Why the same approach works at one trade a day and fails at twenty

Every round trip pays a toll

Opening and closing a position costs something regardless of the result. You cross the spread, or pay a commission, or both. You accept some slippage on entry and usually more on exit, particularly when the exit is a stop. If the position is held overnight you pay financing. None of these depends on being right. They are the entry fee for having an opinion at all.

Stated that way it sounds trivial, and in isolation it is. The problem is that this fee is charged with perfect reliability while your edge is not. Across a hundred trades, cost arrives a hundred times with very little variance. The edge arrives in an uneven scatter, sometimes clustered, sometimes absent for long stretches. Comparing their averages makes them look like comparable quantities. They are not. One is a bill and the other is a hope with a positive expectation attached, and the bill is settled first.

Put edge and cost in the same unit

The only way to make the comparison honestly is to express both in the same unit. Pick one, usually dollars per ounce for gold, or cash for your standard position size. Then write down two quantities for your own trading rather than from an article.

• Gross edge per trade: the average result of your trades measured before costs, in that unit.
• Round trip cost: spread plus commission plus the slippage you actually receive on both sides, plus financing if the position is held, in the same unit.

Net edge is the first minus the second. The sign of that difference decides whether the approach is a business or an expensive hobby, and the sign is not obvious from a profitable looking equity curve, because a curve built from gross results flatters you. Comparing the two numbers only in aggregate also hides things. Gross edge is usually distributed unevenly across setup types, times of day and market conditions, while cost is roughly constant or worse in the conditions you find most exciting. Breaking both down by category is where the uncomfortable discoveries live.

Frequency multiplies cost and dilutes edge

Doubling the number of trades does two things, and only one of them is linear. Cost doubles, exactly and predictably. Gross edge does not double, because the additional trades are not drawn from the same quality distribution as the original ones. They are the marginal setups, the ones that did not quite meet the criteria, taken because the screen was open and the previous trade was frustrating.

So the net result follows a shape rather than a straight line. Up to a point, more trades mean more total profit, because each one carries a positive net edge. Beyond that point the marginal trade has a gross edge smaller than the round trip cost, and every one you add subtracts from the total. The curve rises, flattens and then falls. The figure shows that shape with no numbers on it, because the position of the turning point is specific to your costs, your criteria and your market. What is general is that the turning point exists, and that a trader who feels busy is usually past it.

Cost as a share of the target is the real constraint

Here is the part that decides whether a style is viable at all. Cost is roughly fixed per trade, so what matters is cost as a proportion of the distance you are trying to capture. A target measured in a few dollars per ounce makes the round trip cost a large share of the move. A target measured in tens of dollars per ounce makes the same cost a minor deduction.

This is why short horizon trading is structurally harder rather than merely faster. The trader is often not wrong about the setup, they are attempting to extract a distance that is small relative to the toll. To make it work, something has to change: a lower cost structure, a higher strike rate, or a wider target. Two of those three are not within your gift. The third is, which is why the relationship between target distance and stop distance is a cost question as much as a probability question. Before blaming execution, check whether the arithmetic of the chosen distance ever permitted a profit in the first place.

More trades reveal an edge, they do not create one

There is a legitimate argument for more trades, and it deserves stating properly before it is dismissed. A larger sample reduces the influence of luck on your measured results. With few trades, a good month tells you almost nothing about whether the approach works. With many, the average converges toward whatever the true expectation is, and you learn faster.

The catch is in that last clause. Convergence toward the true expectation is good news only if the expectation is positive. If it is negative, frequency delivers the bad news efficiently and expensively. Frequency is a measuring instrument rather than a source of return. It reveals the edge you have, it does not manufacture one. This is also where a familiar confusion creeps in: a conviction rating on a setup is not a measured frequency of success, and treating a score as though it were a probability encourages exactly the extra trades that push you past the turning point.

The cost lines people leave out

When traders total their costs they usually count spread and commission and stop there. The full list is longer.

• Slippage, which is larger on exits than on entries because exits are more often forced.
• Financing on anything held past rollover, repeated nightly.
• Conversion charges when your account currency differs from the settlement currency.
• Platform, data and subscription charges, which are fixed and therefore weigh more heavily on a small account.
• Error cost, the mistakes made when tired or annoyed, which correlate strongly with high frequency.
• Opportunity cost, the better setup missed because capital and attention were already committed.

The last two resist measurement and are not therefore zero. The figure stacks the countable items against a gross edge bar to show the proportion, unlabelled because your proportions are your own. The exercise is worth doing on paper once: take a month of trades, total every charge, and compare it with gross profit. Most people find the ratio considerably worse than they assumed, and that discovery changes behaviour more effectively than any amount of advice.

Why simulated results flatter frequency

Simulated results are optimistic about frequency in specific, predictable ways. A fixed cost assumption understates what you pay in volatile conditions, which is when many signals occur. Filling at the mid price ignores the spread entirely. Assuming a limit order fills whenever price touches its level ignores that in a fast move it often does not, and that when it does the move is usually continuing against you. Testing on a single symbol and a single period makes a cost structure look stable when it is not.

The effect compounds with frequency. A small per trade optimism is invisible at ten trades and decisive at a thousand. This is why a high frequency approach can look excellent in simulation and fail immediately in practice, while a slower approach degrades only mildly. The sensible test is to rerun the model with costs deliberately set worse than you believe they are and see whether the conclusion survives, which is one of the items on the backtest honesty checklist. If an approach only works at optimistic costs, you have measured the cost assumption rather than the approach.

What actually follows from the arithmetic

Three practical consequences follow, and one caveat that matters as much as they do.

First, raise the bar for what counts as a setup, since removing marginal trades removes cost without removing much gross edge. Second, prefer targets large enough that the round trip cost is a modest share of the move, which usually means being more patient about entries rather than more aggressive about them. Third, measure your real costs by category instead of assuming an average, because the average conceals the expensive conditions.

The caveat is important. Fewer trades is not automatically better. A trader who cuts frequency but keeps the same loose criteria is simply taking a smaller random sample of the same mediocre population, which reduces costs and results together while making the outcome noisier and harder to evaluate. Selectivity only helps if the trades you remove are genuinely the worse ones, which is the real argument in why fewer trades often work better and in knowing when to sit out. Deciding in advance what you will not trade, and watching the rest go past on the live chart without acting, is the version of this that holds up under pressure.