THE SYSTEM · MOMV6.3
One model. Every decision.
MomV6.3 is a systematic momentum rotation strategy on the S&P 500. It was built in Python, backtested across 21 years of point-in-time index data, validated out-of-sample with frozen parameters — and since December 23, 2025, it has been trading real capital. This page explains how it works and what the evidence shows.
EVERY 21 TRADING DAYS
How a rotation cycle works.
STEP 01
Rank the entire index.
The model scores every S&P 500 constituent on multi-horizon price momentum. The universe is point-in-time: stocks that were later removed from the index are still ranked as they were on that date, so the backtest contains no survivorship bias.
STEP 02
Select seven, equal weight.
The strongest seven names each receive one-seventh of the portfolio. No conviction sizing, no overrides, no exceptions. Equal weight keeps any single position from dominating the outcome.
STEP 03
Hold for 21 trading days.
Between rebalances the model holds its positions. A profit-taking rule can trim a position into unusual single-day strength, and the risk layer monitors the portfolio daily — but there is no discretionary trading inside the cycle.
STEP 04
Rotate and repeat.
On schedule, the model re-ranks the index and rotates into the new top seven. Names that remain strong stay; names that faded are replaced. The cycle has repeated on time, every time, since the first live trade.
WHEN MARKETS BREAK
Three independent layers of protection.
LAYER 1 · REGIME
Market regime filter
A trend-following regime model decides whether conditions support full exposure. When the broad market deteriorates, the system reduces or exits equity exposure entirely rather than fighting the tape.
LAYER 2 · DRAWDOWN
Graduated risk-off and re-entry
If portfolio drawdown breaches the model’s threshold, exposure is cut — and re-entry is staged in gradually as conditions repair, instead of jumping back in all at once.
LAYER 3 · EMERGENCY
Position-level hard stop
A deep hard stop protects against single-position disasters. Across 21 years of backtesting it fired only three times — and each time, it was right to.
21.33 YEARS · 2005–2026
The backtest, with nothing hidden.
Figures below assume $3,500 invested monthly during the first year only — $42,000 contributed in total, with nothing added after month twelve. Everything beyond that is the model compounding on its own for two decades. All figures are averaged across 21 different start-date cohorts — no cherry-picked starting point.
| Metric | MomV6.3 | SPY (same contributions) |
|---|---|---|
| Final value | $1,303,168 | $368,629 |
| CAGR | 31.99% | 24.40% |
| Sharpe ratio | 0.969 | 0.789 |
| Max drawdown | -26.15% | -55.19% |
| Calmar ratio | 1.223 | 0.442 |
When it mattered most.
| Period | MomV6.3 | SPY |
|---|---|---|
| 2008 financial crisis (full year) | +1.55% | -36.24% |
| 2022 rate-hike bear (Jan–Oct) | +1.13% | -24.50% |
| COVID crash (Feb–Mar 2020) | -29.03% | -33.72% |
The model is not magic. It drew down almost 30% in the COVID crash, and it lagged the violent 2020 recovery — momentum strategies rotate into strength with a lag, and V-shaped reversals are their hardest environment. We show these numbers because a track record you can trust includes the parts that hurt.
THE TEST THAT COUNTS
Parameters frozen in 2016. Validated on everything after.
Every parameter was locked using data through 2016 only. The model then ran on 2017–2026 — nine years it had never seen. Most optimized strategies fall apart here. This one got stronger.
OOS verdict: PASS. Out-of-sample alpha +8.88% annualized over SPY. Out-of-sample Sharpe 1.155 — higher than in-sample. Robustness sweep across start dates: every variant beat SPY, worst case 29.4% CAGR and -41.6% drawdown.
Now it runs for real.
Backtests earn a hypothesis. Live capital earns a track record. See the live record →