OPEN-SOURCE SCRIPT
Updated Daily EMA60 Standard Error Table-zrbb-
Quickly estimate the stop-loss range required for daily-chart trading.
快速估算日线级别交易需要的止损范围
In quantitative finance / stock market research, "3 standard errors" (requiring a t-statistic ≥ 3, corresponding to roughly a 99.7% confidence level) has a few main uses:
1. Raising the significance bar to guard against data snooping
This is the most important application. When quant researchers backtest large numbers of factors, strategies, or parameter combinations, the conventional 95% confidence threshold (about 2 standard errors) is often too lenient — if you test hundreds or thousands of parameter combinations, some will look "significant" purely by chance (the multiple comparisons problem).
Researchers like Marcos López de Prado have proposed raising the significance threshold to a t-stat ≥ 3 (i.e., 3 standard errors) as a rule of thumb to substantially reduce the probability of "false discoveries" — mistaking luck for real alpha — and to combat backtest overfitting.
2. Building more conservative confidence intervals
For estimating parameters like strategy returns or factor exposures (beta):
Mean ± 1 SE → roughly 68% confidence interval
Mean ± 2 SE → roughly 95% confidence interval
Mean ± 3 SE → roughly 99.7% confidence interval (assuming approximate normality)
Using 3 SE means you require very strong evidence before drawing a conclusion, reducing the risk of mistaking noise for signal.
3. An important distinction to keep in mind
Standard Error (SE) and Standard Deviation (SD) are not the same thing:
Standard deviation measures the volatility of returns themselves (commonly used in Bollinger Bands, risk measures, etc.)
Standard error measures the uncertainty of an estimate (such as an average return or a beta coefficient); SE = SD / √n, so the larger the sample size, the smaller the SE
These two concepts are often conflated, but when judging whether a strategy's historical average return is truly significant, it's the standard error you should use — not the standard deviation of returns.
A practical caveat: even using 3 standard errors as a threshold is just a rule of thumb to reduce overfitting risk — it doesn't eliminate it entirely. Out-of-sample validation, economic rationale, and transaction cost considerations remain essential. Statistical significance alone doesn't guarantee a strategy will actually work in live trading.
在股市/量化投资研究里,"3倍标准误差"（即要求 t 统计量 ≥ 3，对应约 99.7% 的置信区间）主要有几个用途:
1. 提高策略显著性门槛,对抗数据窥探(data snooping)
这是最重要的应用场景。量化研究员在回测大量因子、策略参数组合时,常规的 95% 置信区间(约 2 倍标准误差)门槛太宽松——如果你测试了成百上千种参数组合,总会有一些"看起来显著"其实纯属巧合(多重比较问题)。
Marcos López de Prado 等人在量化文献中提出:把显著性门槛提高到 t-stat ≥ 3(即 3 倍标准误差),可以大幅降低"假发现"(把纯粹运气当作真实alpha)的概率,是应对回测过拟合的一种经验法则。
2. 构建更保守的置信区间
对于策略收益率、因子暴露(beta)等参数的估计:
均值 ± 1倍SE → 约68%置信区间
均值 ± 2倍SE → 约95%置信区间
均值 ± 3倍SE → 约99.7%置信区间(假设近似正态分布)
用3倍SE意味着你要求证据非常强才愿意下结论,减少"把噪音当信号"的风险。
3. 需要注意的一个关键区分
标准误差(Standard Error, SE)和标准差(Standard Deviation, SD)不是一回事:
标准差衡量的是收益率本身的波动性(常用于布林带、风险度量)
标准误差衡量的是"某个估计量"(比如平均收益、beta系数)的不确定性,SE = SD / √n,样本量越大SE越小
很多人会混用这两个概念,但在判断"这个策略的历史平均收益是否真实显著"时,该用的是标准误差,而不是收益率的标准差。
实际应用提醒:即便用3倍标准误差作门槛,也只是降低过拟合风险的经验法则,不能完全消除。样本外验证、经济学逻辑支撑、交易成本考量仍然必不可少——单纯统计显著不代表策略在实盘中一定有效。
快速估算日线级别交易需要的止损范围
In quantitative finance / stock market research, "3 standard errors" (requiring a t-statistic ≥ 3, corresponding to roughly a 99.7% confidence level) has a few main uses:
1. Raising the significance bar to guard against data snooping
This is the most important application. When quant researchers backtest large numbers of factors, strategies, or parameter combinations, the conventional 95% confidence threshold (about 2 standard errors) is often too lenient — if you test hundreds or thousands of parameter combinations, some will look "significant" purely by chance (the multiple comparisons problem).
Researchers like Marcos López de Prado have proposed raising the significance threshold to a t-stat ≥ 3 (i.e., 3 standard errors) as a rule of thumb to substantially reduce the probability of "false discoveries" — mistaking luck for real alpha — and to combat backtest overfitting.
2. Building more conservative confidence intervals
For estimating parameters like strategy returns or factor exposures (beta):
Mean ± 1 SE → roughly 68% confidence interval
Mean ± 2 SE → roughly 95% confidence interval
Mean ± 3 SE → roughly 99.7% confidence interval (assuming approximate normality)
Using 3 SE means you require very strong evidence before drawing a conclusion, reducing the risk of mistaking noise for signal.
3. An important distinction to keep in mind
Standard Error (SE) and Standard Deviation (SD) are not the same thing:
Standard deviation measures the volatility of returns themselves (commonly used in Bollinger Bands, risk measures, etc.)
Standard error measures the uncertainty of an estimate (such as an average return or a beta coefficient); SE = SD / √n, so the larger the sample size, the smaller the SE
These two concepts are often conflated, but when judging whether a strategy's historical average return is truly significant, it's the standard error you should use — not the standard deviation of returns.
A practical caveat: even using 3 standard errors as a threshold is just a rule of thumb to reduce overfitting risk — it doesn't eliminate it entirely. Out-of-sample validation, economic rationale, and transaction cost considerations remain essential. Statistical significance alone doesn't guarantee a strategy will actually work in live trading.
在股市/量化投资研究里,"3倍标准误差"（即要求 t 统计量 ≥ 3，对应约 99.7% 的置信区间）主要有几个用途:
1. 提高策略显著性门槛,对抗数据窥探(data snooping)
这是最重要的应用场景。量化研究员在回测大量因子、策略参数组合时,常规的 95% 置信区间(约 2 倍标准误差)门槛太宽松——如果你测试了成百上千种参数组合,总会有一些"看起来显著"其实纯属巧合(多重比较问题)。
Marcos López de Prado 等人在量化文献中提出:把显著性门槛提高到 t-stat ≥ 3(即 3 倍标准误差),可以大幅降低"假发现"(把纯粹运气当作真实alpha)的概率,是应对回测过拟合的一种经验法则。
2. 构建更保守的置信区间
对于策略收益率、因子暴露(beta)等参数的估计:
均值 ± 1倍SE → 约68%置信区间
均值 ± 2倍SE → 约95%置信区间
均值 ± 3倍SE → 约99.7%置信区间(假设近似正态分布)
用3倍SE意味着你要求证据非常强才愿意下结论,减少"把噪音当信号"的风险。
3. 需要注意的一个关键区分
标准误差(Standard Error, SE)和标准差(Standard Deviation, SD)不是一回事:
标准差衡量的是收益率本身的波动性(常用于布林带、风险度量)
标准误差衡量的是"某个估计量"(比如平均收益、beta系数)的不确定性,SE = SD / √n,样本量越大SE越小
很多人会混用这两个概念,但在判断"这个策略的历史平均收益是否真实显著"时,该用的是标准误差,而不是收益率的标准差。
实际应用提醒:即便用3倍标准误差作门槛,也只是降低过拟合风险的经验法则,不能完全消除。样本外验证、经济学逻辑支撑、交易成本考量仍然必不可少——单纯统计显著不代表策略在实盘中一定有效。
Release Notes
Increase the ratio of the standard error to the price.增加标准误差对价格的比值
Open-source script
In true PulseWire spirit, the creator of this script has made it open-source, so that traders can review and verify its functionality. Kudos to the author! While you can use it for free, remember that republishing the code is subject to our House Rules.
Disclaimer
The information and publications are not meant to be, and do not constitute, financial, investment, trading, or other types of advice or recommendations supplied or endorsed by PulseWire. Read more in the Terms of Use.
Open-source script
In true PulseWire spirit, the creator of this script has made it open-source, so that traders can review and verify its functionality. Kudos to the author! While you can use it for free, remember that republishing the code is subject to our House Rules.
Disclaimer
The information and publications are not meant to be, and do not constitute, financial, investment, trading, or other types of advice or recommendations supplied or endorsed by PulseWire. Read more in the Terms of Use.