Hypothetical backtest at a glance
What did Coil do in its published backtest?
In a hypothetical backtest from Sep 1, 2025 to Aug 31, 2026, Coil took 184 trades on MGC (1 MGC contract). Net of commission and one tick of slippage per contract, the hypothetical net profit was $3,670 and the maximum drawdown was −$972, with a profit factor of 1.40 and a win rate of 62.5%.
| Backtest window | Sep 1, 2025 – Aug 31, 2026 |
| Market | MGC · Gold low-volatility reversion |
| Size tested | 1 MGC contract |
| Trades | 184 |
| Hypothetical net profit | $3,670 |
| Max drawdown | −$972 |
| Profit factor | 1.40 |
| Win rate | 62.5% |
| Average trade | $19.95 |
| Largest win / loss | $118.48 / −$152.52 |
↓ The trade data behind these results: coil-trades.csv — every one of the 184 trades, losing trades included. Equity curve and monthly results: backtests. All 1,113 trades across 7 runs: trade data.
What it actually does
It fades an extreme that nobody showed up for.
Coil looks for gold to make a fresh short-term high or low on below-average volume — a push that the market did not really participate in — and takes the other side, expecting price to come back toward the middle. The volume condition is the whole idea: an extreme made on conviction is a trend, and an extreme made on nothing is a stretch.
It takes one trade a day, and it stops after a win. That is a deliberate ceiling. A reversion engine that keeps re-entering a market which has started trending is how a small loss becomes a bad day.
The honest structure: Coil risks more than it targets on any single trade. It has to win more often than it loses to make money at all. It is a grind by design — lots of small, unremarkable wins — and it will never produce the one enormous trade that makes a screenshot.
The honest two-sided ledger
Where it wins. Where it hurts.
Thrives in
- Quiet, rotational gold. Range with two-way rotation and no conviction is exactly the shape it is built for.
- Overextended pushes on thin volume. The lower the participation behind a new extreme, the better the fade tends to work.
- Sessions that go nowhere. The market conditions that bleed a breakout engine are the ones that feed this one.
Faces duress in
- Volatility expansion. A reversioner gets run over when range starts expanding. This is its #1 weakness, and it is the one thing we actively filter for.
- Strong one-way trends. Every new extreme is a fresh reason to fade, and every fade is on the wrong side of a market that keeps going.
- News spikes. A violent, one-directional move through the window is the worst case for taking the other side of an extreme.
The weakness is the point. Coil is the reversion counterweight in the book — it makes money in the quiet, directionless conditions where the breakout engines are bleeding.
The regime truth
The one filter in this whole book that survived out-of-sample testing.
We are skeptical of entry filters, and we say so on every one of these pages: they nearly always look brilliant on past data and then fail on data they have not seen. We have thrown out a lot of them.
Coil has the exception. Skipping entries while volatility is expanding held up under walk-forward testing — trained on one period, tested on the next, repeatedly. It is also the filter you would predict from first principles, which is why we trust it: a strategy that fades extremes should stand aside when the market starts to run. It costs a little gross profit and buys a materially better expectancy and a shallower drawdown. That trade we will take every time.
The flip side of that honesty: if gold settles into a long, dead, rangeless stretch there is not enough movement to fade either, and Coil grinds toward flat. Quiet is its friend; dead is not.
Risk & bail-out
Drawdown-first, with hard lines that don't move.
Fixed resting stop and target priced from the actual fill, not from a level we hoped to get. There is also a volatility-collapse exit: if the movement the trade depended on disappears, it takes what is there instead of waiting to be right.
One trade per day and stop-after-win is the real risk control. It caps how much any single session can cost, and it stops the engine arguing with a market that has changed character. Anything still open is closed out in the afternoon.
This is the steadiest, shallowest profile in the book — which is exactly why it is here.
How we catch decay before it’s too late
We monitor this like a model, not a hunch.
Every edge decays eventually — ours included. The difference is whether you find out from a dashboard or from your account balance. Coil runs inside a monitoring loop that compares every live trade against its learned baseline and raises a flag the moment behavior drifts.
The promise. We would rather pull Coil back a month early than a day too late. When the data says the edge is thinning, we trim or bench it — publicly — instead of hoping it comes back.