[{"data":1,"prerenderedAt":888},["ShallowReactive",2],{"\u002Fzh\u002Fglossary\u002Fwhat-is-backtesting":3,"\u002Fzh\u002Fglossary\u002Fwhat-is-backtesting-related":264,"\u002Fzh\u002Fglossary\u002Fwhat-is-backtesting-langs":885},{"id":4,"title":5,"body":6,"category":248,"date":249,"description":250,"draft":251,"extension":252,"meta":253,"navigation":254,"path":255,"related":256,"seo":260,"stem":261,"term":262,"updated":249,"__hash__":263},"glossary\u002Fzh\u002F5.glossary\u002Fwhat-is-backtesting.md","什么是回测？定义、局限与实盘数据",{"type":7,"value":8,"toc":240},"minimark",[9,13,17,20,31,34,68,76,79,82,85,88,99,123,134,141,144,148,156,187,197,200,232],[10,11,12],"p",{},"回测是用历史价格数据模拟交易策略，估算它在过去会有怎样的表现。一张订单都不用发到经纪商，就能得出净值曲线、盈利因子和回撤。这些结果只是对策略的一种假设，并不是策略的真实记录。",[14,15,16],"h2",{"id":16},"工作原理",[10,18,19],{},"测试器逐根回放历史 K 线或 tick，在每一步套用策略的入场和离场规则，记下一次模拟成交。模拟和现实之间的所有差距，都藏在成本模型和成交模型里：",[21,22,27],"pre",{"className":23,"code":25,"language":26},[24],"language-text","每笔交易净结果 = 毛盈亏 − 点差 − 佣金 − 库存费 − 滑点\n\n点差成本 = 点差（价格单位）× 合约规模 × 手数\n库存费   = 库存费率 × 手数 × 持仓过夜数\n","text",[28,29,25],"code",{"__ignoreMap":30},"",[10,32,33],{},"输出靠不靠得住，取决于三项设置：",[35,36,37,45,57],"ul",{},[38,39,40,44],"li",{},[41,42,43],"strong",{},"建模质量","。MT4 和 MT5 用百分比表示。如果不提供真实 tick 数据，“每个 tick”（Every tick）模式其实是用 1 分钟 K 线插值出来的；“控制点”（control points）和“仅开盘价”（open prices only）都是近似算法，对依赖 K 线内部走势的策略美化得很厉害。",[38,46,47,50,51,56],{},[41,48,49],{},"点差假设","。多数测试器默认用固定点差。实盘点差会在每日展期（rollover）、新闻公布和时段开盘时扩大，而很多系统恰恰在这些时候交易。按 0.8 点建模、实际却要付 4.0 点的",[52,53,55],"a",{"href":54},"\u002Fglossary\u002Fwhat-is-a-spread","点差","，已经是另一个策略了。",[38,58,59,62,63,67],{},[41,60,61],{},"滑点与成交逻辑","。测试器默认订单都按请求价成交。",[52,64,66],{"href":65},"\u002Fglossary\u002Fwhat-is-slippage","滑点","、拒单、重新报价和部分成交，在测试器里统统不存在。",[10,69,70,71,75],{},"还有第四个问题，软件不会告诉您：这组参数本身就是从一轮搜索里筛出来的。如果您跑了 400 个变体，只留下最好的那个，眼前的结果就是事后挑出来的。这就是",[52,72,74],{"href":73},"\u002Fglossary\u002Fwhat-is-curve-fitting","曲线拟合","，也是测试表现和实盘表现之间最大的一道鸿沟。",[14,77,78],{"id":78},"为什么重要",[10,80,81],{},"要否决一个坏主意，回测是最便宜的办法；要确认一个好主意，它却是最昂贵的办法。它能说明一套规则在机制上前后一致，扛得住粗略的成本假设，回撤幅度也在您忍受得了的范围内。但它证明不了优势真的存在，因为它用到的每一项输入都是重建出来的，每一个参数都是看过数据之后才定下来的。",[10,83,84],{},"决定资金投向的人，应该把回测当成一个有一票否决权的筛选工具，而永远不要把它当作证据。",[14,86,87],{"id":87},"数据显示了什么",[10,89,90,91,94,95,98],{},"ShowMyTrades 以公开页面发布了 ",[41,92,93],{},"1 万+ 个账户","，背后是 ",[41,96,97],{},"15,436,464 笔已同步的交易","（2026 年 8 月）。没有一笔是模拟的：每一笔都由经纪商成交，再从经纪商那里读取回来。",[10,100,101,102,105,106,110,111,114,115,110,119,122],{},"这些已发布账户绝大多数是自动交易：自动交易占比的中位数为 ",[41,103,104],{},"99%","，53.9% 的账户自动化程度超过 90%，42.2% 低于 10%。自动交易的那部分，主要是投入资金之前做过回测的策略。在全部已发布账户中，",[52,107,109],{"href":108},"\u002Fglossary\u002Fwhat-is-a-profit-factor","盈利因子","中位数为 ",[41,112,113],{},"1.28","，",[52,116,118],{"href":117},"\u002Fglossary\u002Fwhat-is-the-sharpe-ratio","夏普比率",[41,120,121],{},"0.05","。测试报告里的数字，真正要比的就是这些。",[10,124,125,126,129,130,133],{},"模拟最容易低估的是成本，而这里的成本是实际扣掉的，不是假设出来的：佣金 ",[41,127,128],{},"$4,782,670","，库存费 ",[41,131,132],{},"$862,547","；在产生过库存费的账户中，86.3% 的净库存费为负。策略要隔夜持仓，回测却把库存费设为零，这不是稍微乐观了一点，而是漏掉了一笔反复发生的费用。",[10,135,136],{},[137,138],"img",{"alt":139,"src":140},"点差、佣金和库存费是三项独立的成本：ShowMyTrades 公开账户共被收取 $4,782,670 的佣金和 $862,547 的库存费","\u002Fimages\u002Farticles\u002Fwhere-costs-hide.svg",[10,142,143],{},"测试器用固定点差、又不算库存费，三项成本就漏掉了两项。",[14,145,147],{"id":146},"在-showmytrades-上的位置","在 ShowMyTrades 上的位置",[10,149,150,151,155],{},"回测有自己单独的区域，从不和账户混在一起。您可以上传 MetaTrader 4、MetaTrader 5 或 cTrader 的策略测试器报告，发布成标有 BACKTEST 的页面。页面会检查有没有篡改迹象，并单独列在您交易者主页的一个标签页里。回测页面从不带验证徽章，也不会出现在小组件、探索、投资组合和 Developer API 中，因为模拟结果不是",[52,152,154],{"href":153},"\u002Fglossary\u002Fwhat-is-a-track-record","业绩记录","。账户页面则不同，每一个都是从经纪商只读同步来的，所以上面的数字描述的是真实执行的交易，而不是模拟出来的交易。",[10,157,158,159,162,163,166,167,170,171,174,175,178,179,182,183,186],{},"回测藏起来的东西，这几个模块能揭出来：高级统计模块里的",[41,160,161],{},"佣金合计","和",[41,164,165],{},"已付库存费合计","，以账户货币显示策略的真实成本；",[41,168,169],{},"终端","面板，报告与交易服务器之间的 ",[41,172,173],{},"Ping"," 和",[41,176,177],{},"重传","，也就是您实际成交时所处的条件；还有",[41,180,181],{},"交易历史与余额变化","，它的",[41,184,185],{},"已平仓交易","标签页按真实成交价列出每一笔已执行的订单，而不是测试器会给的价格。",[10,188,189,192,193,196],{},[41,190,191],{},"业绩记录已验证","徽章（65 个账户持有）和",[41,194,195],{},"交易权限已验证","徽章（264 个账户持有）标出的，是从经纪商到页面这条链路确实经过核查、而不只是自己声称的记录。",[14,198,199],{"id":199},"常见误解",[35,201,202,209,220,226],{},[38,203,204,205,208],{},"“",[41,206,207],{},"它用了每个 tick，所以很准确。","”tick 建模改善的是价格路径，不是成交。经纪商的点差、执行和拒单，仍然没有算进去。",[38,210,204,211,214,215,219],{},[41,212,213],{},"它测试了十年。","”",[52,216,218],{"href":217},"\u002Fguides\u002Fhow-much-history-a-track-record-needs","时长不等于独立性","。如果这十年里只有一种占主导的市场状态，那它只是一次观察，而不是 2,500 次。",[38,221,204,222,225],{},[41,223,224],{},"只优化了三个参数。","”三个参数各取二十个值，就是 8,000 次实验。8,000 次实验里的赢家，从构造上就注定有一部分靠的是运气。",[38,227,204,228,231],{},[41,229,230],{},"模拟账户的前向测试已经证实了它。","”模拟服务器的成交偏乐观，也从不拒单。模拟账户的结果只是第二次模拟，算不上第一次实盘测试。",[10,233,234,235,239],{},"无论是测试结果还是实盘数字，在相信它之前，先按",[52,236,238],{"href":237},"\u002Fguides\u002Fverifying-trading-performance-claims","如何验证交易业绩宣传","逐项核对一遍。",{"title":30,"searchDepth":241,"depth":241,"links":242},2,[243,244,245,246,247],{"id":16,"depth":241,"text":16},{"id":78,"depth":241,"text":78},{"id":87,"depth":241,"text":87},{"id":146,"depth":241,"text":147},{"id":199,"depth":241,"text":199},"策略","2026-08-19T00:00:00.000Z","回测就是拿历史价格数据模拟交易策略。它能证明什么、证明不了什么，以及为什么与经纪商同步的实盘结果几乎总会和它不一样。",false,"md",{},true,"\u002Fzh\u002Fglossary\u002Fwhat-is-backtesting",[257,258,259],"what-is-curve-fitting","what-is-a-verified-track-record","what-is-slippage",{"title":5,"description":250},"zh\u002F5.glossary\u002Fwhat-is-backtesting","回测 (Backtesting)","JFA3C0OuDmaFAnqzJS-I02fA4QcPMlLuq0a6Gjsakv8",[265,454,654],{"id":266,"title":267,"body":268,"category":442,"date":249,"description":443,"draft":251,"extension":252,"meta":444,"navigation":254,"path":445,"related":446,"seo":450,"stem":451,"term":452,"updated":249,"__hash__":453},"glossary\u002Fzh\u002F5.glossary\u002Fwhat-is-a-verified-track-record.md","什么是经验证的业绩记录？定义与核查",{"type":7,"value":269,"toc":435},[270,273,275,298,304,307,313,316,318,330,333,335,345,361,363,384,395,397,428],[10,271,272],{},"经验证的业绩记录，是通过只读连接直接从经纪商持续传来的交易结果记录，整条链路上没有任何一个环节能让发布者编辑、删除或调整顺序。这正是“说法”和“证据”的区别。这个词里其实藏着两个不同的问题，关于验证的争论，大多源于把两者混为一谈。",[14,274,16],{"id":16},[10,276,277,278,282,283,287,288,292,293,297],{},"第一个问题：数据是不是真的。这由连接本身来回答：通道只能读取账户，永远不能写入，而且不间断地运行，所以记录是持续积累起来的，而不是挑个方便的时候导出来的。在 ShowMyTrades 上，这条通道是以下四种之一：安装在 MetaTrader 4 或 5 中的 EA（智能交易系统）、使用 ",[52,279,281],{"href":280},"\u002Fglossary\u002Fwhat-is-an-mt4-investor-password","MT4 投资者密码","的直接连接、",[52,284,286],{"href":285},"\u002Fctrader-account-tracking","cTrader OAuth2 授权","，或 ",[52,289,291],{"href":290},"\u002Ftradelocker-tracking","TradeLocker REST API","。这四种方式都不能下单、改单或平仓。只读凭据是行业里实现这一点的通用机制，",[52,294,296],{"href":295},"\u002Fglossary\u002Fwhat-is-myfxbook","Myfxbook"," 等第三方追踪工具也是这样连接 MetaTrader 账户的。",[21,299,302],{"className":300,"code":301,"language":26},[24],"经纪商服务器 → 只读连接 → ShowMyTrades → 公开页面\n　　　　　　   整条链路中没有任何可编辑的环节\n",[28,303,301],{"__ignoreMap":30},[10,305,306],{},"第二个问题：发布页面的人是不是账户的主人。只读数据流能证明交易确实存在，却证明不了交易是谁的，因为投资者密码可以交给任何人。要补上这个缺口，得做一件只有拥有交易权限的人才能做的事：以一个不可能成交的价格挂一笔单，并在注释栏里填上发布者的用户 ID。在 cTrader 上，一次 OAuth 授权就能同时回答这两个问题。",[10,308,309],{},[137,310],{"alt":311,"src":312},"“业绩记录已验证”回答的是：数据是否通过一条无法下单、改单或平仓的通道从经纪商传来；“交易权限已验证”回答的是：发布者是否控制着这个账户","\u002Fimages\u002Farticles\u002Ftwo-badges.svg",[10,314,315],{},"一份记录可能过了一关、没过另一关，所以两者分开统计。",[14,317,78],{"id":78},[10,319,320,321,162,325,329],{},"连接之后的一切都只是算术。只要输入能被改动，盈利因子、",[52,322,324],{"href":323},"\u002Fglossary\u002Fwhat-is-drawdown","回撤",[52,326,328],{"href":327},"\u002Fglossary\u002Fwhat-is-an-equity-curve","净值曲线","就全是摆设，页面做得再详细也没用。截图、PDF 对账单和电子表格有着同一个毛病：它们都出自请您信任它们的那个人之手。",[10,331,332],{},"连续性和只读访问同样重要。可以暂停、再恢复的记录，能让一个糟糕的月份凭空消失。数据流连续时，中断本身就会在页面上显示出来，这本身就是信息，而不是信息的缺失。",[14,334,87],{"id":87},[10,336,337,338,114,341,344],{},"验证要自己主动去做，还得花点功夫，数字也说明了这一点。截至 2026 年 8 月，在 ShowMyTrades 上发布的 1 万+ 个账户中，只有 ",[41,339,340],{},"65 个带有“业绩记录已验证”",[41,342,343],{},"264 个带有“交易权限已验证”","。证明所有权的账户比证明数据来源的多，因为挂一笔单，比去经纪商那里改密码容易。",[10,346,347,348,352,353,356,357,360],{},"已验证不等于表现好。在有交易历史的公开账户中，",[52,349,351],{"href":350},"\u002Fglossary\u002Fwhat-is-time-weighted-return","时间加权收益率","的中位数为 ",[41,354,355],{},"+3.2%","，63.0% 为正，最深回撤的中位数为 ",[41,358,359],{},"9.7%","，还有 17.6% 的账户曾经亏掉峰值的一半以上。这个分布来自 15,436,464 笔已同步交易，这才是诚实的记录该有的样子，和广告里的那些完全不同。",[14,362,147],{"id":146},[10,364,365,366,369,370,162,372,375,376,379,380,383],{},"两枚徽章都在每个账户页面头部的",[41,367,368],{},"认证","组里，位于各项数字上方：",[41,371,154],{},[41,373,374],{},"交易权限","，获得时显示绿色，未获得时显示灰色，每枚都有提示框，说明它在该平台上认证的是什么。同一组里还会显示",[41,377,378],{},"真实账户","或",[41,381,382],{},"模拟账户","，这是根据经纪商服务器判断的，不是用户自己填的。",[10,385,386,387,390,391,394],{},"验证在账户设置中管理，",[41,388,389],{},"投资者密码（可选）"," 字段就在那里：清空它并保存，业绩记录徽章当即变灰。",[41,392,393],{},"完整控制台","小组件也会显示这两枚徽章，因为它嵌入了账户头部，所以放在别人网站上的小组件，也带着自己的数据来源。",[14,396,199],{"id":199},[35,398,399,410,416,422],{},[38,400,401,404,405,409],{},[41,402,403],{},"指向某个追踪网站的链接，不等于验证","。",[52,406,408],{"href":407},"\u002Fguides\u002Fthird-party-verification-explained","任何平台都能托管未经验证的账户","。要看页面上的徽章，而不是网址里的域名。",[38,411,412,415],{},[41,413,414],{},"“业绩记录已验证”证明不了所有权","。它证明的是交易真实存在。转卖别人投资者密码的人，同样能提供真实的数据流，只是那些交易不是他自己的。",[38,417,418,421],{},[41,419,420],{},"已验证不等于盈利","。徽章只认证数据来源，别的一概不管。之后还得把增长率和回撤放在一起看。",[38,423,424,427],{},[41,425,426],{},"经验证的模拟账户，仍然是模拟账户","。偏偏在实盘成交最吃亏的那些时候，模拟成交显得过于乐观。",[10,429,430,431,404],{},"两枚徽章的验证步骤详解：",[52,432,434],{"href":433},"\u002Fguides\u002Fhow-to-verify-your-account","如何验证您的交易账户",{"title":30,"searchDepth":241,"depth":241,"links":436},[437,438,439,440,441],{"id":16,"depth":241,"text":16},{"id":78,"depth":241,"text":78},{"id":87,"depth":241,"text":87},{"id":146,"depth":241,"text":147},{"id":199,"depth":241,"text":199},"验证","经验证的业绩记录，是通过只读连接直接从经纪商持续获取的业绩数据，发布者无法改动其中任何内容。",{},"\u002Fzh\u002Fglossary\u002Fwhat-is-a-verified-track-record",[447,448,449],"what-is-an-mt4-investor-password","what-is-myfxbook","what-is-drawdown",{"title":267,"description":443},"zh\u002F5.glossary\u002Fwhat-is-a-verified-track-record","经验证的业绩记录 (Verified Track Record)","XXw0XywuXXvGTGPkURwZXnd-tURxBRrjPXvj-rr4h2k",{"id":455,"title":456,"body":457,"category":248,"date":249,"description":643,"draft":251,"extension":252,"meta":644,"navigation":254,"path":645,"related":646,"seo":649,"stem":650,"term":651,"updated":652,"__hash__":653},"glossary\u002Fzh\u002F5.glossary\u002Fwhat-is-curve-fitting.md","什么是曲线拟合？如何识别过度拟合的策略",{"type":7,"value":458,"toc":636},[459,462,464,467,473,476,479,510,512,515,521,523,540,548,555,557,568,601,603,629],[10,460,461],{},"曲线拟合也叫过度拟合（overfitting），指的是反复调整策略参数，直到它贴合测试用的历史数据，连其中的噪音也一并贴合。这样得出的结果把过去描述得分毫不差，对未来却毫无预测力。测试时看起来很出色的系统，一上实盘当月就亏钱，最常见的原因就是它。",[14,463,16],{"id":16},[10,465,466],{},"每一个可调的输入都是一个自由度。在一张取值网格里搜索，测的就不是一个策略，而是每种组合各做一次实验：",[21,468,471],{"className":469,"code":470,"language":26},[24],"实验次数 = v1 × v2 × ... × vk\n\n4 个参数 × 每个 15 个取值 = 50,625 次回测\n50,625 次中的最优者被选中，一部分是因为它的优势，\n一部分是因为它恰好碰上的那些噪音\n",[28,472,470],{"__ignoreMap":30},[10,474,475],{},"在同一段固定的历史上，实验做得越多，就越可以肯定：赢家的排名是运气给的。测试器不会提这一点，它只把赢家当成唯一的候选报告出来。",[10,477,478],{},"四个信号，大致按可靠程度排列：",[35,480,481,492,498,504],{},[38,482,483,486,487,491],{},[41,484,485],{},"参数悬崖","。某个设置往任一方向挪一格，表现就一落千丈。真实的优势会慢慢减弱，拟合出来的优势则是直接掉下悬崖。只在周期 47 时才管用的",[52,488,490],{"href":489},"\u002Fglossary\u002Fwhat-is-a-moving-average","移动平均线","，就是教科书式的例子。",[38,493,494,497],{},[41,495,496],{},"参数太多，样本太少","。一套有六个输入的规则只在 150 笔交易上验证过，就是欠定的。参数和样本之间没有固定比例，但方向毫无争议。",[38,499,500,503],{},[41,501,502],{},"净值曲线平滑得可疑","。真实的收益是参差不齐的。回测里一条近乎笔直的斜线，通常说明要么参数组是拟合出来的，要么系统是靠死扛亏损单活下来的。",[38,505,506,509],{},[41,507,508],{},"样本外崩溃","。在拟合窗口里成立，一出窗口就消失。这是决定性的检验，但只在您第一次看样本外数据时才算数。",[14,511,78],{"id":78},[10,513,514],{},"曲线拟合的系统会以一种特定而又昂贵的方式失败：它不会一开始就慢慢亏。只要市场和拟合期相似，它就表现得还过得去；一旦市场状态变了，它就崩了，而且往往正是在重仓的时候。交易者会归咎于运气不好，殊不知模型只是在做它被造出来要做的事。",[10,516,517,518,520],{},"解药不是更聪明的测试，而是一份实盘记录：真实承担点差、佣金和",[52,519,66],{"href":65},"，而且时间够长，经历过开发者从没见过的行情。",[14,522,87],{"id":87},[10,524,525,526,528,529,531,532,528,534,536,537,404],{},"以 ShowMyTrades 上有交易历史的公开账户为样本（2026 年 8 月）：",[52,527,109],{"href":108},"中位数 ",[41,530,113],{},"，历史最深",[52,533,324],{"href":323},[41,535,359],{},"，每个账户已平仓交易数中位数 ",[41,538,539],{},"171 笔",[10,541,542,543,547],{},"这三个数字构成了一个真实可信的参照分布。38.2% 的账户回撤曾超过 20%，17.6% 超过 50%。如果一个策略宣传盈利因子高于 3.0、",[52,544,546],{"href":545},"\u002Fglossary\u002Fwhat-is-maximum-drawdown","最大回撤","只有 4%，它宣称的不是本事，而是自己站在一个由 1,540 万笔已同步交易构成的分布之外。这种说法可能成立，但需要证据，而证据要看实盘。",[10,549,550,551,554],{},"171 笔已平仓交易这个中位数，值得单独记一下：要验证任何东西，这都是",[52,552,553],{"href":217},"一个很小的样本","，而大多数实盘账户恰恰就是凭这么大的样本被评判的。",[14,556,147],{"id":146},[10,558,559,560,563,564,567],{},"图表模块里的",[41,561,562],{},"逐笔交易增长","图按交易序号而不是日期来画曲线，长期闲置带来的美化效果就此去掉，账户是还在运作，还是只是还开着，一目了然。",[41,565,566],{},"月度收益表","会列出每一个亏损月和每一段“无交易活动”空档：策略在糟糕的月份之后被悄悄重新优化，通常会在那里留下一段明显的停顿。",[10,569,570,571,162,573,579,580,583,584,587,588,162,591,594,595,600],{},"在高级统计模块中，把",[41,572,109],{},[52,574,576],{"href":575},"\u002Fglossary\u002Fwhat-is-standard-deviation-in-trading",[41,577,578],{},"标准差","、",[41,581,582],{},"Z 值（概率）"," 放在一起看：盈利因子很高，单笔收益却异常均匀，说明账户平滑得不像经历过各种行情，而围绕某一段时期调出来的参数组，往往就是这种表现。",[41,585,586],{},"自定义分析","面板里的",[41,589,590],{},"魔术号",[41,592,593],{},"注释筛选","，可以把多策略账户上的某一个 EA（智能交易系统）单独拎出来，只看它的统计数据。",[52,596,598],{"href":597},"\u002Fglossary\u002Fwhat-is-a-verified-track-record",[41,599,191],{},"徽章目前有 65 个账户持有，它证明数据来自经纪商，而不是来自某份报告文件。",[14,602,199],{"id":199},[35,604,605,611,617,623],{},[38,606,204,607,610],{},[41,608,609],{},"它通过了样本外测试。","”只要您看过样本外结果、又据此改动过任何东西，这些数据就变成了样本内数据。机会只有一次。",[38,612,204,613,616],{},[41,614,615],{},"历史越长，越不会过度拟合。","”关键在于观察数量和自由参数的比例，而不是时间跨度。一边加年份一边加过滤条件，反而可能更糟。",[38,618,204,619,622],{},[41,620,621],{},"滚动前向优化（walk-forward）能解决这个问题。","”它能减轻问题，也可能重演问题：在同一个品种上逐个窗口重新优化，搜索的仍然是同一段历史。",[38,624,204,625,628],{},[41,626,627],{},"已经是实盘了，不可能是曲线拟合。","”一出问题就重新调参的系统，是在实盘中被拟合。账户是真的，优势却还在被倒推着编出来。",[10,630,631,632,404],{},"什么样的记录能回答这个问题，什么样的不能，请阅读",[52,633,635],{"href":634},"\u002Fguides\u002Fverified-vs-unverified-track-records","经验证与未经验证的业绩记录",{"title":30,"searchDepth":241,"depth":241,"links":637},[638,639,640,641,642],{"id":16,"depth":241,"text":16},{"id":78,"depth":241,"text":78},{"id":87,"depth":241,"text":87},{"id":146,"depth":241,"text":147},{"id":199,"depth":241,"text":199},"曲线拟合就是把策略反复优化，直到它把过去的数据描述得完美无缺，却什么也预测不了。有哪些警示信号，为什么只有实盘结果才能下定论。",{},"\u002Fzh\u002Fglossary\u002Fwhat-is-curve-fitting",[647,648,258],"what-is-backtesting","what-is-a-profit-factor",{"title":456,"description":643},"zh\u002F5.glossary\u002Fwhat-is-curve-fitting","曲线拟合 (Curve Fitting)","2026-09-10T00:00:00.000Z","seQQ_6nNRjX5XjiWi7epEXN6cZrnk69LgGFRtAA932E",{"id":655,"title":656,"body":657,"category":872,"date":249,"description":873,"draft":251,"extension":252,"meta":874,"navigation":254,"path":875,"related":876,"seo":880,"stem":881,"term":882,"updated":883,"__hash__":884},"glossary\u002Fzh\u002F5.glossary\u002Fwhat-is-slippage.md","什么是滑点（Slippage）？定义、成因与真实数据",{"type":7,"value":658,"toc":865},[659,662,664,670,673,679,685,696,703,705,718,724,726,745,756,758,761,828,830,857],[10,660,661],{},"滑点，是您下单时预期的价格与实际成交价格之间的差额。从您的终端发出请求，到经纪商服务器完成撮合，中间这段时间市场还在动，滑点就是这样产生的。滑点可能对您不利，也可能对您有利，但实际上并不对称：最容易出现大滑点的时候，恰恰也是最需要精确成交的时候。",[14,663,16],{"id":16},[21,665,668],{"className":666,"code":667,"language":26},[24],"滑点（点） = (成交价 − 请求价) \u002F 点的大小\n             卖出订单取相反符号\n",[28,669,667],{"__ignoreMap":30},[10,671,672],{},"滑点有三个来源。",[10,674,675,678],{},[41,676,677],{},"延迟","。从请求到执行，每一毫秒价格都可能变。延迟来自通往交易服务器的网络路径、终端自身的处理，以及经纪商的撮合队列。",[10,680,681,684],{},[41,682,683],{},"流动性","。订单要在订单簿上成交。如果报价上可成交的数量比您的订单小，剩下的部分就会在下一档价格成交。所以仓位越大，滑点越大；市场越深，滑点越小。",[10,686,687,690,691,695],{},[41,688,689],{},"订单类型","。市价单按当时能拿到的价格成交，所以会滑点。限价单从不滑点，但可能成交不了，这本身也是一种成本。Stop 类订单（包括",[52,692,694],{"href":693},"\u002Fglossary\u002Fwhat-is-a-stop-loss","止损","）一触发就变成市价单，这是最重要的一点：止损保证的是您能离场，而不是离场的价格。",[10,697,698,699,702],{},"最糟糕的时段是可以预见的：预定的经济数据公布、每日展期（rollover）前后流动性稀薄的时段、周日开盘跳空，以及一大批止损同时触发、把订单簿抽空的那一刻。",[41,700,701],{},"正滑点","也存在（成交价比请求价更好）。经纪商如果让您承担负滑点，也把正滑点照样给您，做法就是对称的；如果好的成交自己留着，差的推给您，那就不对称。",[14,704,78],{"id":78},[10,706,707,708,712,713,717],{},"滑点就是纸面上的策略和服务器上的同一策略之间的差距。用历史卖出价（bid）、按零滑点跑出来的",[52,709,711],{"href":710},"\u002Fglossary\u002Fwhat-is-backtesting","回测","，假设了一种根本不存在的执行方式；",[52,714,716],{"href":715},"\u002Fglossary\u002Fwhat-is-average-trade-length","平均每笔交易","越短，这个假设凭空造出来的优势就占得越多。",[10,719,720,721,723],{},"目标越小，这笔账越无情。半个点的不利滑点，是 5 点目标的十分之一，却只是 200 点目标的四百分之一；而且它是在",[52,722,55],{"href":54},"和佣金之外另算的，不是替代它们。",[14,725,87],{"id":87},[10,727,728,729,114,732,735,736,739,740,744],{},"不等合适时机就出手的系统，最看重执行质量，而平台上的大多数账户正是这样。截至 2026 年 8 月，ShowMyTrades 上有交易历史的公开账户中，",[41,730,731],{},"自动交易占比的中位数为 99%",[41,733,734],{},"53.9% 的账户自动化程度超过 90%","（42.2% 低于 10%）。这些账户连接了 ",[41,737,738],{},"703 个不同的经纪商服务器","：",[52,741,743],{"href":742},"\u002Fguides\u002Fsame-strategy-different-brokers","同一个策略，发往七百个不同的撮合引擎","，面对七百种不同的延迟情况。",[10,746,747,748,751,752,755],{},"持仓时间又放大了这种风险。这些账户的",[41,749,750],{},"持仓时间中位数为 2.4 小时","，中位数账户已平仓 ",[41,753,754],{},"171 笔交易","：执行成本付得频繁，每笔交易又没有多少时间让价格波动把它消化掉。",[14,757,147],{"id":146},[10,759,760],{},"我们不记录您的终端请求的价格，所以不公布逐笔滑点这个指标；只靠经纪商数据，任何诚实的平台也都公布不了。我们公布的是能衡量的那一半：造成滑点的延迟。",[35,762,763,797,818],{},[38,764,765,767,768,771,772,777,778,783,784,579,787,579,790,793,794,404],{},[41,766,169],{},"面板（MT5 上为",[41,769,770],{},"终端与已打开的图表","）会显示 ",[52,773,775],{"href":774},"\u002Fglossary\u002Fwhat-is-terminal-ping",[41,776,173],{},"，即终端与经纪商交易服务器之间的往返时间，以及",[52,779,781],{"href":780},"\u002Fglossary\u002Fwhat-is-packet-retransmission",[41,782,177],{},"，即需要重新发送的网络数据包比例。Ping 低于 50 ms 属于优秀；高于 200 ms 可能意味着执行更慢、滑点更多。重传率一直偏高，说明网络不稳定，或者交易服务器离得太远。同一面板还会显示",[41,785,786],{},"终端版本号",[41,788,789],{},"内存",[41,791,792],{},"自动交易","和 ",[41,795,796],{},"DLL 导入",[38,798,799,801,802,804,805,162,808,811,812,162,814,817],{},[41,800,181],{},"下的",[41,803,185],{},"表格，每个订单都有",[41,806,807],{},"开仓价",[41,809,810],{},"平仓价","，旁边是",[41,813,694],{},[41,815,816],{},"止盈","列。把离场价位的设定值和实际成交的位置一比，就是从经纪商记录里能读到的最接近滑点的数据。",[38,819,820,823,824,827],{},[41,821,822],{},"高级统计","中的",[41,825,826],{},"最差交易（点）"," 可以用来快速检查。如果最差交易远远超出账户原本设定的止损距离，那就是跳空或滑点离场留下的痕迹，而不是入场没做好。",[14,829,199],{"id":199},[831,832,833,839,845,851],"ol",{},[38,834,835,838],{},[41,836,837],{},"止损保证的是离场，不是价格","。遇到跳空，它会按第一个能成交的价格平仓，可能远远超出您设定的价位。",[38,840,841,844],{},[41,842,843],{},"市价单上的“零滑点”，是谁都兑现不了的承诺","。保证止损确实有，但那是另一种产品，要另外付费。",[38,846,847,850],{},[41,848,849],{},"滑点不一定是亏损","。看看您的成交价有没有比请求价更好的时候；如果从来没有，这本身就说明了这家经纪商的情况。",[38,852,853,856],{},[41,854,855],{},"模拟账户的结果偏乐观","。模拟服务器不模拟排队位置和订单簿深度，所以在模拟账户上完美无缺的策略，到了实盘可能只是勉强能用。",[10,858,859,860,864],{},"想了解 Ping、重传与页面上的其他内容如何结合起来看，请阅读",[52,861,863],{"href":862},"\u002Fguides\u002Freading-a-trading-account-dashboard","如何读懂交易账户控制台","指南。",{"title":30,"searchDepth":241,"depth":241,"links":866},[867,868,869,870,871],{"id":16,"depth":241,"text":16},{"id":78,"depth":241,"text":78},{"id":87,"depth":241,"text":87},{"id":146,"depth":241,"text":147},{"id":199,"depth":241,"text":199},"成本","滑点是您预期的价格与实际成交价之间的差额。本文解释它为什么总在最要命的时候最严重，以及延迟如何让它变得可以衡量。",{},"\u002Fzh\u002Fglossary\u002Fwhat-is-slippage",[877,878,879,647],"what-is-a-spread","what-is-a-stop-loss","what-is-a-vps",{"title":656,"description":873},"zh\u002F5.glossary\u002Fwhat-is-slippage","滑点 (Slippage)","2026-08-29T00:00:00.000Z","_usMaUZBN8x6rtKI1x17K10-_LxzY0duV2FLyHX73gE",[886,887],{"path":710},{"path":255},1793898068485]