Risk terms

Correlation risk

Correlation risk is the exposure created when several open positions move together, so that what looks like diversification is in fact one larger bet.

Three positions each risking 1% look like 3% total risk only if they are independent. If they are all effectively long the same driver, the real exposure is closer to a single 3% position, and it will be resolved by a single event.

Gold correlates strongly and inversely with the dollar and with real yields, so a gold position and a short-dollar position are largely the same trade. Traders discover this during the move that closes both simultaneously.

For automated systems the problem takes a specific form: multiple strategies firing in the same direction on the same symbol within minutes of each other. We found this in our own live results — same-strategy stacking accounted for three of ten trades in one month, carrying roughly half the printed profit and an outsized share of the risk. A concurrent-position cap is the fix, and it is why ours is enforced rather than advisory.

Drawdown

Drawdown is the decline from a peak in account equity to the subsequent trough, expressed as a percentage, and it is the truest measure of what a strategy costs to hold.

Drawdown measures pain rather than performance. Two systems can return the same amount over a year while one drifts down 8% at worst and the other falls 40% before recovering, and no reasonable person would consider those equivalent.

The recovery arithmetic is asymmetric and unforgiving. A 20% drawdown requires a 25% gain to recover, 50% requires 100%, and 80% requires 400%. This is why controlling drawdown matters more than maximising return — the deeper the hole, the less plausible the climb out.

Our own validated figures are 27.4% maximum drawdown over a January-to-August 2026 in-sample period and 13.2% over an August-to-December 2025 out-of-sample one. We publish both because a single-period drawdown figure is close to meaningless, and because roughly one in five is a number a buyer deserves to see before purchase rather than after — the risk disclosure states it too.

How far gold itself has fallen from its peaks, and how long it took to get back, is measured in every gold price crash since 1971.

Equity curve

Also called: Equity curve trading, Trading the equity curve, Equity curve filter, Equity curves.

An equity curve is a chart of an account's value over time or trade by trade. It shows the path a strategy took to its result: how steadily it rose, how far it fell and how long it stayed below its high, which a single profit figure hides.

Read the full entry: Equity curve →

Grid trading

Grid trading places a ladder of orders at fixed intervals above and below price, accumulating positions as the market moves against the initial entry.

In a ranging market a grid harvests small profits repeatedly as price oscillates through the levels, and the results look outstanding. The mechanism is genuinely profitable in the conditions it was designed for.

In a trending market it accumulates an ever-larger losing position in the wrong direction, with no stop, and the account is destroyed by exactly the market behaviour that every other strategy would have profited from. The failure is not a bug in a particular implementation; it is the design.

Grid and martingale are frequently combined, which compounds both problems. The distinguishing question to ask about any such system is simple: what is the maximum adverse excursion it can survive, and at what account size. Systems that cannot answer that are not managing risk, they are deferring it.

Hedging

Hedging is holding offsetting positions in the same or correlated instruments so that a move in one is compensated by the other.

A direct hedge — long and short the same symbol simultaneously — locks the current profit or loss in place while both remain open. What it does not do is improve the position, and it continues to pay spread and swap on both legs while achieving nothing.

Traders typically open one to avoid closing a loser, which converts a decision into a delay. The eventual unwind requires choosing which leg to close and when, which is the same directional decision that was being avoided, now made under worse conditions.

Regulatory treatment varies — US retail accounts operate under FIFO rules that prevent direct hedging entirely, while many other jurisdictions permit it. Automated systems written for hedging accounts frequently misbehave on netting accounts for this reason, which is worth checking before running any EA on an unfamiliar broker.

Kelly criterion

The Kelly criterion is a formula giving the position size that maximises long-run growth rate for a known edge, derived from win probability and payoff ratio.

For a binary outcome the fraction is the win probability minus the ratio of losing to winning payoff. It is mathematically optimal for growth given exact knowledge of the edge, which is the assumption that makes it dangerous in trading.

Full Kelly produces enormous position sizes and correspondingly enormous drawdowns — swings of 50% or more are normal at the theoretically optimal fraction. More importantly, the formula assumes the edge is known exactly, and a trading edge is estimated from a sample and changes over time.

Practitioners who use it at all use a fraction of it, commonly a quarter or less, and treat the output as an upper bound rather than a target. Our fixed 1.5% is far below any Kelly estimate our numbers would produce, which is deliberate: the estimate is built on roughly ten trades per month, and sizing aggressively against a number that uncertain is how a working system produces a blown account.

Martingale

Martingale is a position-sizing method that increases size after a loss, on the reasoning that a win will eventually recover the entire sequence at once.

Doubling after every loss guarantees that the first win recovers everything plus one unit — provided the account can fund the sequence. Ten consecutive losses require a position 1,024 times the original, which no retail account can support.

This is why martingale systems produce beautiful equity curves for months and then a single vertical drop to zero. The curve is not evidence the method works; it is evidence the terminal sequence has not arrived yet. Every martingale account eventually meets one.

It is worth naming plainly because a large share of the EAs sold on marketplaces are martingale or grid systems with the mechanism described in euphemisms — "smart recovery", "averaging", "no stop loss needed". If a product cannot show you its stop loss, it is telling you where the risk went. Ours risks a fixed percentage and never increases after a loss.

Maximum drawdown

Also called: MDD, Max drawdown, Maximum drawdown definition, Max drawdown prop firm, What is max drawdown in prop firm.

Maximum drawdown (MDD) is the largest fall from a peak in account value to the lowest point that follows, usually given as a percentage of the peak. At a prop firm, "max drawdown" also names the rule for the most an account may lose before it is closed.

Read the full entry: Maximum drawdown →

Risk of ruin

Risk of ruin is the probability that an account falls below a defined threshold given a strategy's win rate, payoff ratio and risk per trade.

It converts three familiar inputs into the number that actually matters: the chance the account does not survive long enough for the edge to appear. A system with a genuine positive edge can still be ruined by sizing, and risk of ruin is the calculation that shows where that line is.

The relationship with position size is steeply non-linear. Halving risk per trade does not halve risk of ruin — it typically collapses it by an order of magnitude, which is why the difference between 1% and 3% risk is not three times more aggressive but something closer to categorically different.

The risk of ruin calculator makes this concrete for your own numbers. The result is usually uncomfortable at the sizes most retail traders actually use, and that discomfort is the point of running it before rather than after.

Risk per trade

Risk per trade is the fraction of account equity a single position is allowed to lose if its stop is hit, and it is the setting that determines survival more than any other.

It is expressed as a percentage rather than a fixed amount so that position sizes shrink automatically during a losing run and grow during a winning one. That property alone prevents the most common way accounts are destroyed: increasing size after losses to recover them.

The conventional range is one to two percent, and the arithmetic behind it is worth internalising. At 2% risk, ten consecutive losses cost roughly 18% of the account; at 5%, the same run costs 40%. Ten consecutive losses are not unusual for a system with a 45% win rate — they should be expected within a few hundred trades.

Our engine defaults to 1.5%, scaled down to about 1.0% in mixed and choppy conditions. Both figures assume the stop is respected, which the broker minimum lot can silently prevent on small accounts — the configured percentage is a ceiling the platform can override upward, never downward.

Risk-reward ratio

Also called: R:R, Reward to risk.

The risk-reward ratio compares the distance to a trade's take profit against the distance to its stop loss, expressed as a multiple such as 2:1.

A 2:1 ratio means the target is twice as far away as the stop, so one winner covers two losers. This converts directly into the break-even win rate: at 2:1 you need to be right about 34% of the time to break even before costs, at 1:1 you need 50%, and at 1:2 you need 67%.

Higher ratios are not automatically better because reaching a distant target is less likely than reaching a close one. Pushing the ratio up mechanically lowers the win rate, and past some point the loss in hit rate outweighs the gain in size. The optimum depends on the strategy and cannot be assumed.

We fix it at 2.0 across every regime. An earlier version scaled the ratio by regime — 1.6 in mixed conditions, higher in trends — and validating that against an out-of-sample period showed it was fitting noise. Reverting to a single value across the board is one of the few changes we made that improved results by removing a feature. See expectancy for why the ratio alone is not the whole picture.

Whether a bigger target actually pays on gold is measured in does the risk-reward ratio actually work?