A Reader Sent Us Their Camarilla Reversal Strategy. We Backtested It — Exactly As Written.

This one came from the community. A reader sent us their complete ruleset — genuinely complete, which almost never happens — and asked what the data says. Here is the strategy, condensed but faithful:

MES & MNQ only. Daily candles above the 21 EMA → longs only; below → shorts only. Plot Camarilla pivots, use only S3/S4/R3/R4. Reversals only: for longs, price taps S3 or S4 (or closes slightly past), then a reversal confirmation candle (doji, engulfing, double top/bottom, hammer, shooting star) → enter at the close of that candle. Stop just below the reversal candle. Always 1:3 RRR. Shorts mirror at R3/R4. Trade only 9–11 AM ET. Day is over when there’s no setup, after 3 trades, or the moment a take-profit hits. All or nothing — “even if you are ONE TICK AWAY, DO NOT TOUCH.”

That’s a spec. We test specs. So: 7 years of NQ tick data (MNQ trades the same prices; MES we can’t test — our ES history only starts mid-2026, disclosed), micro economics ($2/pt, $2 round-trip including slippage), management on 1-minute bars, stop-checked-before-target, flat at the close. The submission doesn’t name the confirmation-candle timeframe, so we ran both common readings — 5-minute and 15-minute — and every pattern in the list is implemented (doji, engulfing, hammer, shooting star, double top/bottom). One stat judges everything: t — how far the average day sits from zero relative to its own noise; |t| > 2 is real.

The cleanest zero we’ve ever measured

914 trades. 25.7% win rate. Net +$144 over seven years. t = +0.07.

Not a money-printer, not a account-burner: a coin flip with excellent manners. The 15-minute reading is the same story with fewer trades — +$1,025 over 7 years (≈ $146/year per contract, t = 0.4). No year stands out in either direction.

Why it HAS to be zero: the 1:3 arithmetic

Here’s the part worth internalizing, because it applies to every “always take 1:X” rule you’ll ever hear.

At a fixed 1:3 reward-to-risk, the breakeven win rate is exactly 25% (before costs). The strategy wins 25.7% of the time. That is not a coincidence — it’s what entries with no informational content look like when you strap a 1:3 bracket on them. The RRR decides your win rate; it cannot decide your expectancy. If the entry is random, every bracket geometry converges to the same place: zero, minus costs.

The placebos complete the picture. Same machine, same candles, same management — but Camarilla levels computed with wrong multipliers (0.7 and 1.5 instead of the canonical 1.1), or from two-day-old stale OHLC: all of them sit in the same zero-to-slightly-negative band, win rates 23–25.3%. If the real S3/S4/R3/R4 prices carried information, they’d have to separate from the fakes. They don’t. Even dropping the confirmation candle entirely (enter on any tap) changes nothing (t = 0.4) — and inverting the 21-EMA filter, which should turn a real edge into a reliable loser, produces only noise-grade bleed (t = −1.3).

What the submission gets right

We test a lot of strategies, and most lose money reliably. This one doesn’t — and that’s not luck, it’s the risk discipline doing exactly what risk discipline does:

• Fixed, known risk per trade (stop under the reversal candle) — no blowup tail.
• Three-trade daily cap and “quit after a win” — hard ceiling on daily damage and on overtrading.
• A defined session (9–11 ET) — the two liquid morning hours, where spreads and slippage are smallest.
• No partials, no averaging, no discretion — the reason we could even test it.

That discipline is why the equity curve is flat instead of bleeding. But discipline is a cost cap, not an edge. What’s missing is the part that actually decides expectancy: an entry that knows something about the next hour. The reversal-candle-at-pivot entry, measured against every placebo we could throw at it, doesn’t.

This is the same conclusion our level-family research keeps reaching from the other direction: on NQ, marked levels — volume-derived, pivot-derived, or arithmetic like Camarilla — get absorbed. The mechanical edges we’ve validated live elsewhere: trend-pullback continuation, stretched-move mean reversion, compression breakouts.

Send us yours

This is what we want more of: complete rules, honestly stated, testable. If you have a strategy — yours or your favorite guru’s — send the full, unfiltered ruleset through the strategy submission form. We’ll mechanize it exactly as written, run it on 7 years of real tick data with honest fills, real costs and placebo controls, and publish what comes out — whatever it shows.

Methodology: NQ front-month tick store 2019–2026 (MNQ trades identical prices; MES untestable — our ES recording starts 2026-06). Camarilla from prior full Globex-day OHLC (RTH-range variant tested, same result). Daily 21-EMA on Globex daily closes, evaluated on the prior day’s close (no lookahead). Entries at confirmation-candle close 9:00–11:00 ET; management on 1-minute bars, stop-before-target; flat 16:00 ET. $2/pt, $2.00 RT incl. slippage per MNQ. This is research, not trading advice; our own live, disclosed track records are on /algos.