Part 3: From theory to the account · Chapter 10 of 14 · 11 min
Costs against the stop
The cost of a trade is fixed in dollars, so in R it weighs more the narrower the stop: 0.3 $ is worth 0.007 R with a 40 $ stop and 0.10 R with a 3 $ one. In the series of the books the average cost is 0.015 R per trade (8.5 R in total over 552 trades, 8.5% of a 10,000 $ account at 1%). LD, with +0.03 R on average, breaks even at a cost of 0.87 $ and loses from 1 $ up; SD, with wide stops, holds up to 8 $. A rule: keep the cost under 5% of the stop.
Cost is in dollars, risk is in R
Every trade has a cost: the spread (the difference between buying and selling price), a possible commission, the slippage (an execution price worse than the one seen), and the swap if you hold the position overnight. The first three you pay on every trade, and in dollars per ounce they are almost the same whatever the stop. The cost in R is therefore:
> cost in R = cost in $ per ounce ÷ stop in $ per ounce
With a cost of 0.3 $ (the assumption of all the books, already included in the results in R of the series):
| Stop | Cost 0.3 $ | Cost 0.6 $ | Cost 1 $ | Cost 1.5 $ |
|---|---|---|---|---|
| 3 $ | 0.100 R | 0.200 R | 0.333 R | 0.500 R |
| 5 $ | 0.060 R | 0.120 R | 0.200 R | 0.300 R |
| 10 $ | 0.030 R | 0.060 R | 0.100 R | 0.150 R |
| 20 $ | 0.015 R | 0.030 R | 0.050 R | 0.075 R |
| 40 $ | 0.007 R | 0.015 R | 0.025 R | 0.037 R |
The real cost depends on the broker, on the hour (the spread widens at night and on news) and on the type of order. The 0.3 $ of the books is an assumption, not a measurement.
Read the chart in words
Hourly gold chart of 25-26 December 2025 with the entry, the close stop and the target of a long trade.
| When | What | Price |
|---|---|---|
| 25 Dec 2025 18:00 UTC | entry | 4,489.91 |
| 25 Dec 2025 18:00 UTC | stop: −1 R (8.4 $) | 4,481.51 |
| 25 Dec 2025 18:00 UTC | target: +1 R | 4,498.31 |
Times in UTC.
How much it weighs in our series
Let us compute the cost of each trade as 0.3 $ divided by its stop:
| Series | Trades | Mean cost in R | 90th percentile | Maximum | Total cost |
|---|---|---|---|---|---|
| LA | 552 | 0.0154 R | 0.0263 R | 0.056 R | 8.5 R |
| LD | 367 | 0.0152 R | 0.0240 R | 0.058 R | 5.6 R |
| SD | 55 | 0.0035 R | 0.0049 R | 0.005 R | 0.2 R |
With a 10,000 $ account and 1% risk, the 552 buys and sells of LA cost 851 $ in total: 8.5% of the account in ten months, not counting compound interest. A system that trades often pays more, and pays it for certain: the expected gain is uncertain (chapter 2), the cost is not.
When the cost changes the sign
Let us recompute the series with different costs: the result in R of each trade changes by (0.3 − c) ÷ stop. Mean R per trade and sum of R:
| Cost | LA: mean R / sum | LD: mean R / sum | SD: mean R / sum | SX (stops of LA): mean R |
|---|---|---|---|---|
| 0 $ | −0.02 / −9.7 R | +0.04 / +16.2 R | +0.10 / +5.3 R | +0.21 |
| 0.3 $ | −0.03 / −18.2 R | +0.03 / +10.6 R | +0.09 / +5.1 R | +0.19 |
| 0.6 $ | −0.05 / −26.8 R | +0.01 / +5.0 R | +0.09 / +4.9 R | +0.18 |
| 1 $ | −0.07 / −38.1 R | −0.01 / −2.4 R | +0.08 / +4.6 R | +0.16 |
| 1.5 $ | −0.09 / −52.3 R | −0.03 / −11.7 R | +0.08 / +4.3 R | +0.13 |
| 2 $ | −0.12 / −66.5 R | −0.06 / −21.0 R | +0.07 / +4.0 R | +0.10 |
| 3 $ | −0.17 / −94.9 R | −0.11 / −39.7 R | +0.06 / +3.3 R | +0.05 |
(The error on mean R does not change with the cost: ±0.08 for LA, ±0.10 for LD, ±0.15 for SD.) Four readings.
- LA does not lose because of costs. Even at zero cost it is −0.02 R (−9.7 R): the cost worsens a system that already does not earn.
- LD is sensitive. With +0.03 R at 0.3 $, it breaks even at a cost of 0.87 $ and turns negative at 1 $ (−2.4 R). The margin between the assumed cost and the one that zeroes the mean is less than 0.6 $: the difference between two brokers, or between hours of the day.
- SD holds up because the stops are wide (84 $ on median, cost 0.0035 R) and the trades few (55): breakeven is at 8.1 $.
- SX, with a built-in edge of 0.21 R, falls to 0.10 at 2 $ and to 0.05 at 3 $. A true edge can survive a high cost; a small edge like that of LD cannot.
Read the chart in words
Daily gold chart with, below, four curves of cumulative R of LD with different costs.
The winning share needed to cover the cost
With target and stop at the same distance (1:1), the share of winning trades needed to break even is (1 + cost in R) ÷ 2. A cost of 0.1 R requires 55%, not 50%:
| Stop | Cost 0.3 $ | Cost 0.6 $ | Cost 1 $ | Cost 1.5 $ |
|---|---|---|---|---|
| 3 $ | 55.0% | 60.0% | 66.7% | 75.0% |
| 5 $ | 53.0% | 56.0% | 60.0% | 65.0% |
| 10 $ | 51.5% | 53.0% | 55.0% | 57.5% |
| 20 $ | 50.7% | 51.5% | 52.5% | 53.8% |
| 40 $ | 50.4% | 50.7% | 51.2% | 51.9% |
With a 5 $ stop and a 1 $ cost you need 60% of trades winning just to stay at zero. That is why whoever trades with stops of a few dollars (scalping) needs a large edge and very low costs.
Cost is not visible amid the noise
Let us divide LA and LD into four bands of equal number of trades by stop, and look at cost and net result (mean R ± error):
| Series, band | Stop | Trades | Mean cost | Net mean R | Winning |
|---|---|---|---|---|---|
| LA, 1 | from 5.4 to 15.5 $ | 138 | 0.026 R | 0.00 ± 0.17 | 51% |
| LA, 2 | from 15.5 to 21.7 $ | 138 | 0.016 R | −0.02 ± 0.17 | 50% |
| LA, 3 | from 21.7 to 29.7 $ | 138 | 0.012 R | −0.14 ± 0.17 | 43% |
| LA, 4 | above 29.7 $ | 138 | 0.007 R | +0.02 ± 0.16 | 51% |
| LD, 1 | from 5.1 to 15.6 $ | 91 | 0.024 R | +0.10 ± 0.21 | 56% |
| LD, 2 | from 15.6 to 20.5 $ | 92 | 0.017 R | +0.03 ± 0.21 | 52% |
| LD, 3 | from 20.5 to 29.1 $ | 92 | 0.013 R | +0.01 ± 0.21 | 51% |
| LD, 4 | above 29.1 $ | 92 | 0.007 R | −0.02 ± 0.20 | 48% |
The cost difference between band 1 and band 4 is 0.019 R in LA: a certain effect but much smaller than the error of ±0.17 over 138 trades. In real data the cost does not show in the result: you compute it, you do not observe it. That is why it counts more as a design rule than as a measurement.
What to do with it
- Measure your true cost: average spread at the hour you trade, commission, slippage on stops. Compare several brokers.
- Compute the cost in R of every trade: cost ÷ stop. If it exceeds 5% (stop under 20 times the cost: 6 $ at 0.3 $, 20 $ at 1 $), discard the trade or look for a more sensible stop.
- Recompute the mean R of your system with double the cost and see whether the sign holds: LD does not, SD does.
- Avoid thin hours and news for market entries, where spread and slippage grow.
- Count total costs: number of trades × mean cost. More trades means more certain cost for an uncertain gain.
Procedure in four steps
- Measure the real cost in $ per ounce (spread, commission, slippage) at the hours you trade.
- For every trade compute cost ÷ stop and discard those above 5%.
- Recompute your series with double the cost and see whether mean R stays positive.
- Add up total costs every month and compare them with the gross gain.
Exercises
Chapter exercises
1. A cost of 0.6 $ per ounce, a stop at 12 $. How much is the cost in R?
2. Target and stop at the same distance, stop at 10 $, cost 1 $. What winning share is needed to break even?
3. LD makes +0.03 R per trade with a cost of 0.3 $. What happens if the true cost is 1 $?
4. LA has a total cost of 8.5 R. With a 1% risk per trade, what share of the account does it correspond to, without compound interest?
Have a question or want to share your exercise?
Post in the community, or join the free signals room on Telegram.
Educational content, not financial advice. Trading involves risk.