Stock Trader V7
Sector rotation that trades itself — and the honest numbers behind it
A systematic equity engine running leveraged sector rotation off momentum and volatility targeting. The V7 "PUSH-20" strategy has traded hands-off on an Alpaca paper account since June 9, 2026. Its headline backtest was re-run under an independent validation harness, then put through a formal overfitting audit in August 2026 — both passes corrected the story, and the corrected version is the one published here.
The problem
Discretionary trading is inconsistent and eats your time. Most "systems" never actually run unattended — they live in a backtest notebook and die there. And most published backtests quietly flatter themselves through optimistic cost assumptions and through the dozens of variations that were tried and never mentioned, which is how a strategy looks great on paper and disappoints in practice.
What I built
The validation corrected the headline down: real performance is 19.0% CAGR against a −36.6% max drawdown, not the 20.7% / −30% originally published. The August overfitting audit then split the result in two. The direction is real — a Deflated Sharpe Ratio of 0.9991 says the edge is very unlikely to be an artefact of having tried many variations. The precision is not: many different risk-dial settings fit the history about equally well, so the specific tuned values are not identifiable from the data and should not be treated as optimal. Two independent simulators also disagree on trading-cost assumptions, which is logged as open rather than resolved quietly.
How it works
The engine ranks sectors by momentum, sizes positions with volatility targeting, then applies leverage. Execution runs as a two-speed loop: a 60-second cycle checks open positions and order fills, while a separate daily cycle recomputes signals and rotates the book. The more interesting work was the two rounds of self-criticism. The first re-ran the strategy at a realistic 10 basis points of slippage instead of the original 5, then attacked it with walk-forward testing, a timing-luck study across six different start months, and a parameter sweep. The second, in August 2026, applied a Deflated Sharpe Ratio and a Probability of Backtest Overfitting analysis — the standard tests for whether a result is simply the best of many things that were tried.
Highlights
- 19.0% CAGR / −36.6% max drawdown, validated at realistic 10 bps slippage
- Deflated Sharpe Ratio of 0.9991 — the edge survives correction for how many variants were tried
- The risk-dial tuning is NOT identifiable from the data, and the write-up says so plainly
- Original 20.7% / −30% headline found optimistic and publicly corrected
- Beat the benchmark in 5 of 5 walk-forward sub-periods, with a 0.8pp spread across six start months
- Known weakness documented: loses ~1.6× the benchmark in fast whipsaw selloffs
- Trades hands-off on Alpaca paper on a scheduled loop, with slippage tracking
Let's build yours.
Tell me what you're trying to ship — you'll get a scoped plan and a straight answer.