[{"data":1,"prerenderedAt":1269},["ShallowReactive",2],{"\u002Fzh\u002Fglossary\u002Fwhat-is-the-sortino-ratio":3,"\u002Fzh\u002Fglossary\u002Fwhat-is-the-sortino-ratio-related":255,"\u002Fzh\u002Fglossary\u002Fwhat-is-the-sortino-ratio-langs":1266},{"id":4,"title":5,"body":6,"category":237,"date":238,"description":239,"draft":240,"extension":241,"meta":242,"navigation":243,"path":244,"related":245,"seo":250,"stem":251,"term":252,"updated":253,"__hash__":254},"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":8,"toc":229},"minimark",[9,13,17,28,34,43,46,63,66,69,82,85,93,96,115,118,134,138,174,192,195,222],[10,11,12],"p",{},"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。",[14,15,16],"h2",{"id":16},"工作原理",[18,19,24],"pre",{"className":20,"code":22,"language":23},[21],"language-text","索提诺比率 = (平均收益 − 目标收益) \u002F 下行偏差\n","text",[25,26,22],"code",{"__ignoreMap":27},"",[18,29,32],{"className":30,"code":31,"language":23},[21],"下行偏差 = √( Σ min(rᵢ − 目标收益, 0)² \u002F n )\n",[25,33,31],{"__ignoreMap":27},[10,35,36,37,42],{},"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",[38,39,41],"a",{"href":40},"\u002Fglossary\u002Fwhat-is-standard-deviation-in-trading","标准差","完全一样。",[10,44,45],{},"这个公式里藏着两个选择，每一个都会改变结果：",[47,48,49,57],"ul",{},[50,51,52,56],"li",{},[53,54,55],"strong",{},"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。",[50,58,59,62],{},[53,60,61],{},"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。",[10,64,65],{},"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。",[14,67,68],{"id":68},"为什么重要",[10,70,71,72,76,77,81],{},"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",[38,73,75],{"href":74},"\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",[38,78,80],{"href":79},"\u002Fglossary\u002Fwhat-is-the-sharpe-ratio","夏普比率","的分母撑大了。",[10,83,84],{},"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。",[10,86,87,88,92],{},"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",[38,89,91],{"href":90},"\u002Fglossary\u002Fwhat-is-maximum-drawdown","最大回撤","，最大回撤衡量的，才是投资者真正要熬过去的东西。",[14,94,95],{"id":95},"数据显示了什么",[10,97,98,99,102,103,106,107,110,111,114],{},"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），",[53,100,101],{},"最深回撤的中位数为 9.7%","，但关键在分布：",[53,104,105],{},"38.5%"," 的账户跌幅从未超过 5%，而 ",[53,108,109],{},"38.2%"," 曾下跌超过 20%，",[53,112,113],{},"17.6%"," 亏掉了峰值的一半以上。",[10,116,117],{},"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。",[10,119,120,121,124,125,128,129,133],{},"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 ",[53,122,123],{},"$862,547"," 库存费，在产生过库存费的账户中，",[53,126,127],{},"86.3%"," ",[38,130,132],{"href":131},"\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。",[14,135,137],{"id":136},"在-showmytrades-上的位置","在 ShowMyTrades 上的位置",[10,139,140,141,144,145,148,149,152,153,155,156,158,159,162,163,166,167,173],{},"哪里都没有，这一点值得说清楚。",[53,142,143],{},"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。",[53,146,147],{},"高级统计","模块在",[53,150,151],{},"表现指标","一栏中公布",[53,154,80],{},"和",[53,157,41],{},"；",[53,160,161],{},"账户统计","面板公布",[53,164,165],{},"回撤","（最差的净值回撤）和",[38,168,170],{"href":169},"\u002Fglossary\u002Fwhat-is-drawdown-on-balance",[53,171,172],{},"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。",[10,175,176,177,180,181,180,184,187,188,191],{},"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 ",[53,178,179],{},"Profit","、",[53,182,183],{},"Swap",[53,185,186],{},"Commission"," 和 ",[53,189,190],{},"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。",[14,193,194],{"id":194},"常见误解",[47,196,197,204,210,216],{},[50,198,199,200,203],{},"“",[53,201,202],{},"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。",[50,205,199,206,209],{},[53,207,208],{},"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。",[50,211,199,212,215],{},[53,213,214],{},"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。",[50,217,199,218,221],{},[53,219,220],{},"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。",[10,223,224,225,228],{},"我们确实公布的下行指标是什么、该怎么看，请参阅",[38,226,91],{"href":227},"\u002Fguides\u002Fmaximum-drawdown-explained","指南。",{"title":27,"searchDepth":230,"depth":230,"links":231},2,[232,233,234,235,236],{"id":16,"depth":230,"text":16},{"id":68,"depth":230,"text":68},{"id":95,"depth":230,"text":95},{"id":136,"depth":230,"text":137},{"id":194,"depth":230,"text":194},"指标","2026-08-19T00:00:00.000Z","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。",false,"md",{},true,"\u002Fzh\u002Fglossary\u002Fwhat-is-the-sortino-ratio",[246,247,248,249],"what-is-the-sharpe-ratio","what-is-standard-deviation-in-trading","what-is-maximum-drawdown","what-is-recovery-factor",{"title":5,"description":239},"zh\u002F5.glossary\u002Fwhat-is-the-sortino-ratio","索提诺比率 (Sortino Ratio)","2026-09-17T00:00:00.000Z","yJBnJvTj398d0Ye0udTJCiWcMfFgnOW7pDyPZ_h4c0o",[256,584,810,1065],{"id":257,"title":258,"body":259,"category":237,"date":238,"description":572,"draft":240,"extension":241,"meta":573,"navigation":243,"path":574,"related":575,"seo":580,"stem":581,"term":582,"updated":238,"__hash__":583},"glossary\u002Fzh\u002F5.glossary\u002Fwhat-is-maximum-drawdown.md","什么是最大回撤？公式与回本计算",{"type":7,"value":260,"toc":565},[261,269,271,274,280,283,338,343,362,364,367,428,431,439,441,448,455,458,469,497,499,514,529,531,557],[10,262,263,264,268],{},"最大回撤是账户整个历史上出现过的最大一次",[38,265,267],{"href":266},"\u002Fglossary\u002Fwhat-is-drawdown","从峰值到谷底的跌幅","，以峰值的百分比表示。它是策略实际造成过的最大亏损，而不是账户主人心里预期的最大亏损。它只增不减：一旦出现 34% 的最大回撤，此后无论再创多少次新高，它都停在 34%。",[14,270,16],{"id":16},[10,272,273],{},"沿着净值曲线逐点往前看，同时记下到目前为止的最高值。每到一个点，就量一量它比这个最高值低了多少。全程找到的最深跌幅，就是最大回撤。",[18,275,278],{"className":276,"code":277,"language":23},[21],"最大回撤 % = 对所有 t 取最大值 [ (峰值(0..t) − 数值(t)) \u002F 峰值(0..t) ] × 100\n",[25,279,277],{"__ignoreMap":27},[10,281,282],{},"结果完全取决于用的是哪条曲线，ShowMyTrades 两条都公布。",[284,285,286,304],"table",{},[287,288,289],"thead",{},[290,291,292,295,298,301],"tr",{},[293,294,237],"th",{},[293,296,297],{},"计算基础",[293,299,300],{},"是否计入持仓？",[293,302,303],{},"说明了什么",[305,306,307,323],"tbody",{},[290,308,309,314,317,320],{},[310,311,312],"td",{},[53,313,165],{},[310,315,316],{},"净值（余额 + 浮动盈亏）",[310,318,319],{},"是",[310,321,322],{},"账户在那一刻实际价值的真实跌幅",[290,324,325,329,332,335],{},[310,326,327],{},[53,328,172],{},[310,330,331],{},"仅已平仓余额",[310,333,334],{},"否",[310,336,337],{},"已实现结果的跌幅",[10,339,340,342],{},[38,341,172],{"href":169},"的数字更好看，而且这是计算方式决定的：未实现的亏损它根本看不到。网格、马丁格尔和摊平加仓类系统正是利用了这一点：亏损单永远不平，余额曲线自然一直平滑。",[10,344,345,346,180,349,352,353,356,357,361],{},"所以，一个账户显示",[53,347,348],{},"回撤 41%",[53,350,351],{},"余额回撤 6%","，并不是数据出了问题。它曾经比峰值低 41%，已平仓的结果却只承认有过一段 6% 的低谷。净值那个数字才是诚实的，而",[53,354,355],{},"两者之间的差距","本身就说明了策略是怎么做的。如果两个数字很接近，说明仓位基本是在亏损出现的地方附近就平掉了，这正是",[38,358,360],{"href":359},"\u002Fglossary\u002Fwhat-is-a-stop-loss","止损","的用处。",[14,363,68],{"id":68},[10,365,366],{},"回本是不对称的：亏损是在较大的基数上亏掉的，盈利却得在变小了的基数上靠复利赚回来。",[284,368,369,378],{},[287,370,371],{},[290,372,373,375],{},[293,374,91],{},[293,376,377],{},"回到原峰值所需涨幅",[305,379,380,388,396,404,412,420],{},[290,381,382,385],{},[310,383,384],{},"10%",[310,386,387],{},"11.1%",[290,389,390,393],{},[310,391,392],{},"20%",[310,394,395],{},"25.0%",[290,397,398,401],{},[310,399,400],{},"30%",[310,402,403],{},"42.9%",[290,405,406,409],{},[310,407,408],{},"50%",[310,410,411],{},"100.0%",[290,413,414,417],{},[310,415,416],{},"70%",[310,418,419],{},"233.3%",[290,421,422,425],{},[310,423,424],{},"90%",[310,426,427],{},"900.0%",[10,429,430],{},"回撤在 20% 以内，这种不对称只是有点麻烦。超过 50%，它就成了这个账户压倒一切的事实：您得靠刚刚让资金腰斩的同一套策略，把资金翻倍。",[10,432,433,434,438],{},"最大回撤还决定了您实际能用的杠杆上限。一个系统历史上回撤过 30%，用两倍的",[38,435,437],{"href":436},"\u002Fglossary\u002Fwhat-is-position-sizing","仓位大小","去跑，就意味着 60% 的回撤，而没有任何证据表明您扛得住。",[14,440,95],{"id":95},[10,442,443,444,447],{},"这里的数字来自 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），最深回撤的中位数是 ",[53,445,446],{},"9.7%","，但无论往高还是往低，分布都很散。",[10,449,450],{},[451,452],"img",{"alt":453,"src":454},"ShowMyTrades 公开账户的最深回撤：38.5% 低于 5%，23.3% 在 5% 到 20% 之间，20.6% 在 20% 到 50% 之间，17.6% 超过 50%","\u002Fimages\u002Farticles\u002Fdrawdown-distribution.svg",[10,456,457],{},"中位数 9.7%，尾部却比多数公开宣传估计的更长。",[10,459,460,461,464,465,468],{},"在最差的一端，",[53,462,463],{},"17.6% 的账户曾比峰值低 50% 以上","，",[53,466,467],{},"38.2% 的账户曾回撤超过 20%","。换句话说，大约每六个账户就有一个碰上过上面回本表里 50% 那一行：光是回到原来的水平，就需要 100% 的涨幅。",[10,470,471,472,474,475,464,478,482,483,486,487,491,492,496],{},"最好的一端同样值得怀疑：",[53,473,105],{}," 的账户最大回撤不到 5%。其中有些确实稳健保守，但很多只是资历太浅。中位数水平的已发布账户有 ",[53,476,477],{},"171 笔已平仓交易",[38,479,481],{"href":480},"\u002Fglossary\u002Fwhat-is-average-trade-length","持仓时间","中位数 ",[53,484,485],{},"2.4 小时","；样本就这么大，最大回撤小，记录的只是“还没发生”，而不是“不会发生”。最大回撤说的是分布的尾部，而",[38,488,490],{"href":489},"\u002Fguides\u002Fhow-much-history-a-track-record-needs","尾部需要足够长的历史来支撑","，所以这个数字只有在",[38,493,495],{"href":494},"\u002Fglossary\u002Fwhat-is-a-verified-track-record","经验证的业绩记录","上才真正有分量。",[14,498,137],{"id":136},[10,500,501,502,504,505,155,507,509,510,513],{},"每个已发布账户页面的",[53,503,161],{},"面板里，",[53,506,165],{},[53,508,172],{},"紧挨着排在",[53,511,512],{},"月均 %"," 下方，两者都来自经纪商的数据，而非自行填报。先把这两个数字对比一下：它们之间的差距，是整个页面上最快的判断依据。",[10,515,516,517,519,520,523,524,528],{},"图表模块里有一个专门的",[53,518,165],{},"视图，用柱状图画出每日回撤，回答了醒目百分比回答不了的问题：账户在高水位线下方待了",[53,521,522],{},"多久","。七周内就收复的 25% 回撤，和十四个月后仍未收复的 25% 回撤，显示的是同一个数字，却完全不是同一种账户。",[38,525,527],{"href":526},"\u002Ftools\u002Fdrawdown-calculator","回撤计算器","可以拿您自己的余额套用上面的回本表。",[14,530,194],{"id":194},[47,532,533,539,545,551],{},[50,534,199,535,538],{},[53,536,537],{},"我的最大回撤今年改善了。","”它不可能改善。它是历史最大值，一年业绩再好也抹不掉。",[50,540,199,541,544],{},[53,542,543],{},"两个回撤数字应该一致。","”只有仓位在亏损附近就平掉时，两者才会一致。差距大是死扛亏损单的标志，不是数据出错。",[50,546,199,547,550],{},[53,548,549],{},"最大回撤小，风险就低。","”历史短的话就不一定。把它当风险指标之前，先问问它是多少笔交易、多少个月得出来的。",[50,552,199,553,556],{},[53,554,555],{},"百分比回撤和金额回撤可以互相替代。","”$2,000 的账户和 $200,000 的账户各跌 30%，风险特征相同，感受却天差地别；但能在账户之间比较的，只有百分比。",[10,558,559,560,564],{},"想在实时账户页面上把最大回撤和其他指标放在一起看，请阅读",[38,561,563],{"href":562},"\u002Fguides\u002Freading-a-trading-account-dashboard","如何读懂交易账户控制台","。",{"title":27,"searchDepth":230,"depth":230,"links":566},[567,568,569,570,571],{"id":16,"depth":230,"text":16},{"id":68,"depth":230,"text":68},{"id":95,"depth":230,"text":95},{"id":136,"depth":230,"text":137},{"id":194,"depth":230,"text":194},"最大回撤是账户有史以来从峰值到谷底的最深跌幅。本文介绍计算公式、回本对照表，以及数千个账户的真实分布。",{},"\u002Fzh\u002Fglossary\u002Fwhat-is-maximum-drawdown",[576,577,578,579],"what-is-drawdown","what-is-position-sizing","what-is-a-verified-track-record","what-is-a-profit-factor",{"title":258,"description":572},"zh\u002F5.glossary\u002Fwhat-is-maximum-drawdown","最大回撤 (Maximum Drawdown)","H1pMHY5DcM__vxubkgaLVDhrp8FNU0EbYexdvdzrV1o",{"id":585,"title":586,"body":587,"category":237,"date":238,"description":802,"draft":240,"extension":241,"meta":803,"navigation":243,"path":804,"related":805,"seo":806,"stem":807,"term":808,"updated":253,"__hash__":809},"glossary\u002Fzh\u002F5.glossary\u002Fwhat-is-recovery-factor.md","什么是恢复因子？公式、数值区间与真实数据",{"type":7,"value":588,"toc":795},[589,592,594,600,603,609,612,658,665,667,675,686,688,699,706,716,718,724,743,750,760,762,790],[10,590,591],{},"恢复因子就是净利润除以最大回撤：账户让主人每承受一单位跌幅，换回了多少收益。数值为 2.0，说明账户赚到的是最差时亏掉的两倍；低于 1.0，说明它挖下的坑还没填平。",[14,593,16],{"id":16},[18,595,598],{"className":596,"code":597,"language":23},[21],"恢复因子 = 净利润 \u002F 最大回撤\n",[25,599,597],{"__ignoreMap":27},[10,601,602],{},"公式本身很简单，难的是单位。流行的写法有两种，结果并不一致：",[18,604,607],{"className":605,"code":606,"language":23},[21],"货币形式   = 净利润 ($) \u002F 最大回撤 ($)\n百分比形式 = 总收益 (%) \u002F 最大回撤 (%)\n",[25,608,606],{"__ignoreMap":27},[10,610,611],{},"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。",[284,613,614,624],{},[287,615,616],{},[290,617,618,621],{},[293,619,620],{},"恢复因子",[293,622,623],{},"解读",[305,625,626,634,642,650],{},[290,627,628,631],{},[310,629,630],{},"低于 1.0",[310,632,633],{},"账户赚到的还不及最深那次跌幅",[290,635,636,639],{},[310,637,638],{},"1.0 – 2.0",[310,640,641],{},"有效，但回报和承受的风险差不多",[290,643,644,647],{},[310,645,646],{},"2.0 – 3.0",[310,648,649],{},"稳健，前提是历史够长，真正的最大回撤已经出现过",[290,651,652,655],{},[310,653,654],{},"高于 3.0",[310,656,657],{},"很强：相信之前，先看交易笔数和账户年龄",[10,659,660,661,664],{},"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。",[38,662,663],{"href":489},"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。",[14,666,68],{"id":68},[10,668,669,670,674],{},"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",[38,671,673],{"href":672},"\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。",[10,676,677,678,682,683,685],{},"它还补上了相邻指标的空白。",[38,679,681],{"href":680},"\u002Fglossary\u002Fwhat-is-a-profit-factor","盈利因子","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。",[38,684,80],{"href":79},"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。",[14,687,95],{"id":95},[10,689,98,690,464,697,564],{},[53,691,692,696],{},[38,693,695],{"href":694},"\u002Fglossary\u002Fwhat-is-time-weighted-return","时间加权收益率","的中位数为 +3.2%",[53,698,101],{},[10,700,701,702,705],{},"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 ",[53,703,704],{},"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。",[10,707,708,709,711,712,715],{},"尾部也不能忽视。这些账户中有 ",[53,710,109],{}," 曾下跌超过 20%，长期为正的占 ",[53,713,714],{},"63.0%","，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。",[14,717,137],{"id":136},[10,719,720,723],{},[53,721,722],{},"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。",[10,725,726,727,729,730,733,734,736,737,739,740,742],{},"百分比形式：取",[53,728,161],{},"面板里的",[53,731,732],{},"增长","，除以同一面板中往下第四行的",[53,735,165],{},"。两者都是百分比，所以可以直接相除。注意它们各自的含义：",[53,738,732],{},"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；",[53,741,165],{},"是记录到的最差净值跌幅。",[10,744,745,746,749],{},"货币形式：同一面板里的",[53,747,748],{},"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。",[10,751,752,753,755,756,759],{},"不管用哪种形式，都先打开面板旁边的",[53,754,165],{},"图表视图。它以柱状图逐日画出回撤，能看出最大回撤是",[53,757,758],{},"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。",[14,761,194],{"id":194},[47,763,764,770,776,784],{},[50,765,199,766,769],{},[53,767,768],{},"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。",[50,771,199,772,775],{},[53,773,774],{},"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。",[50,777,199,778,781,782,564],{},[53,779,780],{},"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",[38,783,165],{"href":266},[50,785,199,786,789],{},[53,787,788],{},"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。",[10,791,792,793,228],{},"分母怎么计量、最大值为什么永远不会下降，请参阅",[38,794,91],{"href":227},{"title":27,"searchDepth":230,"depth":230,"links":796},[797,798,799,800,801],{"id":16,"depth":230,"text":16},{"id":68,"depth":230,"text":68},{"id":95,"depth":230,"text":95},{"id":136,"depth":230,"text":137},{"id":194,"depth":230,"text":194},"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",[248,576,579,246],{"title":586,"description":802},"zh\u002F5.glossary\u002Fwhat-is-recovery-factor","恢复因子 (Recovery Factor)","T2aXIYYtEer-h9VriHvWPBlH-AZl3zT5JTJbKQEC33Y",{"id":811,"title":812,"body":813,"category":237,"date":238,"description":1054,"draft":240,"extension":241,"meta":1055,"navigation":243,"path":1056,"related":1057,"seo":1061,"stem":1062,"term":1063,"updated":253,"__hash__":1064},"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":814,"toc":1047},[815,818,820,826,846,881,884,886,889,900,903,930,932,940,943,951,953,976,1008,1010,1041],[10,816,817],{},"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。",[14,819,16],{"id":16},[18,821,824],{"className":822,"code":823,"language":23},[21],"标准差 = √( Σ (rᵢ − 均值)² \u002F n )\n",[25,825,823],{"__ignoreMap":27},[10,827,828,829,832,833,837,838,841,842,845],{},"其中 ",[25,830,831],{},"rᵢ"," 是第 ",[834,835,836],"em",{},"i"," 笔已平仓交易的净结果，",[25,839,840],{},"均值"," 是这些结果的平均值，",[25,843,844],{},"n"," 是已平仓交易笔数。这个数字代表什么，取决于四个细节：",[47,847,848,856,862,875],{},[50,849,850,564,853,855],{},[53,851,852],{},"每笔交易都按扣除成本后的净额计入",[25,854,831],{}," 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。",[50,857,858,861],{},[53,859,860],{},"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。",[50,863,864,867,868,870,871,874],{},[53,865,866],{},"它用的是总体标准差","，除以 ",[25,869,844],{}," 而不是 ",[25,872,873],{},"n − 1","。已平仓交易不足两笔时，显示为 0.00。",[50,876,877,880],{},[53,878,879],{},"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。",[10,882,883],{},"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。",[14,885,68],{"id":68},[10,887,888],{},"“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。",[10,890,891,892,894,895,899],{},"它也是我们的",[38,893,80],{"href":79},"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",[38,896,898],{"href":897},"\u002Fglossary\u002Fwhat-is-expectancy","期望值","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。",[10,901,902],{},"最有用的读法是比较，拿它和几行之外的两个平均值对照：",[47,904,905,911,924],{},[50,906,907,910],{},[53,908,909],{},"接近平均盈利","：结果很集中，账户大部分时候表现一致。",[50,912,913,916,917,920,921,564],{},[53,914,915],{},"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",[53,918,919],{},"最佳交易（盈亏）"," 和",[53,922,923],{},"最差交易（盈亏）",[50,925,926,929],{},[53,927,928],{},"远大于平均亏损","：至少有一笔亏损远远超出平常水平。",[14,931,95],{"id":95},[10,933,98,934,464,937,564],{},[53,935,936],{},"夏普比率中位数为 0.05",[53,938,939],{},"已平仓交易数中位数为 171 笔",[10,941,942],{},"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。",[10,944,945,950],{},[53,946,947,949],{},[38,948,481],{"href":480},"中位数为 2.4 小时","，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。",[14,952,137],{"id":136},[10,954,955,956,958,959,961,962,964,965,967,968,970,971,975],{},"账户页面的",[53,957,147],{},"模块里，",[53,960,151],{},"一栏中有一行，标签正是",[53,963,41],{},"，以账户货币显示，并附有提示说明。它紧挨在以它为分母的",[53,966,80],{},"上方、",[53,969,681],{},"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",[38,972,974],{"href":973},"\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。",[10,977,978,979,982,983,180,986,180,989,180,991,920,993,996,997,1000,1001,1007],{},"把它和同一模块的",[53,980,981],{},"交易","一栏对照着看：",[53,984,985],{},"平均盈利",[53,987,988],{},"平均亏损",[53,990,919],{},[53,992,923],{},[53,994,995],{},"总交易数","为这个数字提供了参照。模块里的",[53,998,999],{},"持续时间","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。",[38,1002,1004],{"href":1003},"\u002Fguides\u002Fcustom-analysis-and-filter-presets",[53,1005,1006],{},"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、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。",[14,1009,194],{"id":194},[47,1011,1012,1018,1029,1035],{},[50,1013,199,1014,1017],{},[53,1015,1016],{},"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。",[50,1019,199,1020,1023,1024,1028],{},[53,1021,1022],{},"它衡量的是市场波动率。","”它衡量的是您交易结果的",[38,1025,1027],{"href":1026},"\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。",[50,1030,199,1031,1034],{},[53,1032,1033],{},"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。",[50,1036,199,1037,1040],{},[53,1038,1039],{},"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。",[10,1042,1043,1044,228],{},"一组统计数据要有多少笔交易才有意义，请参阅",[38,1045,1046],{"href":489},"一份业绩记录需要多长的历史",{"title":27,"searchDepth":230,"depth":230,"links":1048},[1049,1050,1051,1052,1053],{"id":16,"depth":230,"text":16},{"id":68,"depth":230,"text":68},{"id":95,"depth":230,"text":95},{"id":136,"depth":230,"text":137},{"id":194,"depth":230,"text":194},"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",[246,1058,1059,1060],"what-is-volatility","what-is-expectancy","what-is-the-sortino-ratio",{"title":812,"description":1054},"zh\u002F5.glossary\u002Fwhat-is-standard-deviation-in-trading","标准差 (Standard Deviation)","d23H7esNsSG175eWPyoZzoJTWWR3SEfy5D-ZKP-x9wU",{"id":1066,"title":1067,"body":1068,"category":237,"date":238,"description":1258,"draft":240,"extension":241,"meta":1259,"navigation":243,"path":1260,"related":1261,"seo":1262,"stem":1263,"term":1264,"updated":253,"__hash__":1265},"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":1069,"toc":1251},[1070,1073,1075,1078,1084,1087,1093,1099,1102,1104,1107,1113,1121,1123,1133,1146,1149,1151,1186,1189,1209,1211,1246],[10,1071,1072],{},"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。",[14,1074,16],{"id":16},[10,1076,1077],{},"教科书上的定义用的是周期收益，并减去无风险利率：",[18,1079,1082],{"className":1080,"code":1081,"language":23},[21],"夏普比率 = (平均收益 − 无风险利率) \u002F 收益的标准差\n",[25,1083,1081],{"__ignoreMap":27},[10,1085,1086],{},"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：",[18,1088,1091],{"className":1089,"code":1090,"language":23},[21],"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",[25,1092,1090],{"__ignoreMap":27},[10,1094,1095,1096,1098],{},"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，",[38,1097,41],{"href":40},"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。",[10,1100,1101],{},"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。",[14,1103,68],{"id":68},[10,1105,1106],{},"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。",[10,1108,1109,1110,1112],{},"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",[38,1111,681],{"href":680},"的同一批已平仓交易。",[10,1114,1115,1116,1120],{},"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。",[38,1117,1119],{"href":1118},"\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。",[14,1122,95],{"id":95},[10,1124,98,1125,464,1127,464,1129,564],{},[53,1126,936],{},[53,1128,939],{},[53,1130,1131,949],{},[38,1132,481],{"href":480},[10,1134,1135,1136,1138,1139,1142,1143,1145],{},"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：",[38,1137,695],{"href":694},"中位数为 ",[53,1140,1141],{},"+3.2%","，长期为正的占 ",[53,1144,714],{},"。真实的优势落到单笔交易上，就是这个样子。",[10,1147,1148],{},"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。",[14,1150,137],{"id":136},[10,1152,955,1153,958,1155,961,1157,1159,1160,1162,1163,1166,1167,180,1169,180,1172,180,1174,187,1177,1180,1181,1185],{},[53,1154,147],{},[53,1156,151],{},[53,1158,80],{},"，并附有提示说明。它的正上方是它自己的分母",[53,1161,41],{},"，正下方是",[53,1164,1165],{},"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",[53,1168,681],{},[53,1170,1171],{},"Z 值（概率）",[53,1173,898],{},[53,1175,1176],{},"AHPR",[53,1178,1179],{},"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；",[38,1182,1184],{"href":1183},"\u002Fguides\u002Ftrading-widgets-and-embeds","可嵌入的小组件","里没有这个模块。",[10,1187,1188],{},"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。",[10,1190,1191,1192,180,1195,180,1198,155,1201,1203,1204,1208],{},"同一模块提供",[53,1193,1194],{},"表格",[53,1196,1197],{},"星期",[53,1199,1200],{},"小时",[53,1202,999],{},"视图；",[38,1205,1206],{"href":1003},[53,1207,1006],{},"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。",[14,1210,194],{"id":194},[47,1212,1213,1225,1234,1240],{},[50,1214,199,1215,1218,1219,1221,1222,1224],{},[53,1216,1217],{},"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",[53,1220,1165],{}," 那一行。普通的",[53,1223,80],{},"是逐笔计算的，这个门槛对它不适用。",[50,1226,199,1227,1230,1231,1233],{},[53,1228,1229],{},"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",[38,1232,165],{"href":266},"，另有单独的数字。",[50,1235,199,1236,1239],{},[53,1237,1238],{},"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。",[50,1241,199,1242,1245],{},[53,1243,1244],{},"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。",[10,1247,1248,1249,228],{},"夏普比率和账户页面上其他数字怎么对照着看，请参阅",[38,1250,563],{"href":562},{"title":27,"searchDepth":230,"depth":230,"links":1252},[1253,1254,1255,1256,1257],{"id":16,"depth":230,"text":16},{"id":68,"depth":230,"text":68},{"id":95,"depth":230,"text":95},{"id":136,"depth":230,"text":137},{"id":194,"depth":230,"text":194},"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。",{},"\u002Fzh\u002Fglossary\u002Fwhat-is-the-sharpe-ratio",[247,1060,249,579],{"title":1067,"description":1258},"zh\u002F5.glossary\u002Fwhat-is-the-sharpe-ratio","夏普比率 (Sharpe Ratio)","rF1ahLb2_s2YlvPIZt9lNpcU3FX-JK93QxZ6RXkVz7s",[1267,1268],{"path":1118},{"path":244},1793898057182]