Two programs, one discipline
MarketFactor runs two research programs against the same market, on the same curated database, under the same evidence gates. One builds a book. The other tries to destroy its own findings before they ever reach it. Both trade on paper. Neither is a recommendation.
The HedgeFund Program
A multi-sleeve book rather than a strategy. Thirteen sleeves run side by side, each with its own narrow mandate — trend participation, relative strength, defensive rotation, income structures, and hedges — and each with its own budget. No sleeve can size to conviction; it sizes to its mandate. The book is allowed to be long and short at the same time, because the sleeves are answering different questions.
What decides a position is not a signal on its own. Every sleeve proposal passes through the same sequence:
- A registry gate — a sleeve that has not earned a state cannot reach the order path at all. States are promoted one step at a time, on evidence, and the final step to real capital is a human decision that no model can make for itself.
- Broker truth, not bookkeeping — target sizes are computed against what the account actually holds, so the book can never over-buy on a stale internal record.
- An exit harness that owns the exit — entries and exits are deliberately separated. Winners are left alone to run; losses are cut by a trailing rule, not by a change of opinion.
- A reconciler that can halt everything — if the book and the broker disagree, or the equity feed is stale, the program refuses to trade blind rather than guessing.
Current paper stance — 21 positions, 6 long and 15 short, across 13 sleeves. A net-defensive posture. Position counts and stance are published here; the names and the per-position reasoning are the subscriber product.
The Quant Program
The Quant pipeline is not a search for what worked. It is a machine for killing things that merely look like they worked. Anyone can find a rule that made money in the past; the entire difficulty is telling that apart from luck, and almost nothing survives being asked properly.
Our most recent full run, end to end:
| Backtests run | 4,386 |
| Survived the robustness screen | 650 |
| Survived multi-fold walk-forward | 289 |
| Survived as statistically independent | 65 |
| Passed every gate | 0 |
| Held on watch | 9 |
We publish the zero on purpose. Sixty-five survivors sounds like a discovery until you measure it against chance: across the full trial count those 65 are only 1.88× what randomness alone would have produced, and an earlier version of our own folds had been overstating survivorship by 4.4× until we corrected it. A vendor with a marketing department sells you the 65. We would rather show you why we did not.
The gates are named here; their thresholds are not:
- Deflated Sharpe against the real number of trials attempted — not a flattering subset
- Probability of backtest overfitting, measured across cohorts rather than a single split
- Purged walk-forward folds, so a rule never sees data adjacent to what it is tested on
- Cost and slippage stress — commission, spread and market impact, then stressed again
- Minimum trade and exposure floors — a rule with four trades has not demonstrated anything
- A point-in-time universe, so a rule cannot quietly trade on membership it would not have known about
Persisted Edges is the nine that carried out of sample. It is published as a persistence diagnostic, not an alpha claim — a list of what kept working, with the honest note that none of them cleared the full statistical bar.
How the two connect
The Quant program does not trade. It produces candidates, and the registry gate decides whether any of them are allowed near the HedgeFund order path. Nothing skips a state, nothing is promoted without evidence on the record, and nothing reaches real capital without a person deciding.
Paper trading — and the live account
All automation runs in a paper account, opened 7 July 2026 at approximately $1.0 million notional. No customer capital is managed, pooled or solicited, and no automated system of ours places an order with real money.
Disclosure of positions. Over time, one or more of the positions discussed in this research will at times correspond to real positions held in the personal margin account of the publisher's principal. We will say so when that is the case.
As of 11 September 2026, that live portfolio is invested entirely in SGOV, a short-term U.S. Treasury ETF. There are currently no equity, short or options positions in it that correspond to anything shown here.
On publishing performance
We are not going to print a hypothetical return number on a public page before our disclosure language has cleared counsel. Paper results carry real limitations — no live fills, no true slippage, no emotional cost — and a figure without those caveats attached is marketing, not evidence. When the scorecard publishes, it will publish with the misses in it and the limitations beside it. We would rather be late than be the kind of shop that advertises the number first and explains it afterwards.
Nothing on this page is investment advice, a recommendation, an offer, or a solicitation. Hypothetical and paper results have inherent limitations and do not represent actual trading. Past performance and hypothetical results do not guarantee future results.