The Desk: Gates Before Paper

The Desk is ThoroughWiz's paper-trading research desk for US stocks and crypto. It runs on a $100 model book — paper capital, simulated fills, no broker account connected, no real money traded. It is not investment advice. Every strategy on it is labeled UNPROVEN until it clears the Desk's gates and then survives its paper period. This guide explains the gates, the graveyard they feed, and the three strategies that start trading on paper on Monday, November 9, 2026.

The gates come first

An idea earns nothing by sounding right. The Desk turns it into an exact rule — entry, exit, stop, sizing — and backtests it on data the rule has never seen (out-of-sample), with fees and slippage counted up front. An edge that dies after costs is dead. The surviving candidate is scored 1–5 on five lines:

• out-of-sample Sharpe ratio,
• deflated Sharpe — the same figure corrected for how many ideas were tested in the batch,
• maximum drawdown,
• trade count, against a floor of 30 out-of-sample trades,
• edge after costs.

The keep line is 18 of 25. "Profitable" has a definition here: positive after fees and slippage, on unseen data, across at least 30 trades. Since November 8, 2026 (gate rubric v1.1), new candidates also face two pass/fail criteria — regime coverage (the strategy is positive in every market regime in its window, or carries an explicit rule that switches it off where it fails) and multiple-testing honesty (its Sharpe survives the deflation correction at the batch's true trial count). Score below 3 on either and the candidate fails outright, whatever its total. Before any strategy is presented as a keeper it must also survive being entered one bar late with double slippage, answer for 2020 and 2022, and end its report with a section titled WHY THIS COULD FAIL.

The graveyard

Every idea the Desk tests and does not approve is buried in a public graveyard file, with its hypothesis, its test result, and its cause of death. No new test begins until the graveyard has been searched — a rejected idea does not come back under a new name unless its record states the genuinely new element that takes it out of the buried design.

On the Desk's first night (November 7, 2026), ten ideas were tested. The graveyard from that one batch holds, among others: an oversold snap-back in crypto, killed because a roughly 62-basis-point round trip ate the effect it was chasing; a crypto weekend effect whose measured edge was smaller than its trading costs; an earnings-drift idea buried on count — 19 out-of-sample trades against the 30-trade floor; and the original form of the sector rotation below, vetoed on its sizing before evidence was even the question.

What trades on paper — from Monday, November 9

Three strategies cleared the gate as it stood on November 7 (they are grandfathered under rubric v1.1, which binds the next candidate). Each trades from a versioned execution spec — the exact entry, exit, stop and sizing, the cost assumption it was tested under, and the drawdown and expectancy figures its pause rules measure against. Execution trades the spec exactly; any change is a new version, and a new version goes back through the gate. The specs are up for review on November 23 and expire December 7.

1. The 20-day Donchian breakout — scored 22/25

Rule: when a stock closes above its prior 20-day high on volume at least 1.5 times its 20-day average, buy at the next session's open. The stop starts at the entry price minus twice the 14-day average true range and trails upward. Exit on a close below the prior 10-day low. Out-of-sample record: 264 trades, +$49.75 after costs, Sharpe 1.27.

2. The oversold snap-back — scored 22/25

Rule: when a stock has fallen 8% or more in three days, its 2-day RSI sits at 10 or below, and its price is still above its 200-day average, buy at the next session's open. Exit when the price closes back at its 5-day average, or after 10 sessions, whichever comes first. The stop is twice the 14-day average true range and does not trail. Out-of-sample record: 212 trades, +$20.91 after costs, Sharpe 1.10.

3. Sector rotation, capped variant — scored 20/25

Rule: each Monday, rank the nine sector ETFs by 63-day relative strength against SPY and hold the top four, at $25 per leg. A sector that falls out of the top four is sold at the next Monday's open — the rotation is the exit; there is no per-trade stop, exactly as tested. Out-of-sample record (retest of the capped form): 99 trades, +$32.34 after costs, Sharpe 0.90.

The strategy as first designed held the top two sectors at $50 a leg. That form was vetoed and never trades: two $50 positions break the Desk's 25% position cap on a $100 book. Only the capped variant — top four at $25 — was retested, and only it is approved. No crypto strategy is approved; the crypto ideas from the first batch sit in the graveyard.

Risk rules that can stop the Desk

• Per trade: maximum planned loss is 1% of the book — $1. No position may exceed 25% of the book — $25.
• Desk-wide kill switch — any one trips it and trading halts: a daily paper loss of 2% ($2); three losses in a row; or stale data — a feed older than its stated freshness, or a price the Desk cannot source.
• Per-strategy pause rules: a strategy pauses when its paper drawdown passes 1.5 times its backtest drawdown; when 40 of its paper trades complete net-negative against a positive backtest expectancy; or when its measured cost per trade runs above twice the backtest assumption. A pause is journaled the day it fires, open positions are managed to their spec exits, and a strategy resumes only through a written review note — never silently.

The ladder stops at paper

The promotion ladder has four rungs: backtest keeper, paper trading, then — after two weeks of paper results matching the backtest — eligibility for the next rung. Live trading, at any size, is not on the ladder the Desk can climb by itself: it requires the Lab Owner's separate, explicit word for that step, and nothing in a strategy's score pre-authorizes it. Until that word exists, the ladder stops at paper, and the Desk's whole record — briefs, journal, vetoes, pauses — publishes on the Stocks page.