],{},"两者接近","。仓位在出问题的位置附近就平掉了，没扛着多少亏损。",[49,710,711,714],{},[20,712,713],{},"净值回撤大得多","。亏损单一直拿着不平。账户的亏损报得晚，余额数字反映的是账面记录，而不是风险。",[49,716,717,720],{},[20,718,719],{},"两者完全相同","。要么浮亏从来没有和高水位时刻撞在一起，要么那段时间根本没有浮动盈亏的历史数据。",[49,722,723,726],{},[20,724,725],{},"余额回撤反而更大","。通常是采样方式造成的：余额每笔交易都检查，净值只看每日收盘值，所以一天之内出现又填平的坑，前者看得到，后者看不到。",[14,728,76],{"id":76},[10,730,731],{},"两者之间的差距，是识别“从不平亏损单”的系统最清楚的一个信号。网格、马丁格尔和摊平加仓策略，天生就会画出平滑的余额曲线：一笔永远不平的仓位，永远碰不到余额。它们的净值说的是另一回事，而最终让账户完蛋的，正是净值。",[10,733,734],{},[735,736],"img",{"alt":737,"src":738},"平滑的余额线旁边，一条净值线被亏损持有的仓位拉到其下方","\u002Fimages\u002Farticles\u002Fequity-vs-balance.svg",[10,740,741],{},"两条线之间的差距，就是余额还没有记上账的亏损。",[10,743,744,745,749,750,754],{},"这不是见仁见智的解读。保证金按净值计算，所以",[81,746,748],{"href":747},"\u002Fglossary\u002Fwhat-is-margin-level","保证金比例","和任何",[81,751,753],{"href":752},"\u002Fglossary\u002Fwhat-is-a-margin-call","追加保证金通知","都跟着净值线走，完全不看余额线。一份记录收益可观、余额回撤只有 3%，旁边的净值回撤却有 45%，这不是低风险系统：它曾经亏掉 45%，只是一直没认。两个数字贴得近，这本身就是证据：亏损一出现就了结了。",[14,756,107],{"id":107},[10,758,759,760,763,764,766,767,770,771,774,775,778,779,782],{},"以下数据描述的是在 ShowMyTrades 上公开的账户，并不代表所有交易者。在",[20,761,762],{},"有交易历史的已发布账户","中（2026 年 8 月），最深",[81,765,84],{"href":83},"的中位数为 ",[20,768,769],{},"9.7%","。两头分得很开：",[20,772,773],{},"38.5%"," 的账户从未从峰值回落超过 5%，而 ",[20,776,777],{},"38.2%"," 曾回撤超过 20%，",[20,780,781],{},"17.6%"," 超过 50%。",[10,784,785,786,789,790,793,794,796],{},"正是在第一组里，这一对数字最能派上用场：回撤低于 5% 的记录，要么是风控严格，要么是浮亏还没兑现，而这两种情况在余额图上看起来一模一样。这批账户大多是自动化交易：中位账户 ",[20,787,788],{},"99%"," 的交易由",[81,791,792],{"href":145},"自动化系统","执行，",[20,795,137],{}," 的账户自动化程度超过 90%。机器人扛着亏损单，既不会累，也不会慌。",[14,798,151],{"id":150},[10,800,801,803,804,807,808,810,811,815],{},[20,802,92],{},"位于",[20,805,806],{},"账户统计","面板中",[20,809,84],{},"的正下方，每个已发布账户的页面上都有，",[81,812,814],{"href":813},"\u002Fguides\u002Ftrading-widgets-and-embeds","完整控制台小组件","里也有。这两行就是面板的整个风险区块，用一条分隔线隔开，上面没有标题。数值由我们的统计服务根据账户已同步的历史计算得出，没有任何人手动填写，显示时不带负号。",[10,817,818,819,822,823,825,826,829],{},"页面上另有两处能把情况补全。",[20,820,821],{},"净值","一行显示净值占余额的百分比：低于 100%，说明此刻持仓处于浮亏，而这块缺口正是余额回撤不计入的部分。图表模块里的",[20,824,84],{},"视图用柱状图画出账户整个生命周期的日回撤，下跌",[20,827,828],{},"持续了多久","一眼就能看到，不用去猜。",[14,831,181],{"id":181},[46,833,834,840,846,852],{},[49,835,186,836,839],{},[20,837,838],{},"数字小的那个才是真实风险。","”数字小的通常是余额回撤，而它天生就是偏乐观的那一个。",[49,841,186,842,845],{},[20,843,844],{},"它没算交易成本。","”算了。每笔已平仓交易计入余额时，佣金和库存费都已经扣掉。",[49,847,186,848,851],{},[20,849,850],{},"出金会显示成回撤。","”不会。资金变动会让高水位线和余额按同样的比例调整。入金也同样修复不了回撤。",[49,853,186,854,857],{},[20,855,856],{},"两个数字看到的细节一样多。","”净值每天只采样一次，所以在一个交易日内跌下去又完全涨回来，它不会留下任何痕迹。",[10,859,860,861,131],{},"完整的计算方法，包括为什么回撤深度和持续时间是两个不同的问题，请参阅",[81,862,864],{"href":863},"\u002Fguides\u002Fmaximum-drawdown-explained","最大回撤详解",{"title":34,"searchDepth":217,"depth":217,"links":866},[867,868,869,870,871],{"id":16,"depth":217,"text":16},{"id":76,"depth":217,"text":76},{"id":107,"depth":217,"text":107},{"id":150,"depth":217,"text":151},{"id":181,"depth":217,"text":181},"风险","余额回撤（DD on Balance）只看已平仓结果，衡量余额从峰值到谷底的最深跌幅。本文讲解如何对照净值回撤来看它，以及公开账户的数据。",{},"\u002Fzh\u002Fglossary\u002Fwhat-is-drawdown-on-balance",[877,878,879,880],"what-is-drawdown","what-is-maximum-drawdown","what-is-margin-level","what-is-absolute-gain",{"title":660,"description":873},"zh\u002F5.glossary\u002Fwhat-is-drawdown-on-balance","余额回撤 (DD on Balance)","AEfaiFChn3YA1Ovbc3Qe6AeT3M2CoD7NND9m3xVT6kU",{"id":886,"title":887,"body":888,"category":224,"date":225,"description":1111,"draft":227,"extension":228,"meta":1112,"navigation":230,"path":1113,"related":1114,"seo":1116,"stem":1117,"term":1118,"updated":240,"__hash__":1119},"glossary\u002Fzh\u002F5.glossary\u002Fwhat-is-expectancy.md","什么是期望值？公式、示例与真实数据",{"type":7,"value":889,"toc":1104},[890,893,895,901,908,911,917,920,922,928,936,944,946,958,974,993,995,1007,1049,1067,1069,1098],[10,891,892],{},"期望值就是一笔交易的平均结果，可以用货币或点数表示：账户每开一笔仓位，平均能赚多少。它由胜率、平均盈利、亏损率、平均亏损四个数字算出，回答的是“这套系统每做一笔到底赚不赚钱”，而且结果可以直接乘以交易笔数。正期望值重复得足够多，就是一门生意；负期望值则是在倒数计时。",[14,894,16],{"id":16},[25,896,899],{"className":897,"code":898,"language":30},[28],"期望值 = (胜率 × 平均盈利) − (亏损率 × 平均亏损)\n\n亏损率 = 1 − 胜率\n",[32,900,898],{"__ignoreMap":34},[10,902,903,904,907],{},"假设一个系统胜率 40%，每笔盈利单赚 $300，每笔亏损单亏 $100：(0.40 × $300) − (0.60 × $100) = $120 − $60 = ",[20,905,906],{},"每笔交易 +$60","。一年做四百笔，预期结果就是 $24,000，这还没考虑仓位大小的变化，也没算上任何意外。",[10,909,910],{},"把同一个公式改用 R 倍数来写（R 即一笔平均亏损），账户货币的影响就去掉了，不同系统也能放在一起比较：",[25,912,915],{"className":913,"code":914,"language":30},[28],"期望值 (R) = (胜率 × 盈亏比) − (1 − 胜率)\n\n盈亏比 = 平均盈利 \u002F 平均亏损\n",[32,916,914],{"__ignoreMap":34},[10,918,919],{},"算出来的数字有没有意义，取决于两点。第一，输入里必须已经扣除成本：点差、佣金和库存费每笔都要付，盈利单因此变小，亏损单因此变大。第二，平均值背后要有足够多的交易，不能让某一笔特别大的结果单独左右它。",[14,921,76],{"id":76},[10,923,924,925,927],{},"统计数字要变成交易计划，靠的就是期望值。",[81,926,305],{"href":304},"告诉您赚的钱和亏的钱之比；期望值告诉您再多做一笔交易值多少钱，把它乘以交易频率，就能得出一个月的预期结果。",[10,929,930,931,935],{},"胜率掩盖的取舍，也会在期望值里暴露出来。胜率只是优势的一半，另一半是",[81,932,934],{"href":933},"\u002Fglossary\u002Fwhat-is-risk-reward-ratio","风险回报比","，期望值正是两者的交汇点。动其中任何一个，结果都会跟着变；两个都不动，故事讲得再好，结果也不会变。",[10,937,938,939,943],{},"期望值还划出了成本的底线。如果一个策略每笔的期望是 +$8，而一个来回的",[81,940,942],{"href":941},"\u002Fglossary\u002Fwhat-is-a-spread","点差","加佣金就要 $7，剩下的那点东西已经算不上策略了。",[14,945,107],{"id":107},[10,947,948,949,114,952,114,955,131],{},"以下数字描述的是 ShowMyTrades 上已发布的账户，不代表全体交易者。在有交易历史的已发布账户中（2026 年 8 月）取中间水平：",[20,950,951],{},"胜率中位数 68.8%",[20,953,954],{},"盈利因子中位数 1.28",[20,956,957],{},"已平仓交易数中位数 171 笔",[10,959,960,961,964,965,968,969,131],{},"有了这几个数字，这个“中间账户”的期望值就能算出来。把盈利因子公式反推回去，平均盈利约为平均亏损的 ",[20,962,963],{},"0.58 倍","；剩下的用 R 形式一算便知：(0.688 × 0.58) − 0.312 = ",[20,966,967],{},"+0.087R","。也就是说，每做一笔交易，只赚到一笔平均亏损的百分之九。171 笔交易下来，全部利润加起来大约只相当于十五笔平均亏损，所以尽管近七成交易以盈利平仓，",[20,970,971,973],{},[81,972,553],{"href":552},"的中位数也只有 +3.2%",[10,975,976,977,980,981,984,985,988,989,992],{},"优势薄到这个程度，成本就不是可以忽略的零头了。已发布账户累计支付了 ",[20,978,979],{},"$4,782,670 佣金","和 ",[20,982,983],{},"$862,547 库存费","；在产生过库存费的账户中，",[20,986,987],{},"86.3%"," 的净库存费为负。持仓时间中位数只有 ",[20,990,991],{},"2.4 小时","，在这种节奏下，0.087R 的优势要想保住，一个来回的成本相对平均亏损就必须足够小。",[14,994,151],{"id":150},[10,996,997,998,1000,1001,1003,1004,1006],{},"每个已发布账户页面的",[20,999,157],{},"模块里都有",[20,1002,542],{},"这一行，位于",[20,1005,165],{},"分组。它在同一行里给出两种单位（先点数，后账户货币）：点数便于在不同品种之间比较，货币金额便于在不同仓位大小之间比较，单看哪一个都不完整。",[10,1008,1009,1010,1012,1013,1015,1016,1015,1019,1022,1023,1025,1026,1032,1033,1022,1036,1039,1040,1015,1043,1022,1045,1048],{},"它的四个输入在同一模块的",[20,1011,615],{},"分组里：",[20,1014,118],{},"、",[20,1017,1018],{},"平均盈利",[20,1020,1021],{},"平均亏损","和",[20,1024,619],{},"，旁边还有",[81,1027,1029],{"href":1028},"\u002Fglossary\u002Fwhat-is-average-trade-length",[20,1030,1031],{},"平均持仓时间","，能看出这点薄薄的单笔优势是几小时赚到的，还是几周才赚到的。已经扣除的成本显示在",[20,1034,1035],{},"佣金合计",[20,1037,1038],{},"已付库存费合计","两行，每笔交易的明细则在交易表格的",[20,1041,1042],{},"盈利（毛额）",[20,1044,489],{},[20,1046,1047],{},"佣金","列。",[10,1050,1051,1052,1056,1057,1015,1060,1022,1063,1066],{},"想检验一个期望值到底是一种优势，还是几种优势平均出来的，可以打开",[81,1053,1054],{"href":392},[20,1055,395],{},"侧边面板：按品种、魔术号、方向、日期范围、星期或小时筛选出子集，整个区块就会重新计算。高级统计还提供",[20,1058,1059],{},"星期",[20,1061,1062],{},"小时",[20,1064,1065],{},"持续时间","视图，不用筛选就能把同一结果拆开来看。",[14,1068,181],{"id":181},[46,1070,1071,1080,1086,1092],{},[49,1072,186,1073,1076,1077,1079],{},[20,1074,1075],{},"正期望值就意味着我会赚钱。","”它只说明平均每笔交易是盈利的。仓位定错了，正期望值的系统照样会亏光：能不能活下来，是",[81,1078,529],{"href":528},"的问题，不是期望值的问题。",[49,1081,186,1082,1085],{},[20,1083,1084],{},"期望值就是我的平均每笔交易。","”前提是平均值按净额计算。按毛利润算就漏掉了佣金和库存费，而薄优势恰恰是在这里消失的。",[49,1087,186,1088,1091],{},[20,1089,1090],{},"高胜率就意味着正期望值。","”单看胜率说明不了任何问题。盈亏比为 0.58 时，60% 的胜率是个亏钱的系统。",[49,1093,186,1094,1097],{},[20,1095,1096],{},"这个数字很稳定。","”只有几百笔交易时并不稳定。一笔特别大的盈利单就能把整个记录的期望值抬上去，把它去掉再算一遍，是个公平的检验。",[10,1099,1100,1101,131],{},"哪些指标值得长期追踪、该按什么顺序看，请参阅",[81,1102,1103],{"href":638},"交易表现追踪完全指南",{"title":34,"searchDepth":217,"depth":217,"links":1105},[1106,1107,1108,1109,1110],{"id":16,"depth":217,"text":16},{"id":76,"depth":217,"text":76},{"id":107,"depth":217,"text":107},{"id":150,"depth":217,"text":151},{"id":181,"depth":217,"text":181},"期望值是单笔交易的平均盈亏：(胜率 × 平均盈利) − (亏损率 × 平均亏损)。本文给出公式、完整算例和公开账户的真实数据。",{},"\u002Fzh\u002Fglossary\u002Fwhat-is-expectancy",[457,1115,652,877],"what-is-risk-reward-ratio",{"title":887,"description":1111},"zh\u002F5.glossary\u002Fwhat-is-expectancy","期望值 (Expectancy)","4gLLgYBgHrrvWKTKFixk2tUfM9kXODaMx4LcyeirnhQ",[1121,1123],{"path":1122},"\u002Fglossary\u002Fwhat-is-a-z-score-in-trading",{"path":231},1793898057189]