Part 3: From theory to the account · Chapter 11 of 14 · 11 min
Several positions at once
Two positions open together do not halve the risk: they add it up according to how correlated they are. Running LA and LD together, 62% of LA trades overlap with an LD one; when they go the same way the results are correlated (+0.41) and coincide (both winning or both losing) in 65% of cases, against 47% expected if they were independent. The drawdown of the combined series (42.7 R) exceeds that of each (35.3 and 18.9), and the worst day (16 September 2026) closed 7 trades at −5.1 R.
Three ways of having several positions
So far we have reasoned about one trade at a time. In practice you end up with several open positions in three ways:
- More entries of the same system on the same instrument (adding to a position in progress).
- More systems on the same instrument (for example a system on levels and one on swings, like LA and LD).
- More instruments (gold, silver, indices, currencies).
In all three the total risk is not the sum of the risks of the single positions, nor its root: it depends on how much the results move together, that is on the correlation. Two positions on the same instrument in the same direction are almost a single, larger position; two positions in opposite directions largely cancel.
In the series of the books each system opens one position at a time (within the system, a trade closes before another enters). The problem appears when you run the two systems together, which is what happens to you if you follow two methods on the same chart.
LA and LD together
Trades are short: from entry to exit, median 3 hourly candles in both series (mean 4.5, maximum 24). Yet they overlap often:
- 344 of the 552 LA trades (62.3%) fall during at least one LD trade, and 321 of the 367 LD trades (87.5%) during an LA one;
- overall there are 432 overlapping pairs;
- hours with no open position: 23%; with one: 51%; with two: 26% (the maximum is two, because each series opens one at a time).
The overlapping pairs are 180 in the same direction and 252 in the opposite direction. And they behave in opposite ways:
| Pairs | Number | Correlation of results in R |
|---|---|---|
| All overlapping pairs | 432 | −0.03 |
| Same direction | 180 | +0.41 |
| Opposite direction | 252 | −0.28 |
Overall the correlation is close to zero because the two halves cancel, but it is an average that deceives: you cannot choose when one or the other happens. The deviation of the pair's result (sum of the two R) is 1.62 R in the same direction, 1.16 R in the opposite direction; two independent results would give 1.41 R. In the 180 pairs in the same direction the outcomes are:
| Outcome of the pair | Observed | Expected if independent |
|---|---|---|
| Both winning | 57 | 41 |
| Both losing | 60 | 44 |
| One winning, one losing | 63 | 95 |
The two positions have the same outcome in 65% of cases (117 out of 180), against 47% expected if they were independent. Opening a second position in the same direction is not diversifying: it is increasing the size of the first.
In the opposite-direction pairs the gain of one position partly offsets the loss of the other: net exposure shrinks (stops and sizes are not identical, so it does not cancel), but you pay the costs twice (chapter 10). A long and a short position on the same instrument, at the same time, is not a strategy: it is two costs.
What happens to the combined series
Let us put the two series together, in order of entry:
| Series | Trades | Mean R ± error | Sum | Maximum drawdown | Longest losing streak |
|---|---|---|---|---|---|
| LA | 552 | −0.03 ± 0.08 | −18.2 R | 35.3 R | 7 |
| LD | 367 | +0.03 ± 0.10 | +10.6 R | 18.9 R | 6 |
| LA + LD | 919 | −0.01 ± 0.06 | −7.6 R | 42.7 R | 8 |
The maximum drawdown of the combined series (42.7 R) is larger than that of either of the two. The negative periods of the two systems partly overlap (August-September, for example) and the drawdowns add up; the longest losing streak rises to 8. Mean R stays close to zero, and with more trades the error falls (±0.06): one more system does not create an edge that is not there.
Let us look at the R of each day (sum of the trades closed that day):
| Measure | Value |
|---|---|
| Days with trades | 232 |
| Mean daily R | −0.03 |
| Daily standard deviation | 1.82 R |
| Best day | +7.9 R |
| Worst day | −5.1 R |
| Days at −2 R or worse | 36 (15%) |
| Days at −3 R or worse | 18 (8%) |
With 1% risk per trade, one day in 13 you lose 3% or more of the account; with 2%, 6%. The worst day in the data is 16 September 2026: seven trades closed, one winning (+0.99 R) and six losing, sum −5.1 R. With 2% that would have been −10% in a day.
Read the chart in words
Daily gold chart with, below, the bars of the daily R of LA and LD together and two horizontal lines at 0 and −3.
Read the chart in words
Hourly gold chart of 15-17 September 2026 with seven numbers on the entry candles of the LA and LD trades closed on 16 September.
| When | What | Price |
|---|---|---|
| 15 Sept 2026 20:00 UTC | Point 1 | 4,304.74 |
| 16 Sept 2026 00:00 UTC | Point 2 | 4,291.61 |
| 16 Sept 2026 01:00 UTC | Point 3 | 4,298.82 |
| 16 Sept 2026 02:00 UTC | Point 4 | 4,328.45 |
| 16 Sept 2026 03:00 UTC | Point 5 | 4,346.46 |
| 16 Sept 2026 15:00 UTC | Point 6 | 4,368.12 |
| 16 Sept 2026 16:00 UTC | Point 7 | 4,358.99 |
Times in UTC.
The risk of several positions
With N positions each with risk f, results with the same deviation and average correlation ρ, the deviation of the total result is:
> f × σ × √(N × (1 + (N − 1) × ρ))
The effective total risk, in units of the risk of a single position:
| Positions | ρ = 0 | ρ = 0.3 | ρ = 0.5 | ρ = 0.8 | ρ = 1 |
|---|---|---|---|---|---|
| 2 | 1.41 | 1.61 | 1.73 | 1.90 | 2.00 |
| 3 | 1.73 | 2.19 | 2.45 | 2.79 | 3.00 |
| 5 | 2.24 | 3.32 | 3.87 | 4.58 | 5.00 |
With three positions at 1% and a correlation of 0.5 the true risk is about 2.5%, not the 3% of the worst case and not the 1.7% of the independent case. For positions in the same direction on the same instrument the correlation is close to 1: they are a single position and the risk is the sum. For different instruments the correlation must be measured, not assumed: in the books we have no data on other instruments.
A cap on positions
The simple way to limit all this is a cap: at most k open positions, or a cap on the total risk of the open positions (the portfolio "heat"). Let us try it on LA + LD, skipping a trade if k are already open:
| Cap | Trades taken | Sum | Mean R | Maximum drawdown |
|---|---|---|---|---|
| None (maximum 2) | 919 | −7.6 R | −0.01 | 42.7 R |
| 1 position | 606 | −11.6 R | −0.02 | 29.7 R |
With a cap of one position the drawdown falls from 42.7 to 29.7 R (about −30%), and 34% fewer trades are made: costs and risk fall. The sum worsens a little (from −7.6 to −11.6 R), because the two series, taken one by one, have no edge. The cap reduces the risk, it does not create the edge.
What to do with it
- Set a cap on the total open risk (for example 2-3% of the account, the sum of the stops of all positions).
- Count two positions in the same direction on the same instrument as one, of double size: the risk of the sum is the sum of the risks.
- Do not hold a long and a short position on the same instrument: net exposure shrinks and costs double.
- For different instruments measure the correlation on your data before counting on diversification, and use the highest value you have seen.
- Do not add to a losing position: it is a larger position against you.
Procedure in four steps
- List the open positions and, for each, the risk as a percentage of the account (stop distance × ounces ÷ capital).
- Group the positions in the same direction on the same instrument (or on highly correlated instruments) and add the risks.
- Check that the total open risk stays under the cap you set.
- Before opening a new position, if the cap would be exceeded, skip the trade or reduce the size.
Exercises
Chapter exercises
1. Three positions of 1% each, with an average correlation of 0.5. What total effective risk do you have, roughly?
2. A long and a short position on gold are open at the same moment. What does it mean?
3. On 16 September 2026 LA and LD closed 7 trades for −5.1 R in total. With a 2% risk per trade, what effect did it have on the account?
Read the chart in words
| When | What | Price |
|---|---|---|
| 15 Sept 2026 20:00 UTC | Point 1 | 4,304.74 |
| 16 Sept 2026 00:00 UTC | Point 2 | 4,291.61 |
| 16 Sept 2026 01:00 UTC | Point 3 | 4,298.82 |
| 16 Sept 2026 02:00 UTC | Point 4 | 4,328.45 |
| 16 Sept 2026 03:00 UTC | Point 5 | 4,346.46 |
| 16 Sept 2026 15:00 UTC | Point 6 | 4,368.12 |
| 16 Sept 2026 16:00 UTC | Point 7 | 4,358.99 |
Times in UTC.
4. The combined series LA + LD has a maximum drawdown of 42.7 R, larger than that of either of the two. Why?
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.