Beat the index.
Lose less doing it.
A systematic position book, published in the open. Two sleeves, fourteen names, one rebalance a month. Math, not headlines.
$ in. $669 out. SPY made $545.
Try any amount. Both books recompute from 2013.
Same start. Same point-in-time Russell 1000. Same fifteen basis points of cost per rebalance. Lavender is the book. Dashed is SPY.
$669.62 vs SPY $545.63
Backtest results computed via walk-forward + combinatorial purged cross-validation on the point-in-time Russell 1000 universe. Hypothetical results do not reflect actual investment results and are not a guarantee of future returns. Live track record begins May 7, 2026.
Loading the live record…
The live record is shown win or lose, straight from the published book. A few weeks is a small sample; the backtest above is the evidence, this is the receipt.
What the model holds today.
Two sleeves, seven names each, rebalanced on the seventh of every month. Hedge defends, growth attacks, and the regime dial sets the split. This is the live book, picked by the same engine the track record above was built from.
Two sleeves. One book. Fourteen at the cap.
Hedge defends. Growth attacks. Same universe. Same monthly cadence. One cash pool. The regime dial sets the split. The blend, not the forecast, is the edge.
Point-in-time Russell 1000
About 1,140 U.S. tickers per snapshot, sourced from monthly iShares IWB membership lists. Names need eighteen months of history before they can be ranked. Survivor bias is excluded by construction, not by hope.
Two signal stacks, two top sevens
The hedge sleeve ranks the full Russell 1000 with a defensive-leaning composite. The growth sleeve ranks the same universe with a faster, momentum-heavier stack tuned for mid caps. Each picks its own seven. Sleeve-to-sleeve correlation is +0.09.
Three buckets size the split
A sixty-bar rolling read of market breadth and trend classifies the tape as risk-on, cautious, or defensive. Risk-on tilts to growth with ten percent cash. Defensive lifts cash above half and leans on hedge. The dial is mechanical.
Seven slots, three caps
Per sleeve: maximum five per GICS sector, three per industry group, plus a per-name idiosyncratic-vol cap. Whatever does not fit the caps goes to cash via the BIL proxy.
Boring discipline. Published proof.
Every backtest number on this page comes from one fourteen-year run, frozen in March 2026 so it can never quietly drift. The live book republishes monthly on top of it. Below: the evidence it works, and the methodology that keeps it honest.
Evidence
Fourteen years out-of-sample. Twelve-month embargo on the lookback signals. Combinatorially disjoint paths. Not a single lucky window. López de Prado (2018).
First half beat SPY by 1.64 percentage points of CAGR. Second half beat it by 3.01. The book is not riding one regime.
Hedge and growth win on different slices of the calendar. Blending them cuts max drawdown five points below SPY. Costs less than a point of CAGR versus growth alone.
Average alpha 1.25 percentage points per year across every rolling twelve-month window. Best window +41.5pp. Worst -26.3pp.
Methodology
Every rebalance uses the index membership as it actually stood on that date. Never the latest list. Names with fewer than eighteen months of history are excluded, not zero-filled.
Each name's momentum is stripped of market, size, value, profitability, investment, and the momentum factor itself. What's left is name-specific drift. Fama-French (2015); Blitz-Huij-Martens (2011).
Frazzini-Israel-Moskowitz (2018) real-money execution estimate. Every Sharpe and CAGR on this page is reported after the cost drag.
Three buckets driven by rolling breadth and trend over the index. Per-name caps every rebalance. Cash slack to the BIL proxy. The blend is the discipline. The algorithm does not forecast.
References. Jegadeesh & Titman (1993); Asness, Moskowitz & Pedersen (2013); Blitz, Huij & Martens (2011); Fama & French (2015); Faber (2007); Bailey & López de Prado (2014); López de Prado (2018); Frazzini, Israel & Moskowitz (2018); George & Hwang (2004); Gettleman & Marks (2006).
Data. Backtest numbers are frozen at the March 2026 run and stamped where shown. The live record and holdings republish on every monthly rebalance.
Toggle the regime. Watch the book reshape.
Three regimes, three allocations. The dial below starts on whichever regime the book holds today. In risk-on, ten percent sits in cash and the sleeves run nearly equal. In defensive, more than half the book is parked. The regime classifier picks the bucket. The bucket picks the allocation.
Versioned. Dated. No silent edits.
The backtest is frozen and the page is versioned. When the methodology changes, the change lands here with a date. What you read above is v1.
The regime dial now opens on the regime the book holds today, straight from the API. Decorative motion removed; every element that stays has a job. One rebalance a month, stated plainly.
Backtest frozen at the March 2026 run. The live record renders straight from the published book, win or lose. Future methodology revisions get a dated entry here before they touch anything above.
First live rebalance. NAV tracking starts at $10,000, marked to market daily against SPY from the same start.
Fourteen years of walk-forward + purged CPCV locked in: 15.56% CAGR against SPY's 13.69%, after costs, on the point-in-time universe.
The whole book, in the open.
Every position, every rebalance, republished each month from the same code that runs the live book. No app, no login, just the page.