Introduction

We replicated research 21469 in LEAN and in an independent vectorbt engine, then tested it outside the post's 2021-09 to 2026-09 window.

 

Method

Rule unchanged: equal weight in every US common stock a Senator disclosed buying in the last 30 days (by disclosure date), rebalanced when the set changes. Disclosures from FMP's senate feed (cross-checked later against Quiver). Local runs charge 5 bp fee plus 5 bp slippage per side.

 

Results, 2021-09 to 2026-07

  • LEAN (attached backtest): CAGR 30.17%, Sharpe 1.157, max drawdown -24.53%, alpha t-stat vs SPY 1.94, beta 1.06.
  • vectorbt: CAGR 28.72%, Sharpe 1.164, max drawdown -24.39%, alpha t-stat 1.96, beta 1.00.
  • Daily return correlation between the two engines: 0.979.
  • SPY over the same window: CAGR 12.38%, Sharpe 0.761.

Where the return came from

  • 2025 alone returned +94% to +114% (AI and semiconductor names disclosed right after the April 2025 tariff sell-off). The top 10 names made 71% of the profit.
  • Excluding 2025, Sharpe is 0.737, the same as SPY's 0.76.

Out of sample, 2018-07 to 2021-08 (before the post's window)

CAGR 23.23%, Sharpe 0.848 vs SPY 0.917, alpha t-stat 0.28, max drawdown -41.4%.

 

Conclusion

The engines agree, but the edge is beta of about 1 plus one exceptional year. Across 42 holding-period and lag variants the deflated Sharpe ratio is 0.39-0.55, below a 0.95 bar. We would not trade it.