Customized Momentum Strategy Back Testing with last 5 years

看班上讨论, 自己建了一个简单策略。打字快些,换了英文。这个策略很简陋,可以再改进。就作为抛砖引玉。。。

Inspired by the SPMO, I build a customized momentum strategy, back tested with last 5 years data (9/2020 ~ 9/2025).

Stragegy: from Mag7 pick the top 5 stocks based on previous 9 months stock price performance, and buy equally weights. then rebalance every 3 months.

The momentum strategy significantly outperformed all benchmarks, capturing AI/tech surges (e.g., heavy NVDA weighting in 2023–2025) while rotating out of laggards.

Summary Performance:

Metric Momentum Strategy No Balance (7 Stocks) SPY (Buy & Hold) VGT (Buy & Hold)
Final Portfolio Value $14.28 $10.85 $2.12 $4.56
Total Return +1,328% +985% +112% +356%
Annualized Return ~88% ~72% ~16% ~34%
Max Drawdown -32% (Q2 2022) -36% (Q2 2022) -25% (Q2 2022) -33% (Q2 2022)
Volatility (Std Dev) 26% 28% 15% 22%

The strategy's edge comes from dynamic selection, e.g., favoring NVDA/TSLA during rallies and avoiding META/AMZN in 2022 downturns. VGT (tech-focused) outperformed SPY but trailed the strategy due to fixed weighting.

Risks: Higher volatility from concentration in growth stocks

所有跟帖: 

Updated strategy, added 100MA, stock selected from top20 -olive2025- 给 olive2025 发送悄悄话 (3342 bytes) () 09/14/2025 postreply 18:01:43

Andreas Clenow 在他的第二本书Stocks on the Move: Beating the Market -jenning- 给 jenning 发送悄悄话 jenning 的博客首页 (355 bytes) () 09/14/2025 postreply 18:32:43

我记得好像看过你的文章,今天就把这个付诸实践跑了一下。 -olive2025- 给 olive2025 发送悄悄话 (95 bytes) () 09/14/2025 postreply 20:18:05

这个容易有future leak。应为MAG7是现在定义的,十年前不知道他们是MAG7. -mobius- 给 mobius 发送悄悄话 (96 bytes) () 09/14/2025 postreply 18:04:33

改进的第二版选股是从全美股票市值TOP20 -olive2025- 给 olive2025 发送悄悄话 (0 bytes) () 09/14/2025 postreply 18:28:07

马上试投起来! -jingke- 给 jingke 发送悄悄话 jingke 的博客首页 (0 bytes) () 09/14/2025 postreply 19:32:54

小仓位试投:)很多参数可以FINE TUNE -olive2025- 给 olive2025 发送悄悄话 (0 bytes) () 09/14/2025 postreply 20:19:18

为什么看过去九个月的performance, 而不是过去一年? -jingke- 给 jingke 发送悄悄话 jingke 的博客首页 (0 bytes) () 09/14/2025 postreply 19:45:24

一般这样的系统会用不同的lookback window回测 -jenning- 给 jenning 发送悄悄话 jenning 的博客首页 (167 bytes) () 09/14/2025 postreply 19:50:08

赞。另外注意前两个annualized return疑似有误可以核对一下,应是70%和61%,但不影响结论。 -ybdddnlyglny- 给 ybdddnlyglny 发送悄悄话 (0 bytes) () 09/14/2025 postreply 19:39:12

好,我再把程序和数据对一下 -olive2025- 给 olive2025 发送悄悄话 (0 bytes) () 09/14/2025 postreply 20:22:00

不用查程序:14.28^(1/5)-1=70%。 :) -ybdddnlyglny- 给 ybdddnlyglny 发送悄悄话 (0 bytes) () 09/14/2025 postreply 20:37:42

牛! -olive2025- 给 olive2025 发送悄悄话 (0 bytes) () 09/16/2025 postreply 13:06:47

作为Quant Field工作多年的,评价一下主要的issue -lionhill- 给 lionhill 发送悄悄话 lionhill 的博客首页 (773 bytes) () 09/14/2025 postreply 20:42:35

赞! -olive2025- 给 olive2025 发送悄悄话 (0 bytes) () 09/16/2025 postreply 13:07:45

赞! -Zaplet- 给 Zaplet 发送悄悄话 Zaplet 的博客首页 (0 bytes) () 09/15/2025 postreply 05:04:00

请您先登陆,再发跟帖!