What is an ATR stop?
An ATR stop is a stop loss placed at the entry price plus or minus a multiple of the average true range, the average over a stated number of bars of each bar's full range including any gap from the previous close.
The point of scaling to ATR is consistency across conditions. A twenty-pip stop is generous on a quiet Asian morning and absurdly tight during a volatile afternoon; a stop of one and a half times ATR is roughly the same distance relative to the market's current movement in both. The rule then means the same thing on every day and, with the same multiple, on every instrument.
The ATR period is a parameter, and a short one tracks recent conditions closely while a long one smooths them. The multiple is the parameter that matters more. Together with the target, it sets the geometry of every trade.
Because both are numbers, the stop can be examined across a grid, which is more than can be said for a stop placed by eye below the last swing.
Why does the multiple change everything at once?
The multiple changes the win rate, the size of each loss, the size of each win in R, and the number of trades that reach any given target, all together, because it sets the distance that ordinary noise has to travel to end a trade.
A tight multiple is hit by noise. Trades that would have gone on to reach the target are stopped by a pullback that was inside the market's normal movement, and the win rate falls. Each loss is small in price terms, but because the R-multiple of every trade is measured against the stop distance, the wins get larger in R while becoming rarer.
A wide multiple survives the noise and is hit only by genuine adverse moves. The win rate rises, each loss is larger in price, and the same target is a smaller number of R because the stop it is measured against is bigger. The strategy's expectancy in R can move in either direction, and which direction depends on how price tends to move after entry on that instrument.
This is why the multiple cannot be reasoned to. It has to be measured, and the measurement has to be in R after costs, because a tight stop makes the spread a large fraction of every trade.
How does the stop interact with the target?
The stop and the target together fix the reward-to-risk ratio in ATR units, and a change to either changes the win rate needed for the strategy to break even, so the two have to be examined as a pair rather than one at a time.
A target of three ATR against a stop of one is a three-to-one trade that needs to win more than a quarter of the time after costs. The same target against a stop of two is a one-and-a-half-to-one trade that needs to win more than forty percent. Whether either is achievable is a property of how far price travels after entry on that rule, which is what the distribution of maximum favourable excursion describes.
The cost term enters through the stop. A round-trip spread and slippage that is a tenth of an ATR is a tenth of an R on a one-ATR stop and a twentieth of an R on a two-ATR stop. Tight stops pay proportionally more to the market on every trade, and on lower timeframes that cost can consume the entire edge.
The honest picture is a surface across both parameters, in R after costs, with the trade count in each cell, examined for a stable region rather than for its brightest cell.
How do you test the multiple honestly?
Test the multiple by running the same entries across a range of stop and target multiples, grading each combination in R after realistic costs, resolving the entry bar honestly when the stop is close, and confirming the chosen region on held-back history.
The entry bar is the trap specific to tight stops. When the stop is a fraction of an ATR from the entry, it frequently falls inside the bar that contains the entry, and so does the target on a good trade. A tester that assumes the target was reached first flatters every tight-stop rule; a tester that assumes the stop was reached first punishes it. Only finer data settles the order, and where there is none the conservative reading is the loss.
Costs have to be in the same units. A spread stated in pips has to become a fraction of the stop distance for each trade, so that the surface shows what is left after the market's cut rather than before it.
The confirmation step is the same as for any parameter. Draw the surface on the in-sample segment, choose the centre of the stable region, and grade that choice once on the out-of-sample segment.
How does QuantParadox place ATR stops?
QuantParadox states stops and targets as multiples of the average true range on the entry timeframe, resolves any bar containing both the entry and its stop with finer data where the archive has it, and books the loss where the finer data does not settle the order.
That resolution rule is the property that makes tight-stop results on the platform believable rather than flattering. A trade whose stop and target were both inside one candle is not assumed to have won; it is checked against the minute record, and if the minute record is itself ambiguous the trade is labelled as such and counted against the strategy.
The consequence is that some rules look worse on the platform than on a tester that guesses in the trader's favour, and that difference is the point. A stop rule that only works when the tester is generous does not work.
The stop-and-target grid is a set of runs of the same rule, and the selection among them is counted by the platform's variant ledger, so the best cell is graded as the best of that many rather than as a single test.