Part 2: How much to risk · Chapter 5 of 14 · 13 min
Fixed size and fixed fraction
Two rules to decide how much to risk: the same amount on every trade (fixed size) or the same percentage of current capital (fixed fraction). On the real series the two rules give the same result when risk is small (LA at 1%: 82% and 81%) and diverge as it grows (SX at 5%: 670% against 15,049%, drawdown 17.8% against 37.4%). The martingale, which doubles after every loss, closed in profit on the real LA series and wiped out the account or took it below 5% in 72% of 4,000 random orders of the same trades (LA, base 1%).
Two rules to choose how much to risk
Once the stop is decided, what remains is to decide how much capital to put at risk. The two simplest rules are:
- Fixed size (or fixed risk): you always risk the same amount, for example 100 $ per trade, whatever has happened to the account. It is the rule you use without noticing when you always buy the same number of ounces.
- Fixed fraction: you always risk the same percentage of the capital of that moment, for example 1%. After a loss you risk fewer dollars, after a gain you risk more.
In both cases the quantity to buy comes from the stop:
> ounces = capital × risk percentage ÷ stop distance in dollars per ounce
With a 10,000 $ account, 1% (100 $) and the median stop of LA (21.7 $), you buy 4.6 ounces. If the stop is 12 $ you buy 8.3; if it is 40 $ you buy 2.5. The stop changes the quantity, the dollar risk stays the same. (Chapter 8 explains how to get to lots and ounces.)
What changes on the real series
Let us put the four series, in the order in which they happened, in an account with the two rules: in fixed size 1 R is worth a percentage of the initial capital; in fixed fraction it is worth the same percentage of current capital. Final capital as a percentage of the initial one, and maximum drawdown:
| Series | 2%, fixed size | 2%, fixed fraction | 5%, fixed size | 5%, fixed fraction |
|---|---|---|---|---|
| LA | 64% / dd 52.6% | 62% / dd 53.3% | 9% / dd 95.3% | 20% / dd 87.9% |
| LD | 121% / dd 25.6% | 115% / dd 32.2% | 153% / dd 43.1% | 108% / dd 63.7% |
| SD | 110% / dd 3.2% | 110% / dd 3.4% | 125% / dd 7.1% | 126% / dd 8.3% |
| SX | 328% / dd 10.7% | 876% / dd 16.7% | 670% / dd 17.8% | 15,049% / dd 37.4% |
What the table says.
- At small risk the two rules coincide. With SD, where losses are small, they cannot be told apart even at 5%. With LA at 1% they are 82% and 81%.
- When the risk grows, fixed fraction amplifies everything. With a true edge (SX) capital compounds: at 5% it goes from 670% to 15,049%, and the drawdown from 17.8% to 37.4%, because the dollar risk grows with the account. It is the effect of compound interest, in both directions.
- Without an edge, fixed fraction protects from the bottom. With LA at 5% fixed size closes at 9%, fixed fraction at 20%: after a loss you risk a smaller amount. With fixed size the account can even fall below zero: in random paths of LA at 5% the 5th percentile of the final result is −169%.
- With a mean close to zero (LD) fixed fraction earns less. At 5% LD closes at 153% with fixed size and at 108% with fixed fraction. The reason is in the next section.
A fact that surprises: final capital does not depend on the order of the trades (addition and multiplication are commutative), but the drawdown does. LA at 2% with fixed size closes at 64% in any order; with the same trades reversed, the drawdown goes from 52.6% to 70.6%.
In dollars, for a 10,000 $ account: LD at 2% with fixed fraction reaches 11,505 $, and 1 R goes from 200 $ to 235 $. SX at 2% reaches 87,573 $, and 1 R goes from 200 $ to 1,717 $: at that point you are risking in a single trade almost a sixth of the initial account. It happens when the edge is true and repeats for 552 trades: it is a hypothesis, not a promise.
Read the chart in words
Daily gold chart with, below, two curves of capital as a percentage of the initial one.
Why fixed fraction earns less with a mean close to zero
With fixed fraction, capital is multiplied at every trade by (1 + risk × R). A 5% gain followed by a 5% loss does not bring you back to the start: 1.05 × 0.95 = 0.9975. You lose a quarter of a point every time you alternate. For results with mean μ and deviation σ (in R) and risk f, the growth per trade is about:
> f × μ − f² × σ² / 2
The first term is the expected gain, the second is the cost of volatility. With LD (μ = 0.03, σ = 1) and f = 5% the first term is 0.0015, the second 0.00125: the cost eats almost all the gain, and in 500 trades growth is about +13%, instead of +75% with fixed size. The risk at which growth and cost balance is f = 2μ/σ²: for LD 6%, for LA zero (the mean is negative). Beyond that risk, capital falls on average even with a positive mean. Chapter 7 shows that the optimal risk is half this value.
On random paths of 500 trades (median, with the 5th to 95th percentile of final capital in brackets):
| Series, risk | Fixed size | Fixed fraction | Below initial capital (fixed / fraction) |
|---|---|---|---|
| LA, 2% | 67% (from −7% to 138%) | 65% (from 31% to 132%) | 77% / 84% |
| LD, 2% | 129% (from 53% to 205%) | 121% (from 56% to 258%) | 27% / 34% |
| LD, 5% | 172% (from −18% to 362%) | 111% (from 16% to 741%) | 27% / 47% |
| SX, 1% | 204% (from 166% to 240%) | 276% (from 189% to 396%) | 0% / 0% |
| SX, 2% | 308% (from 232% to 380%) | 724% (from 339% to 1,488%) | 0% / 0% |
| SX, 5% | 620% (from 430% to 800%) | 9,737% (from 1,454% to 58,996%) | 0% / 0% |
With a true edge fixed fraction wins; with a null or small edge (LA, LD) it tends to lose a bit because of volatility, but its advantage is that it never takes the account below zero. The 95th percentile of LD at 5% with fixed fraction is 741%: the possible gain is large, and the cost is that almost one trial in two ends below the initial capital (47%).
The martingale
The martingale is a sizing rule born at the roulette table: after every loss you double the stake, after a win you return to the base. The reasoning is that the first win recovers all the losses plus the initial stake. Let us apply it to the real series with a base of 1% of current capital:
| Series | Maximum risk reached | Final capital | Minimum capital | Same trades at fixed fraction 1% |
|---|---|---|---|---|
| LA | 128% of the account | 118% | 60% | 81% |
| LD | 64% | 300% | 75% | 109% |
| SX | 128% | 822% | 94% | 304% |
On the real series the martingale won in all three cases, even with LA, which loses 18 R. On LA and SX it reached a risk of 128% of the account (after seven losses in a row: 1% × 2⁷), that is more than all the capital. On LA it survived because the eighth trade of that series was a winner. Had it been a loss, the account would have been wiped out. The chart shows the path of LA: it looks like a money machine (+262% at the peak) and then gives back almost everything.
Read the chart in words
Daily gold chart with, below, two curves of capital as a percentage of the initial one.
To know what would have happened with a different order of the same trades, I repeat the path of 552 trades 4,000 times drawing at random from the results and count how many times the account ends wiped out or below 5%:
| Series | Base | Wiped out or below 5% | Median capital | Below 50% | Same trades at fixed fraction (below 50%) |
|---|---|---|---|---|---|
| LA | 0.5% | 46.2% | 30% | 55% | 0% |
| LA | 1% | 71.8% | 0% | 77% | 2.2% |
| LD | 0.5% | 33.0% | 111% | 37% | 0% |
| LD | 1% | 57.1% | 0% | 60% | 0% |
| SX (60% winning) | 0.5% | 7.7% | 427% | 8% | 0% |
| SX (60% winning) | 1% | 18.6% | 1,298% | 19% | 0% |
Even with a true edge (SX, 60% winning) the martingale at 1% wipes out the account one time in five, while fixed fraction with the same trades never falls below 50%. The median capital of SX with the martingale is very high (1,298%) because the series that do not blow up grow a lot: it is a distribution with extreme tails, in which the last 20% destroys everything.
The cause is arithmetic: risk grows as 2^k after k losses, and a streak of 8 losses appears, as seen in chapter 3, in 62% of 500-trade paths at a 50% winning share. With a base of 1% the eighth loss requires 128% of the account. No broker accepts that, and no account bears it.
The opposite, increasing size after gains, is what fixed fraction does: risk grows when the account grows, and comes back down if the account falls.
What to do with it
- Use fixed fraction to decide the size, with a small risk (0.5% to 2%). At these values the difference from fixed size is small and fixed fraction cannot take the account below zero.
- Recompute capital at regular intervals (for example at the start of each month or after a 10% change), if recomputing at every trade is inconvenient. What matters is not to recompute asymmetrically, only after gains.
- Never increase size after a loss, neither to recover nor to "get even". The martingale wipes out the account with high probability.
- If you use fixed size, check the real percentage: if the account halves, your 100 $ per trade is no longer 1% but 2%.
- Compute ounces from the stop, not from instinct: capital × risk ÷ stop distance.
Procedure in four steps
- Choose the risk per trade as a percentage of capital (0.5-2%).
- Before every trade compute the dollar risk (capital × percentage) and divide it by the stop distance in dollars per ounce: that is the ounces to buy.
- Recompute capital at fixed intervals and in the same way after gains and losses.
- Do not change the percentage halfway, neither after a streak of losses nor after one of wins.
Exercises
Chapter exercises
1. A 10,000 $ account, 1% risk, stop at 21.7 $ per ounce. How many ounces do you buy?
2. On LD with 5% risk, fixed fraction closes at 108% and fixed size at 153%. Why does fixed fraction earn less?
3. With a martingale at a 1% base, how much would you risk after seven losses in a row?
4. A 10,000 $ account falls to 5,000 $. With fixed size (1 R = 100 $) what percentage of the current account do you now risk on every trade?
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.