Most trading products sell hope.
We publish our homework.
Everything in SentryQ — every strategy, every default parameter, every risk limit — exists because it survived a validation process designed to kill it. This page explains that process, including the parts that are usually hidden: the failures, the drawdowns, and the limits of backtesting itself.
Designed to kill bad ideas
- Split-sample discipline. Strategies are tuned on one span of history, then must prove themselves on a separate, untouched span. If the edge doesn't survive data it has never seen, it doesn't ship — however good the first half looked.
- Real track-record anchoring. Our simulator was calibrated against genuine broker statements — thousands of real fills — so we know exactly how optimistic backtests run, and we discount for it before drawing any conclusion.
- Cost honesty. Every test pays realistic spread and commission. Several "edges" we investigated were profitable until costs were applied — those are exactly the mirages this process exists to catch.
- Survival simulation. Sizing guidance comes from thousands of simulated account paths, not a single lucky equity curve — including explicit measurement of the chance of ruin at each risk level.
Drawdown: the price of every return
Profit is what a strategy gives you. Drawdown is what it demands from you first — and the arithmetic of recovering from losses is brutally asymmetric. This is why we state expected drawdowns before we mention returns, and why the Equity Guardian exists.
| Drawdown | Gain needed to recover |
|---|---|
| −10% | +11% |
| −25% | +33% |
| −50% | +100% |
| −80% | +400% |
Every strategy page states its drawdown character in plain words — "routine ~20%, occasional ~35%" — so the first rough patch is something you expected, not a shock that shakes you out at the worst moment. Losing streaks are a statistical certainty in any real strategy; knowing their expected shape in advance is what makes them survivable. What drawdown you should actually expect →
What didn't survive — and why we tell you
A strategy library you can trust is defined as much by what's absent as what's present. Three researched candidates failed validation and were rejected:
Rejected by the evidence
- Fibonacci retracements — 144 configurations across two distinct designs; zero profitable out-of-sample after costs
- Turtle-style daily breakouts — deeply unprofitable on modern FX majors across our whole window
- Daily pullback swing — every tested configuration net-negative after costs
Why publishing failures matters
- It proves the validation gate is real — things actually fail it
- It tells you a "missing" popular strategy is a decision, not an oversight
- It sets the standard for everything we ship next: same gate, no exceptions
Fibonacci retracement is the tool traders ask us for most. We built it two different ways and tested 144 configurations across our whole history. After realistic spread and commission, not one was profitable out-of-sample. That is why it isn't in the library — not an oversight, a result. The most-requested feature and the most thoroughly disproven, documented here rather than quietly skipped.
Four strategies we shipped — then took back
Rejecting an idea before launch is easy; nobody has seen it. Pulling a strategy that is already live — already in the briefs, already something you could switch on — is the real test of whether an evidence standard means anything. In July 2026 we retired four, for two different kinds of reason.
| Strategy | Why it was withdrawn |
|---|---|
| Session Fade | A lookahead error had flattered its validation. A profit factor of 1.98 collapsed to 0.77 once the simulation could no longer see a bar's close before that bar had formed. |
| Regime Breakout | The same class of error. The honest, causal signal scored 0.59 out-of-sample — and live results had been quietly saying the same thing at 0.40. |
| News Momentum | No measurement error — simply no edge. Its best out-of-sample variant reached 1.05, indistinguishable from breakeven, with most variants below 1.0. |
| Momentum Rejoin | Also measured correctly. Tested against random entries held the same length of time, it scored p = 0.102 when it shipped and p = 0.158 a month later — drifting away from proof, not toward it. |
Two were caught by fixing how we measure; two were honestly measured and never earned their place — the less dramatic, more common reason to cut something. No position in any of them was dumped: each was managed out to its stop, target and time-stop exactly as before, and each keeps its label so your history stays attributed correctly. The full post-mortems, every figure, are on the strategy library; the lookahead story is worked through in our research notes.
What we will never promise
- No profit guarantees. Backtests are evidence about the past, not prophecy. Market regimes change; every strategy can and will have losing periods.
- No "passive income" language. This is trading. It carries real risk of loss, and it deserves your attention even when automated.
- No pressure into live money. The trial is demo-only by design, and live trading arrives as a deliberate, separate step.
- No hidden numbers. Wherever we cite a figure — win rates, drawdowns, rejected configs — it comes from the validation work described on this page.
Our research notes work through the figures on this site — why a 94% win rate tells you almost nothing, and the exact bug that turned a 1.98 profit factor into 0.77.
Judge us by the evidence.
Fourteen days on a demo account — every claim on this page, testable by you, for free.
Start your free trial