[{"data":1,"prerenderedAt":1223},["ShallowReactive",2],{"\u002Fzh\u002Fglossary\u002Fwhat-is-the-sharpe-ratio":3,"\u002Fzh\u002Fglossary\u002Fwhat-is-the-sharpe-ratio-related":267,"\u002Fzh\u002Fglossary\u002Fwhat-is-the-sharpe-ratio-langs":1220},{"id":4,"title":5,"body":6,"category":249,"date":250,"description":251,"draft":252,"extension":253,"meta":254,"navigation":255,"path":256,"related":257,"seo":262,"stem":263,"term":264,"updated":265,"__hash__":266},"glossary\u002Fzh\u002F5.glossary\u002Fwhat-is-the-sharpe-ratio.md","什么是夏普比率？公式、局限与真实数据",{"type":7,"value":8,"toc":241},"minimark",[9,13,17,20,31,34,40,49,52,55,58,66,74,77,96,112,115,119,163,166,190,193,233],[10,11,12],"p",{},"夏普比率是平均收益除以这些收益的标准差，也就是账户每承受一单位波动，换来多少收益。两个账户一年下来增长一样，夏普比率却可能差很多：比率高的那个，一路走得更平稳。",[14,15,16],"h2",{"id":16},"工作原理",[10,18,19],{},"教科书上的定义用的是周期收益，并减去无风险利率：",[21,22,27],"pre",{"className":23,"code":25,"language":26},[24],"language-text","夏普比率 = (平均收益 − 无风险利率) \u002F 收益的标准差\n","text",[28,29,25],"code",{"__ignoreMap":30},"",[10,32,33],{},"ShowMyTrades 按已平仓交易计算，而不是按日历周期，也不减去无风险利率：",[21,35,38],{"className":36,"code":37,"language":26},[24],"夏普比率 = 每笔交易结果的均值 \u002F 每笔交易结果的标准差\n",[28,39,37],{"__ignoreMap":30},[10,41,42,43,48],{},"每笔交易按“盈利 + 佣金 + 库存费”计入序列，所以输入已经扣除了成本。只统计已平仓的买入和卖出交易：入金、出金和余额操作在生成序列之前就已剔除。分母是同一序列的总体标准差（距离平方和除以交易笔数，而不是笔数减一）。所以在我们的统计表里，",[44,45,47],"a",{"href":46},"\u002Fglossary\u002Fwhat-is-standard-deviation-in-trading","标准差","就排在这个比率正上方：它就是那个除数。已平仓交易不足两笔的账户显示 0.00。",[10,50,51],{},"按笔而不是按月计算，意味着这个数字没有年化，常见的经验标准也就用不上。“高于 1.0 算好，高于 2.0 算优秀”，说的是基于周期收益的年化夏普比率。在逐笔序列上，1.00 意味着平均每笔交易值整整一个交易结果标准差：交易笔数足以说明问题的账户，几乎没有能达到的。",[14,53,54],{"id":54},"为什么重要",[10,56,57],{},"只看增长，只知道账户最后停在哪里；夏普比率还能说明它是怎么走过来的。优势小但可靠、反复执行的策略，比率不高，但为正；靠两笔运气交易撑起来的策略，均值很大，离散程度却更大，比率随之一落千丈。",[10,59,60,61,65],{},"有些图表只是看起来平滑，夏普比率能诚实地戳穿它，而它用的正是计算",[44,62,64],{"href":63},"\u002Fglossary\u002Fwhat-is-a-profit-factor","盈利因子","的同一批已平仓交易。",[10,67,68,69,73],{},"它做不到的，是区分好的波动和坏的波动。一笔向上大幅偏离的交易，和一笔向下大幅偏离的交易，把分母抬高的幅度完全一样。",[44,70,72],{"href":71},"\u002Fglossary\u002Fwhat-is-the-sortino-ratio","索提诺比率","正是为了回应这一质疑而诞生的。",[14,75,76],{"id":76},"数据显示了什么",[10,78,79,80,84,85,84,88,95],{},"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），",[81,82,83],"strong",{},"夏普比率中位数为 0.05","，",[81,86,87],{},"已平仓交易数中位数为 171 笔",[81,89,90,94],{},[44,91,93],{"href":92},"\u002Fglossary\u002Fwhat-is-average-trade-length","持仓时间","中位数为 2.4 小时","。",[10,97,98,99,103,104,107,108,111],{},"用我们自己的公式来读，0.05 的含义很具体：在中位数账户上，平均每笔交易约值交易结果标准差的二十分之一，差不多是二十份噪音对一份信号。这并不说明这些账户都很失败：",[44,100,102],{"href":101},"\u002Fglossary\u002Fwhat-is-time-weighted-return","时间加权收益率","中位数为 ",[81,105,106],{},"+3.2%","，长期为正的占 ",[81,109,110],{},"63.0%","。真实的优势落到单笔交易上，就是这个样子。",[10,113,114],{},"这也说明了短样本为什么什么都证明不了。在这样的信噪比下，一个毫无优势的账户连赢二十笔也很平常，而一个确有优势的账户亏上一周，也说明不了任何问题。",[14,116,118],{"id":117},"在-showmytrades-上的位置","在 ShowMyTrades 上的位置",[10,120,121,122,125,126,129,130,133,134,136,137,140,141,143,144,143,147,143,150,153,154,157,158,162],{},"账户页面的",[81,123,124],{},"高级统计","模块里，",[81,127,128],{},"表现指标","一栏中有一行，标签正是",[81,131,132],{},"夏普比率","，并附有提示说明。它的正上方是它自己的分母",[81,135,47],{},"，正下方是",[81,138,139],{},"夏普比率（年化）","（以净值的日收益率计算，再乘以 √252 年化；账户历史不足 30 个工作日时显示为 n\u002Fa），旁边是",[81,142,64],{},"、",[81,145,146],{},"Z 值（概率）",[81,148,149],{},"期望值",[81,151,152],{},"AHPR"," 和 ",[81,155,156],{},"GHPR","。它由我们的服务器根据存储的已平仓交易计算，保留两位小数，1.00 及以上显示绿色，0 到 1 显示黄色，低于零显示红色。在已发布的账户页面上，这个模块默认显示，不过账户主人可以隐藏它；",[44,159,161],{"href":160},"\u002Fguides\u002Ftrading-widgets-and-embeds","可嵌入的小组件","里没有这个模块。",[10,164,165],{},"关于我们自己的页面，要提醒一点：逐笔那一行标签的提示说明里，引用的仍是常规区间，即高于 1.0 算好、高于 2.0 算优秀。这些区间适用于下面的年化那一行，而不是旁边显示的逐笔数值。",[10,167,168,169,143,172,143,175,178,179,182,183,189],{},"同一模块提供",[81,170,171],{},"表格",[81,173,174],{},"星期",[81,176,177],{},"小时","和",[81,180,181],{},"持续时间","视图；",[44,184,186],{"href":185},"\u002Fguides\u002Fcustom-analysis-and-filter-presets",[81,187,188],{},"自定义分析","会在筛选出的子集上重新计算整个模块，可按日期范围、品种、魔术号、交易类型、交易时段、手数等筛选。把账户表现最好的品种去掉再算一次夏普比率，就能看出这个比率描述的是一套系统，还是一段插曲。",[14,191,192],{"id":192},"常见误解",[194,195,196,210,221,227],"ul",{},[197,198,199,200,203,204,206,207,209],"li",{},"“",[81,201,202],{},"高于 1.0 就是好。","”这话适用于基于周期收益的年化夏普比率，也就是",[81,205,139],{}," 那一行。普通的",[81,208,132],{},"是逐笔计算的，这个门槛对它不适用。",[197,211,199,212,215,216,220],{},[81,213,214],{},"夏普比率衡量的是风险。","”它衡量的是已平仓交易结果的离散程度，至于账户曾经跌得多深，它一无所知：那是",[44,217,219],{"href":218},"\u002Fglossary\u002Fwhat-is-drawdown","回撤","，另有单独的数字。",[197,222,199,223,226],{},[81,224,225],{},"夏普比率高，账户就安全。","”网格和摊平系统的已平仓结果之所以紧凑均匀，恰恰是因为它们不肯把亏损单平掉。浮动亏损永远进不了这个序列。",[197,228,199,229,232],{},[81,230,231],{},"我的夏普比率可以和任何账户比较。","”只有交易频率和持仓周期相当时才行：每月平仓 200 笔的剥头皮交易者和每月平仓 8 笔的波段交易者，用的根本不是同一把尺子。",[10,234,235,236,240],{},"夏普比率和账户页面上其他数字怎么对照着看，请参阅",[44,237,239],{"href":238},"\u002Fguides\u002Freading-a-trading-account-dashboard","如何读懂交易账户控制台","指南。",{"title":30,"searchDepth":242,"depth":242,"links":243},2,[244,245,246,247,248],{"id":16,"depth":242,"text":16},{"id":54,"depth":242,"text":54},{"id":76,"depth":242,"text":76},{"id":117,"depth":242,"text":118},{"id":192,"depth":242,"text":192},"指标","2026-08-19T00:00:00.000Z","夏普比率是平均收益除以这些收益的波动率。本文介绍公式、我们如何逐笔计算，以及为什么我们的中位数是 0.05。",false,"md",{},true,"\u002Fzh\u002Fglossary\u002Fwhat-is-the-sharpe-ratio",[258,259,260,261],"what-is-standard-deviation-in-trading","what-is-the-sortino-ratio","what-is-recovery-factor","what-is-a-profit-factor",{"title":5,"description":251},"zh\u002F5.glossary\u002Fwhat-is-the-sharpe-ratio","夏普比率 (Sharpe Ratio)","2026-09-17T00:00:00.000Z","rF1ahLb2_s2YlvPIZt9lNpcU3FX-JK93QxZ6RXkVz7s",[268,548,775,1016],{"id":269,"title":270,"body":271,"category":249,"date":250,"description":536,"draft":252,"extension":253,"meta":537,"navigation":255,"path":538,"related":539,"seo":544,"stem":545,"term":546,"updated":250,"__hash__":547},"glossary\u002Fzh\u002F5.glossary\u002Fwhat-is-a-profit-factor.md","什么是盈利因子？公式、合理数值与真实数据",{"type":7,"value":272,"toc":529},[273,276,278,284,287,290,296,299,366,368,371,379,386,388,402,409,412,419,430,432,458,484,486,522],[10,274,275],{},"盈利因子就是总盈利除以总亏损：所有盈利交易赚的钱加起来，除以所有亏损交易亏的钱。盈利因子 1.50，意思是账户每亏 $1.00，就赚回 $1.50。低于 1.00，不管胜率多高，账户从结构上就注定亏钱。",[14,277,16],{"id":16},[21,279,282],{"className":280,"code":281,"language":26},[24],"盈利因子 = 总盈利 \u002F |总亏损|\n",[28,283,281],{"__ignoreMap":30},[10,285,286],{},"注意这个公式没有考虑什么：账户交易了多久、用了多少资金、中途跌得有多深。",[10,288,289],{},"它也可以用胜率、平均盈利和平均亏损来写，这样写就能拿来诊断问题：",[21,291,294],{"className":292,"code":293,"language":26},[24],"盈利因子 = (胜率 × 平均盈利) \u002F ((1 − 胜率) × 平均亏损)\n",[28,295,293],{"__ignoreMap":30},[10,297,298],{},"两个截然不同的系统，可能算出同一个数值。胜率 30%、盈利单平均是亏损单的四倍，结果是 1.71。胜率 80%、盈利单只有亏损单的一半，结果是 2.00。两者都能赚钱，但出问题的方式不同，做起来的感受也完全两样。",[300,301,302,314],"table",{},[303,304,305],"thead",{},[306,307,308,311],"tr",{},[309,310,64],"th",{},[309,312,313],{},"解读",[315,316,317,326,334,342,350,358],"tbody",{},[306,318,319,323],{},[320,321,322],"td",{},"低于 1.00",[320,324,325],{},"亏钱，而且是结构性的，不是差一点",[306,327,328,331],{},[320,329,330],{},"1.00 – 1.10",[320,332,333],{},"还在噪声范围内，一周行情不好就没了",[306,335,336,339],{},[320,337,338],{},"1.10 – 1.30",[320,340,341],{},"优势很薄但真实存在，前提是样本够大",[306,343,344,347],{},[320,345,346],{},"1.30 – 2.00",[320,348,349],{},"策略有效",[306,351,352,355],{},[320,353,354],{},"2.00 – 3.00",[320,356,357],{},"很强。请核对交易笔数，以及成本是否算全",[306,359,360,363],{},[320,361,362],{},"高于 3.00",[320,364,365],{},"在长期历史里很少见：通常是样本太短、靠几笔超大盈利撑着，或者还有亏损单没平仓",[14,367,54],{"id":54},[10,369,370],{},"要回答“这个系统到底赚不赚钱”，盈利因子是最干净的单一指标。它也比胜率难包装得多，因为每一笔亏损都按全额计入分母。",[10,372,373,374,378],{},"但它不是风险指标。盈利因子不管顺序：一个账户一个月赚够全部利润、另外十一个月一直失血，和一个每周稳步上涨的账户，数值可以一模一样。所以要和",[44,375,377],{"href":376},"\u002Fglossary\u002Fwhat-is-maximum-drawdown","最大回撤","一起看，否则只看到了终点，没看到过程。",[10,380,381,385],{},[44,382,384],{"href":383},"\u002Fglossary\u002Fwhat-is-backtesting","回测","给出的盈利因子同样要小心：优化器追求的就是把这个比值做到最大，所以回测里的高数值，往往反映的是拟合做得多好，而不是策略本身多好。",[14,387,76],{"id":76},[10,389,390,391,84,394,401],{},"以下数字只描述在 ShowMyTrades 上发布的账户，不代表全体交易者。截至 2026 年 8 月，有交易历史的已发布账户中，",[81,392,393],{},"盈利因子中位数为 1.28",[81,395,396,400],{},[44,397,399],{"href":398},"\u002Fguides\u002Fwin-rate-is-not-an-edge","胜率中位数","为 68.8%","。两个数字都是扣除成本后的：每笔交易先计入库存费和佣金，再判定是盈利单还是亏损单，所以 1.28 是净值，不是毛值。",[10,403,404],{},[405,406],"img",{"alt":407,"src":408},"ShowMyTrades 公开账户的中位数：68.8% 的胜率与 1.28 的盈利因子并列","\u002Fimages\u002Farticles\u002Fwin-rate-vs-payoff.svg",[10,410,411],{},"同一组数据的两个中位数：胜率看着很高，背后的优势却很薄。",[10,413,414,415,418],{},"要看的正是这两个数放在一起。中位数账户里，将近七成的交易以盈利平仓，可整体优势也只是每亏 $1.00 赚 $1.28。拿一个两项都正好处在中位数的账户，把公式倒推回去：它的平均盈利只有平均亏损的约 ",[81,416,417],{},"0.58 倍","。赢得多和赚得多是两回事，前者要容易制造得多：亏损单扛着不平，盈利单早早落袋，胜率就往上走，盈利因子却往下掉。",[10,420,421,422,425,426,429],{},"同一组账户的夏普比率中位数是 ",[81,423,424],{},"0.05","，中位数账户有 ",[81,427,428],{},"171 笔已平仓交易","。样本这么小，盈利因子高于 3.00 并不能证明系统更强，只能说明样本还太少，最差的那笔交易还没出现。",[14,431,118],{"id":117},[10,433,434,435,437,438,440,441,143,444,143,447,143,450,143,452,178,455,457],{},"每个已发布账户页面的",[81,436,124],{},"模块里都有",[81,439,64],{},"，旁边是能解释它的几个数字：",[81,442,443],{},"胜率",[81,445,446],{},"平均盈利",[81,448,449],{},"平均亏损",[81,451,149],{},[81,453,454],{},"总交易数",[81,456,132],{},"。这六个数字合起来，能说清单一比值说不清的事。",[10,459,460,461,471,472,178,475,478,479,483],{},"有两个产品细节值得注意。第一，已经算进比值的成本会另外逐项列出：交易表格有",[81,462,463,464,470],{},"盈利（毛额）",[81,465,466,467,178],{},"列，",[81,468,469],{},"库存费","佣金","各占一列；高级统计还汇总了",[81,473,474],{},"佣金合计",[81,476,477],{},"已付库存费合计","，可以看出净值数字里扣掉了多少成本。在所有已发布账户中，这两项合计分别是佣金 $4,782,670、库存费 $862,547。第二，",[44,480,481],{"href":185},[81,482,188],{},"可以先筛选（按日期范围、品种、魔术号、方向或手数），再对子集重算整个统计区块。用它可以检验：去掉表现最好的那个季度，或者去掉最赚钱的那个品种，盈利因子是否依然站得住。",[14,485,192],{"id":192},[194,487,488,499,510,516],{},[197,489,199,490,493,494,498],{},[81,491,492],{},"盈利因子高于 1，就说明我在赚钱。","”前提是成本已经算进比值。如果没有，一个毛值 1.05 的高频系统，",[44,495,497],{"href":496},"\u002Fglossary\u002Fwhat-is-cost-percentage","计入佣金和库存费之后","可能持平甚至亏损。",[197,500,199,501,504,505,509],{},[81,502,503],{},"盈利因子高，风险就低。","”它反映不了回撤、仓位大小和盈亏顺序。马丁格尔式逐级加仓的",[44,506,508],{"href":507},"\u002Fglossary\u002Fwhat-is-position-sizing","仓位管理","方案，可以一直显示 4.00，直到哪天突然失灵。",[197,511,199,512,515],{},[81,513,514],{},"不同时间周期之间可以直接比。","”不能，它没有年化。剥头皮交易者 20,000 笔交易得出的 1.20，和波段交易者 60 笔交易得出的 1.20，说的不是一回事。",[197,517,199,518,521],{},[81,519,520],{},"未平仓的交易不影响它。","”会影响，恰恰因为它们不在里面。一直不平的浮动亏损完全不计入总亏损，在平仓之前会一直把比值抬高。",[10,523,524,525,95],{},"该长期追踪哪些指标、按什么顺序看，请参阅",[44,526,528],{"href":527},"\u002Fguides\u002Ftracking-trading-performance","交易表现追踪完全指南",{"title":30,"searchDepth":242,"depth":242,"links":530},[531,532,533,534,535],{"id":16,"depth":242,"text":16},{"id":54,"depth":242,"text":54},{"id":76,"depth":242,"text":76},{"id":117,"depth":242,"text":118},{"id":192,"depth":242,"text":192},"盈利因子是总盈利除以总亏损，低于 1.0 的账户注定亏损。本文给出各数值区间的解读，以及数千个账户的中位数 1.28。",{},"\u002Fzh\u002Fglossary\u002Fwhat-is-a-profit-factor",[540,541,542,543],"what-is-drawdown","what-is-maximum-drawdown","what-is-position-sizing","what-is-backtesting",{"title":270,"description":536},"zh\u002F5.glossary\u002Fwhat-is-a-profit-factor","盈利因子 (Profit Factor)","OLz2UdKZBQ-s3NM0RQISYNe6Jrhtmy0tM84z39S0Thk",{"id":549,"title":550,"body":551,"category":249,"date":250,"description":766,"draft":252,"extension":253,"meta":767,"navigation":255,"path":768,"related":769,"seo":771,"stem":772,"term":773,"updated":265,"__hash__":774},"glossary\u002Fzh\u002F5.glossary\u002Fwhat-is-recovery-factor.md","什么是恢复因子？公式、数值区间与真实数据",{"type":7,"value":552,"toc":759},[553,556,558,564,567,573,576,621,629,631,639,649,651,661,668,678,680,686,706,713,723,725,753],[10,554,555],{},"恢复因子就是净利润除以最大回撤：账户让主人每承受一单位跌幅，换回了多少收益。数值为 2.0，说明账户赚到的是最差时亏掉的两倍；低于 1.0，说明它挖下的坑还没填平。",[14,557,16],{"id":16},[21,559,562],{"className":560,"code":561,"language":26},[24],"恢复因子 = 净利润 \u002F 最大回撤\n",[28,563,561],{"__ignoreMap":30},[10,565,566],{},"公式本身很简单，难的是单位。流行的写法有两种，结果并不一致：",[21,568,571],{"className":569,"code":570,"language":26},[24],"货币形式   = 净利润 ($) \u002F 最大回撤 ($)\n百分比形式 = 总收益 (%) \u002F 最大回撤 (%)\n",[28,572,570],{"__ignoreMap":30},[10,574,575],{},"分子和分母必须用同一种口径。货币分子配百分比分母，算出来的数字看着像比值，其实毫无意义。",[300,577,578,587],{},[303,579,580],{},[306,581,582,585],{},[309,583,584],{},"恢复因子",[309,586,313],{},[315,588,589,597,605,613],{},[306,590,591,594],{},[320,592,593],{},"低于 1.0",[320,595,596],{},"账户赚到的还不及最深那次跌幅",[306,598,599,602],{},[320,600,601],{},"1.0 – 2.0",[320,603,604],{},"有效，但回报和承受的风险差不多",[306,606,607,610],{},[320,608,609],{},"2.0 – 3.0",[320,611,612],{},"稳健，前提是历史够长，真正的最大回撤已经出现过",[306,614,615,618],{},[320,616,617],{},"高于 3.0",[320,619,620],{},"很强：相信之前，先看交易笔数和账户年龄",[10,622,623,624,628],{},"有一点比整张表都重要：恢复因子会随着时间自己变大。只要账户还在赚钱，分子就不断累积；分母却是历史最大值，只有情况变得更糟时才会变。",[44,625,627],{"href":626},"\u002Fguides\u002Fhow-much-history-a-track-record-needs","五年的账户和六个月的账户在这一点上没法比","：老账户只是有更长的时间，让分子在一个固定的分母上不断累积。",[14,630,54],{"id":54},[10,632,633,634,638],{},"利润和回撤通常分开报，分开来看，两者都很容易包装得好看。恢复因子偏要把它们放在一起。每一份",[44,635,637],{"href":636},"\u002Fglossary\u002Fwhat-is-a-track-record","业绩记录","都得回答一个问题：吃的这些苦，值不值？恢复因子就是这个问题的浓缩版。",[10,640,641,642,644,645,648],{},"它还补上了相邻指标的空白。",[44,643,64],{"href":63},"比较总盈利和总亏损，不管先后顺序，所以一个月赚完全部利润的账户，和每周稳步上涨的账户，得分一样。",[44,646,132],{"href":647},"\u002Fglossary\u002Fwhat-is-the-sharpe-ratio","衡量离散程度，却不管跌得多深。在常用指标里，只有恢复因子把最差的那一刻明明白白放进了分母。",[14,650,76],{"id":76},[10,652,79,653,84,658,95],{},[81,654,655,657],{},[44,656,102],{"href":101},"的中位数为 +3.2%",[81,659,660],{},"最深回撤的中位数为 9.7%",[10,662,663,664,667],{},"这两个中位数出自不同的账户，所以算不出恢复因子在总体中的中位数。但如果一个账户恰好两项都在中位数上，按百分比形式算，恢复因子约为 ",[81,665,666],{},"0.33","：大约要承受三单位的跌幅，才换来一单位的收益。按惯例，低于 3.0 就算平平无奇，两者不妨对照着看。",[10,669,670,671,674,675,677],{},"尾部也不能忽视。这些账户中有 ",[81,672,673],{},"38.2%"," 曾下跌超过 20%，长期为正的占 ",[81,676,110],{},"，也就是说三分之一以上的账户分子是负数。对它们来说，这个比值谈不上低，而是根本没有意义：拿亏损除以回撤，得出的数字哪个区间都解释不了。",[14,679,118],{"id":117},[10,681,682,685],{},[81,683,684],{},"恢复因子不是 ShowMyTrades 上的指标","。任何面板、图表或小组件都不显示它，无论用这个名字还是别的名字，我们的统计引擎也不计算它。不过，用账户页面公布的数字一步就能算出来，这些数字全部来自平台数据，而不是账户主人手动填写的。",[10,687,688,689,692,693,696,697,699,700,702,703,705],{},"百分比形式：取",[81,690,691],{},"账户统计","面板里的",[81,694,695],{},"增长","，除以同一面板中往下第四行的",[81,698,219],{},"。两者都是百分比，所以可以直接相除。注意它们各自的含义：",[81,701,695],{},"是时间加权收益率，已剔除入金和出金的影响，往亏损账户里追加资金也抬不高它；",[81,704,219],{},"是记录到的最差净值跌幅。",[10,707,708,709,712],{},"货币形式：同一面板里的",[81,710,711],{},"盈利","是扣除佣金和库存费后的已平仓净盈亏，但我们不以货币金额公布最大回撤，所以缺了做除数的那个数字。可以改用百分比形式，或者用已平仓交易列表重建净值走势，交易表格支持下载 CSV。",[10,714,715,716,718,719,722],{},"不管用哪种形式，都先打开面板旁边的",[81,717,219],{},"图表视图。它以柱状图逐日画出回撤，能看出最大回撤是",[81,720,721],{},"什么时候","出现的、账户在峰值下方待了多久，而这段持续时间正是比值表达不了的。",[14,724,192],{"id":192},[194,726,727,733,739,747],{},[197,728,199,729,732],{},[81,730,731],{},"总是越高越好。","”历史长短不同时就不成立。风险不变，这个比值也会随时间变大，所以只能在业绩记录长度相当的账户之间比较。",[197,734,199,735,738],{},[81,736,737],{},"它和盈利因子是一回事。","”盈利因子是总盈利除以总亏损；恢复因子是最终结果除以最差那一次从峰值到谷底的跌幅。",[197,740,199,741,744,745,95],{},[81,742,743],{},"它说明我恢复得有多快。","”它说明的是幅度，不是时间。为什么让交易者离场的往往是回撤的持续时间而不是深度，请参阅",[44,746,219],{"href":218},[197,748,199,749,752],{},[81,750,751],{},"用余额回撤做分母更干净。","”那是更好看的分母：亏损单永远不平的策略，余额回撤天生就小。请优先用净值数字，并注明用的是哪一个。",[10,754,755,756,240],{},"分母怎么计量、最大值为什么永远不会下降，请参阅",[44,757,377],{"href":758},"\u002Fguides\u002Fmaximum-drawdown-explained",{"title":30,"searchDepth":242,"depth":242,"links":760},[761,762,763,764,765],{"id":16,"depth":242,"text":16},{"id":54,"depth":242,"text":54},{"id":76,"depth":242,"text":76},{"id":117,"depth":242,"text":118},{"id":192,"depth":242,"text":192},"恢复因子是净利润除以最大回撤，即每承受一单位跌幅换来的收益。本文介绍公式、数值区间，以及计算所需的数字在哪里找。",{},"\u002Fzh\u002Fglossary\u002Fwhat-is-recovery-factor",[541,540,261,770],"what-is-the-sharpe-ratio",{"title":550,"description":766},"zh\u002F5.glossary\u002Fwhat-is-recovery-factor","恢复因子 (Recovery Factor)","T2aXIYYtEer-h9VriHvWPBlH-AZl3zT5JTJbKQEC33Y",{"id":776,"title":777,"body":778,"category":249,"date":250,"description":1006,"draft":252,"extension":253,"meta":1007,"navigation":255,"path":1008,"related":1009,"seo":1012,"stem":1013,"term":1014,"updated":265,"__hash__":1015},"glossary\u002Fzh\u002F5.glossary\u002Fwhat-is-standard-deviation-in-trading.md","交易中的标准差是什么？公式与数据",{"type":7,"value":779,"toc":999},[780,783,785,791,811,846,849,851,854,864,867,894,896,902,905,912,914,934,960,962,993],[10,781,782],{},"标准差衡量的是单笔交易的结果通常离账户的平均交易结果有多远。ShowMyTrades 以账户货币显示它，所以它回答的问题很直接：在这个账户上，一笔交易和平常的交易通常相差多少？数值小，说明结果比较集中；数值大，说明这段历史是由离群值堆出来的。",[14,784,16],{"id":16},[21,786,789],{"className":787,"code":788,"language":26},[24],"标准差 = √( Σ (rᵢ − 均值)² \u002F n )\n",[28,790,788],{"__ignoreMap":30},[10,792,793,794,797,798,802,803,806,807,810],{},"其中 ",[28,795,796],{},"rᵢ"," 是第 ",[799,800,801],"em",{},"i"," 笔已平仓交易的净结果，",[28,804,805],{},"均值"," 是这些结果的平均值，",[28,808,809],{},"n"," 是已平仓交易笔数。这个数字代表什么，取决于四个细节：",[194,812,813,821,827,840],{},[197,814,815,95,818,820],{},[81,816,817],{},"每笔交易都按扣除成本后的净额计入",[28,819,796],{}," 等于盈利 + 佣金 + 库存费，所以佣金或库存费高的交易，压缩了多少盈利，就同样放大了多少离散程度。",[197,822,823,826],{},[81,824,825],{},"只统计已平仓的买入和卖出交易","。入金和出金在生成序列之前就已剔除，所以往账户里注资不会变成一个巨大的离群值。",[197,828,829,832,833,835,836,839],{},[81,830,831],{},"它用的是总体标准差","，除以 ",[28,834,809],{}," 而不是 ",[28,837,838],{},"n − 1","。已平仓交易不足两笔时，显示为 0.00。",[197,841,842,845],{},[81,843,844],{},"它以货币计，不是百分比","。同样是 $50 的标准差，在 $2,000 的账户和 $200,000 的账户上，说明的是完全不同的交易方式。一定要对照账户余额来看。",[10,847,848],{},"由于每个距离都先平方再平均，一笔异常交易对结果的影响远超十几笔普通交易，所以它特别容易揪出那笔格格不入的交易。",[14,850,54],{"id":54},[10,852,853],{},"“稳定”这个词是真是假，标准差一测便知：谁都可以说自己的净值曲线很平稳，标准差则用一个货币数值告诉您，这些交易到底有多平稳。",[10,855,856,857,859,860,863],{},"它也是我们的",[44,858,132],{"href":647},"的分母：夏普比率就是用这个数字作单位来衡量的平均交易结果。它还是",[44,861,149],{"href":862},"\u002Fglossary\u002Fwhat-is-expectancy","的天然搭档：期望值给出每笔交易的平均结果，却看不出这个平均结果有多靠得住。",[10,865,866],{},"最有用的读法是比较，拿它和几行之外的两个平均值对照：",[194,868,869,875,888],{},[197,870,871,874],{},[81,872,873],{},"接近平均盈利","：结果很集中，账户大部分时候表现一致。",[197,876,877,880,881,884,885,95],{},[81,878,879],{},"是平均盈利的好几倍","：整段历史靠少数几笔交易撑着。相信总数之前，先看",[81,882,883],{},"最佳交易（盈亏）"," 和",[81,886,887],{},"最差交易（盈亏）",[197,889,890,893],{},[81,891,892],{},"远大于平均亏损","：至少有一笔亏损远远超出平常水平。",[14,895,76],{"id":76},[10,897,79,898,84,900,95],{},[81,899,83],{},[81,901,87],{},[10,903,904],{},"按我们自己的定义倒推，这组数字能把离散程度说得很准：夏普比率是平均交易结果除以标准差，中位数 0.05 意味着，中间水平的账户平均每笔交易只值交易标准差的二十分之一左右。也就是说，一笔交易和下一笔之间的正常波动，大约是从中赚到的优势的二十倍。",[10,906,907,911],{},[81,908,909,94],{},[44,910,93],{"href":92},"，所以看一整天的结果，交易者学不到任何东西：一天的结果完全落在正常波动范围之内。同样道理，新账户的标准差低，也证明不了风险受控：样本里可能只是还没出现这个策略会产生的那个离群值，而距离要平方，它一旦出现，动静就会很大。",[14,913,118],{"id":117},[10,915,121,916,125,918,129,920,922,923,925,926,928,929,933],{},[81,917,124],{},[81,919,128],{},[81,921,47],{},"，以账户货币显示，并附有提示说明。它紧挨在以它为分母的",[81,924,132],{},"上方、",[81,927,64],{},"下方，由我们的服务器根据存储的已平仓交易计算。在已发布的账户页面上，这个模块默认显示，不过",[44,930,932],{"href":931},"\u002Fguides\u002Fpublic-profile-and-custom-slugs","账户主人可以隐藏它","；可嵌入的小组件里没有这个模块。",[10,935,936,937,940,941,143,943,143,945,143,947,884,949,951,952,954,955,959],{},"把它和同一模块的",[81,938,939],{},"交易","一栏对照着看：",[81,942,446],{},[81,944,449],{},[81,946,883],{},[81,948,887],{},[81,950,454],{},"为这个数字提供了参照。模块里的",[81,953,181],{},"视图能把离散程度直接画出来：每笔已平仓交易是一个点，横轴是持仓小时数，纵轴是盈亏，离散程度就从一个数字变成了一个形状。",[44,956,957],{"href":185},[81,958,188],{},"会在筛选出的子集上重新计算整个模块，可按日期范围、品种、魔术号、交易类型、手数等筛选，由此看出离散程度来自策略本身，还是来自某一个交易品种。",[14,961,192],{"id":192},[194,963,964,970,981,987],{},[197,965,199,966,969],{},[81,967,968],{},"它是一个百分比。","”我们的标准差是每笔交易的金额。旁边没有余额作参照，在不同规模的账户之间比较毫无意义。",[197,971,199,972,975,976,980],{},[81,973,974],{},"它衡量的是市场波动率。","”它衡量的是您交易结果的",[44,977,979],{"href":978},"\u002Fglossary\u002Fwhat-is-volatility","波动率","。市场再平静，只要仓位越加越大，数值照样会很大。",[197,982,199,983,986],{},[81,984,985],{},"标准差低，风险就低。","”只有已平仓交易计入其中。扛着亏损单不平的系统，已平仓结果个个都小而整齐，真正的风险敞口却藏在浮动仓位里。",[197,988,199,989,992],{},[81,990,991],{},"30 笔交易下数值很低，可以放心。","”不到几百笔交易时，只能把它当作暂时的结果。",[10,994,995,996,240],{},"一组统计数据要有多少笔交易才有意义，请参阅",[44,997,998],{"href":626},"一份业绩记录需要多长的历史",{"title":30,"searchDepth":242,"depth":242,"links":1000},[1001,1002,1003,1004,1005],{"id":16,"depth":242,"text":16},{"id":54,"depth":242,"text":54},{"id":76,"depth":242,"text":76},{"id":117,"depth":242,"text":118},{"id":192,"depth":242,"text":192},"标准差衡量单笔交易的结果离您的平均交易结果有多远。本文介绍我们使用的公式（已扣除成本），以及如何在数千个账户上解读它。",{},"\u002Fzh\u002Fglossary\u002Fwhat-is-standard-deviation-in-trading",[770,1010,1011,259],"what-is-volatility","what-is-expectancy",{"title":777,"description":1006},"zh\u002F5.glossary\u002Fwhat-is-standard-deviation-in-trading","标准差 (Standard Deviation)","d23H7esNsSG175eWPyoZzoJTWWR3SEfy5D-ZKP-x9wU",{"id":1017,"title":1018,"body":1019,"category":249,"date":250,"description":1212,"draft":252,"extension":253,"meta":1213,"navigation":255,"path":1214,"related":1215,"seo":1216,"stem":1217,"term":1218,"updated":265,"__hash__":1219},"glossary\u002Fzh\u002F5.glossary\u002Fwhat-is-the-sortino-ratio.md","什么是索提诺比率？下行风险详解",{"type":7,"value":1020,"toc":1205},[1021,1024,1026,1032,1038,1044,1047,1061,1064,1066,1077,1080,1086,1088,1104,1107,1123,1125,1156,1172,1174,1200],[10,1022,1023],{},"索提诺比率可以看作分母更窄的夏普比率：收益不再除以全部结果的波动率，而是只除以低于某个选定目标的那部分结果的波动率。向上的波动不算风险。有人批评说，交易者因为大赚而被扣分，扣错了地方；索提诺比率回应的正是这一点。",[14,1025,16],{"id":16},[21,1027,1030],{"className":1028,"code":1029,"language":26},[24],"索提诺比率 = (平均收益 − 目标收益) \u002F 下行偏差\n",[28,1031,1029],{"__ignoreMap":30},[21,1033,1036],{"className":1034,"code":1035,"language":26},[24],"下行偏差 = √( Σ min(rᵢ − 目标收益, 0)² \u002F n )\n",[28,1037,1035],{"__ignoreMap":30},[10,1039,1040,1041,1043],{},"高于目标的观测值，对分母的贡献都是零；低于目标的，则贡献差额的平方，做法和",[44,1042,47],{"href":46},"完全一样。",[10,1045,1046],{},"这个公式里藏着两个选择，每一个都会改变结果：",[194,1048,1049,1055],{},[197,1050,1051,1054],{},[81,1052,1053],{},"目标","。可以是零、无风险利率，或者一个要求达到的收益率。以 0% 为目标和以每年 4% 为目标算出的索提诺比率，名字一样，其实是两种不同的统计量。",[197,1056,1057,1060],{},[81,1058,1059],{},"除数","。有的实现除以全部观测值的个数，有的只除以低于目标的观测值个数。后一种算出的比率总是更大、更好看。",[10,1062,1063],{},"这两个选择都没有统一标准，所以不注明目标和计算惯例的索提诺比率，根本没法拿来比较。这不是吹毛求疵，而是实际问题：两个平台对同一个账户报出不同的索提诺比率，而且都没算错，原因就在这里。",[14,1065,54],{"id":54},[10,1067,1068,1069,1073,1074,1076],{},"索提诺比率的出发点没错。投资者不会把上涨 5% 的月份当成风险，把它当成风险的指标，就会错估正偏度策略的价值。比如",[44,1070,1072],{"href":1071},"\u002Fglossary\u002Fwhat-is-trend-following","趋势跟踪","：少数几笔特大盈利撑起全部结果，同时也把",[44,1075,132],{"href":647},"的分母撑大了。",[10,1078,1079],{},"但把向上的波动当作没有代价，只是一种建模选择，不是定律；而对一整类交易系统来说，它恰恰用反了。网格、马丁格尔和逆势加仓摊平策略，涨和跌都出自同一个机制：逆着持仓方向不断加码。用索提诺比率给它们打分，平常的小额盈利留在分子里，大部分离散程度却从分母里消失了，直到这个机制失灵的那一天。索提诺比率美化的，恰恰是风险最难从已平仓交易里看出来的那些策略。",[10,1081,1082,1083,1085],{},"稳妥的看法是：索提诺比率可以作为收益序列的第二意见，但绝不是定论，更不能取代",[44,1084,377],{"href":376},"，最大回撤衡量的，才是投资者真正要熬过去的东西。",[14,1087,76],{"id":76},[10,1089,79,1090,1092,1093,1096,1097,1099,1100,1103],{},[81,1091,660],{},"，但关键在分布：",[81,1094,1095],{},"38.5%"," 的账户跌幅从未超过 5%，而 ",[81,1098,673],{}," 曾下跌超过 20%，",[81,1101,1102],{},"17.6%"," 亏掉了峰值的一半以上。",[10,1105,1106],{},"两个规模几乎相当的群体，分处两端。正因为是这种形态，只看下行的统计量才在同一批数据上既诱人又危险：深尾部的账户并不是左右对称地晃到那里的，而是朝一个方向一路跌下去的。任何指标要丢掉一半分布，就必须丢对那一半。",[10,1108,1109,1110,1113,1114,1117,1118,1122],{},"原始数据里本身就已经有另一种不对称。已发布账户共支付了 ",[81,1111,1112],{},"$862,547"," 库存费，在产生过库存费的账户中，",[81,1115,1116],{},"86.3%"," ",[44,1119,1121],{"href":1120},"\u002Fguides\u002Fswap-and-carry-trade","净库存费为负","。还没等任何策略做决定，平均持仓成本就已经在往下拖了。",[14,1124,118],{"id":117},[10,1126,1127,1128,1131,1132,1134,1135,1137,1138,178,1140,1142,1143,1145,1146,1148,1149,1155],{},"哪里都没有，这一点值得说清楚。",[81,1129,1130],{},"ShowMyTrades 上没有任何面板、图表、小组件或导出文件显示索提诺比率","，我们的统计引擎也不计算它：界面背后并没有一个存好了、只等显示的数值。",[81,1133,124],{},"模块在",[81,1136,128],{},"一栏中公布",[81,1139,132],{},[81,1141,47],{},"；",[81,1144,691],{},"面板公布",[81,1147,219],{},"（最差的净值回撤）和",[44,1150,1152],{"href":1151},"\u002Fglossary\u002Fwhat-is-drawdown-on-balance",[81,1153,1154],{},"余额回撤","（只按已平仓结果计算的最差回撤）。这些就是我们页面目前提供的下行指标。",[10,1157,1158,1159,143,1162,143,1165,153,1168,1171],{},"如果想要某个已发布账户的索提诺比率，可以自己算。账户页面的交易表格支持把已平仓交易列表下载为 CSV，每笔交易一行，包含 ",[81,1160,1161],{},"Profit",[81,1163,1164],{},"Swap",[81,1166,1167],{},"Commission",[81,1169,1170],{},"Net profit"," 列：最后一列正是我们计算夏普比率所用的序列。选好目标并写明，任何人拿到同一份文件都能算出同样的数字。这比页面上挂一个来历不明的比率要好，也是我们自己发布这个指标之前必须达到的标准。",[14,1173,192],{"id":192},[194,1175,1176,1182,1188,1194],{},[197,1177,199,1178,1181],{},[81,1179,1180],{},"索提诺比率就是更好的夏普比率。","”它回答的是另一个问题，不是同一个问题的更好答案。它衡量相对目标的不足，夏普比率衡量离散程度，两者都不管回撤有多深。",[197,1183,199,1184,1187],{},[81,1185,1186],{},"索提诺比率高，说明亏损小。","”它只说明低于目标的结果少或小。一笔还没发生的灾难性交易，在这两个比率里都看不到。",[197,1189,199,1190,1193],{},[81,1191,1192],{},"不同网站的索提诺比率可以互相比较。","”只有目标和除数惯例都一致时才行，而这两项通常都不公布。",[197,1195,199,1196,1199],{},[81,1197,1198],{},"ShowMyTrades 把自己的索提诺比率藏起来了。","”没什么可藏的：我们根本不计算它。用已平仓交易的导出文件，您可以自己算。",[10,1201,1202,1203,240],{},"我们确实公布的下行指标是什么、该怎么看，请参阅",[44,1204,377],{"href":758},{"title":30,"searchDepth":242,"depth":242,"links":1206},[1207,1208,1209,1210,1211],{"id":16,"depth":242,"text":16},{"id":54,"depth":242,"text":54},{"id":76,"depth":242,"text":76},{"id":117,"depth":242,"text":118},{"id":192,"depth":242,"text":192},"索提诺比率用下行偏差取代总波动率。本文介绍公式、公式里藏着的建模选择，以及我们为什么不公布这个指标。",{},"\u002Fzh\u002Fglossary\u002Fwhat-is-the-sortino-ratio",[770,258,541,260],{"title":1018,"description":1212},"zh\u002F5.glossary\u002Fwhat-is-the-sortino-ratio","索提诺比率 (Sortino Ratio)","yJBnJvTj398d0Ye0udTJCiWcMfFgnOW7pDyPZ_h4c0o",[1221,1222],{"path":647},{"path":256},1793898069639]