The rule that wrote this post
In June, when we started paper-trading a basket of strategies under pre-committed gates, we drafted the outline for this post and left ourselves one instruction at the top: do not write it early. The whole point is that the data finishes the post.
Every strategy entered its trial with the verdict criteria frozen in advance — profit-factor floors, drawdown lines, minimum trade counts, a calendar. No moving targets, no "let's give it another month" after seeing the numbers. The gates would speak, and we would transcribe.
The gates have spoken. Here is the transcript, in full — including the parts that flatter us and the parts that don't.
One graduate
The short-side intraday strategy — the one that fades morning euphoria — finished its trial at forty-nine days with a profit factor of 1.43 against a pre-committed floor of 1.3, on fifty-seven trades. It is the only strategy out of everything we have ever built to pass a live-forward trial end to end. It kept trading on paper while we built the plumbing for real money, sealed its record at sixty-six trades, and last week became the first strategy to trade our real account. Its paper record is archived, closed, and will never be blended with what comes next.
Two kills
The cascade strategy died at day sixty-one. Its profit factor came in at 1.00 — exactly break-even — against a floor of 1.3. Sixty days of real-time evidence saying "this makes no money after costs" is not a nuance. What made it instructive: extending the backtest that had originally justified it revealed the strong early numbers were a regime artifact. The paper trial didn't fail the strategy; it exposed that the strategy had never deserved the trial.
The momentum rotation died of silence. Four trades in sixty-five days — a cadence so far below its backtest's expectation that the trial couldn't even accumulate the sample its verdict needed. It was liquidated politely, slightly profitable, and killed for inactivity. A strategy that doesn't trade enough to be judged is not "safe." It's unfalsifiable, which is worse.
One extension — the verdict we didn't expect to like
The pairs-trading sleeve reached its sixty-day review this morning: six round trips, six wins, zero losses, zero drawdown. Spectacular, right? Here is what the gates said anyway: not enough evidence. Its profit-factor verdict is defined at twenty trades; six perfect ones have roughly a three-percent chance of happening by luck even if the strategy is merely mediocre. So it got neither a promotion nor a funeral — it got an extension to twenty trades, with a calendar backstop so the extension can't drift forever.
Writing "6-for-6, verdict: insufficient" felt almost rude. That feeling is exactly the thing the pre-committed gates exist to override. Small samples flatter; the gates don't.
Two clean records
The calendar strategy and the panic-fade strategy both completed their sixty-day windows without ever touching a gate — steady profits, zero drawdown against their kill lines. They now sit in the promotion queue behind the graduate, waiting their turn on a ladder where each rung must be earned by the one before it.
What sixty days actually bought
Count it: seven strategies entered serious trials. One graduated to real money. Two are dead with numbered headstones. One is in overtime. Two are queued. And one — this is the quiet lesson — was prevented from being deployed by nothing more than a rule that says verdicts come from evidence, not enthusiasm.
None of these outcomes required judgment in the moment. Every one was decided by criteria written by calmer versions of ourselves, months earlier, who didn't know the numbers yet. That's the entire technology. Patience isn't a virtue here; it's infrastructure — built once, in advance, precisely so that on verdict day nobody has to be wise.
The data finished the post, as promised. It just took until mid-September to file its copy.