Risk and position size
Why fees and spread decide whether a small edge exists at all
Costs are not a rounding error subtracted at the end. They are a fixed toll on every round trip, and the smaller the move you are trying to catch, the larger the share of it the toll takes. That ratio decides whether a plan can work before any analysis.
What a round trip costs
Opening and closing one position crosses the market twice and pays commission twice. There are three components and they are all measured as percentages of notional, which makes them directly comparable with the move you expect.
- Commission - a fee per side. A taker fee if your order crosses the spread and executes immediately; a maker fee if it rests in the book and waits. Maker fees are lower, sometimes zero or negative, and the price of that is that the order may not fill.
- The spread - the gap between the best bid and the best ask. If you buy at the ask and later sell at the bid you have paid the full spread once over the round trip, and nothing on your statement will call it a cost.
- Funding - on a perpetual contract, a periodic payment between longs and shorts, charged on notional. Zero if you are flat by the next payment, and otherwise a cost of time rather than of trading.
Add the first two and you have the number that matters. Suppose your venue charges 0.05% per side as a taker and the instrument shows a spread of 0.03%.
commission in = 0.05% commission out = 0.05% the spread, once = 0.03% round trip = 0.13%
That 0.13% is the move you need before anything is yours. Price has to go your way by 0.13% for the trade to be flat. Use your own venue’s numbers rather than these - they differ by exchange, by tier and by instrument - but compute the total once, because everything else in this lesson is that total divided by something.
The cost as a share of the move you are after
A cost of 0.13% means nothing on its own. It only becomes information when you divide it by the size of the move you are trying to capture.
target cost / target net if the target is reached 0.20% 65.0% 0.07% 0.50% 26.0% 0.37% 1.00% 13.0% 0.87% 3.00% 4.3% 2.87% 10.00% 1.3% 9.87% round-trip cost = 0.13% throughout
This is the general principle of the lesson and it has no exceptions: the tighter your target, the larger the share of it the venue takes. At a 0.20% target almost two thirds of a successful trade is cost. At a 10% target it is barely more than one percent of the move. The cost did not change. Your choice of how far to aim changed what it means.
The same arithmetic applies in the other direction, which is the part that gets forgotten: a losing trade pays the cost too. A stop reached at 4.00% costs 4.13% after the round trip, so costs widen your losses as well as shrinking your gains. They are subtracted from both columns, never just one.
Frequency multiplies the toll
Because the cost is per round trip, it scales with how often you trade, while the moves you are capturing do not get any larger. Two plans, the same notional, the same 0.13%:
4 round trips a month 4 x 0.13% = 0.52% of notional 40 round trips a month 40 x 0.13% = 5.20% of notional 200 round trips a month 200 x 0.13% = 26.00% of notional
The third row is the one worth sitting with. A plan that opens and closes a position roughly ten times a week hands over 26% of its notional in a month, and has to find that much before it has found anything. Nothing about the strategy is wrong; the arithmetic simply has to be cleared first, and clearing it is harder than a chart suggests.
What costs do to the break-even arithmetic
The R lesson derived break-even with no costs in it. Add them. With a target T, a stop S and a round-trip cost C, all in percent, and a share p of occasions reaching the target:
average = p x (T - C) - (1 - p) x (S + C) setting that to zero and solving for p: without cost p = S / (T + S) with cost p = (S + C) / (T + S) the extra share required = C / (T + S)
That last line is the whole relationship in one expression, and it says something precise: the damage done by costs depends on how wide the brackets are, not on how much you trade in currency. Two examples, both with C = 0.13%.
T = 1.00% S = 1.00% without cost 1.00 / 2.00 = 50.0% with cost 1.13 / 2.00 = 56.5% extra needed 0.13 / 2.00 = 6.5 points T = 5.00% S = 3.00% without cost 3.00 / 8.00 = 37.5% with cost 3.13 / 8.00 = 39.1% extra needed 0.13 / 8.00 = 1.63 points
Six and a half points against one and six tenths. The cost was identical; the narrow bracket had to find four times as much extra accuracy to pay for it. This is why tight targets are harder than they look: they are not only harder to reach, they demand more of you before the toll is covered.
None of this says wide brackets are better - wide brackets mean larger losses when the stop is reached, and the position-size lesson already showed what a wider stop does to the size you can hold. It says that the bracket and the cost have to be considered together, and that a plan which looks sound on a chart with no costs on it has not been evaluated yet.
Reading someone else’s numbers, including ours
Any set of results you are shown - from a backtest, a screenshot, a spreadsheet, a signal provider - was computed with some assumption about cost, and often with none. The questions to ask are short and they are the same every time.
- What fee per side was assumed, and is it the taker fee or the maker fee? If the maker fee was used, what happened to the orders that would not have filled?
- Was the spread included at all, or were entries and exits priced at a single number per bar?
- Was slippage included, and at what size? A result computed at exact stop prices is understating losses by the amount the stop lesson describes.
- For anything held overnight, was funding charged?
- How many round trips produced the result? The frequency table above says what that costs, and a plan with many trades is affected far more by a cost assumption than a plan with few.
If the answers are not available, the result is not evaluable - which is a stronger statement than "be careful". A set of numbers computed without costs is not an optimistic version of the truth; it is arithmetic about a market that does not charge anything, and no such market exists. Apply the same question to us: see how a signal is built and the risk disclaimer.
The terms, defined
- Round-trip cost
- Commission on both sides plus the spread paid once, as a percent of notional. The move price must make before a trade is flat.
- Taker fee
- Charged when your order crosses the spread and executes immediately against resting orders. Higher than the maker fee.
- Maker fee
- Charged when your order rests in the book and is executed against. Lower, and the trade-off is that the order may never fill.
- Spread
- The distance between best bid and best ask. A real cost that appears on no statement, because you pay it in the price.
- Funding
- A periodic payment between longs and shorts on a perpetual contract, charged on notional. A cost of holding rather than of trading.
- Break-even share
- The share of occasions a target must be reached for the average result to be zero. (S + C) / (T + S) once costs are included.
What this means in practice
- Compute your own round-trip cost once, from your venue’s fee tier and the spread you actually get, and write it down as a percent.
- Divide that cost by the move you are aiming at before you take the trade. The ratio is the honest description of what you are attempting.
- Remember costs widen losses as well as shrinking gains. They come out of both columns.
- Multiply the round-trip cost by how often you trade. Frequency is the lever that turns a small toll into a large one.
- Add C / (T + S) to any break-even share you compute, and notice that narrow brackets pay the most for the same cost.
- Ask of any result you are shown what fee, what spread, what slippage and what funding were assumed. If nobody can say, the result is not evaluable.
Nothing in this lesson is financial advice or an instruction to trade. See the risk disclaimer. We never recommend a leverage setting - that decision is yours.