KW focus: position sizing strategies • Intent: Informational • Angle: Kelly-lite, fixed-fractional, max drawdown caps
If markets are where edges get compounded, position sizing is the throttle. Get the size wrong and even a good strategy can self-destruct. Get it right and mediocre signals can still survive long enough to improve. This guide breaks down the three most practical position sizing strategies—fixed-fractional, Kelly-lite, and max drawdown caps—and shows exactly how to use the math without blowing up your account.
Why size matters more than you think
Survival beats brilliance. You can be directionally right and still go bankrupt with oversized trades.
Variance is the enemy. Position size determines how much of each trade’s variance hits your equity curve.
Compounding prefers stability. The higher the volatility of returns, the lower your long-run geometric growth for a given average return (volatility drag).
Goal: Choose sizes that (1) keep your risk of ruin negligible, (2) control drawdowns to your tolerance, and (3) allow compounding to work.
The risk budget: start here (then stick to it)
Before any formula, set a risk budget:
Per-trade risk (r%): the maximum percentage of equity you’re willing to lose if a trade hits its stop. Typical for discretionary traders: 0.25%–1.0%; for diversified, low-correlation systems: sometimes 1.0%–2.0%.
Daily/weekly loss limit: e.g., −2% per day or −4% per week across all positions.
Max drawdown cap: the point at which you must de-risk (more below), e.g., −12% on the total equity curve.
Everything that follows plugs into this budget.
Strategy 1: Fixed-Fractional (risk-per-trade)
What it is: Risk a fixed percentage of current equity on each trade. This adapts position size as your account grows or shrinks and is the most widely used of all position sizing strategies.
Core formula
- Let E = current equity
- Let r = risk per trade (as a decimal, e.g., 0.01 for 1%)
- Let D = stop distance in price units (entry − stop), per share/coin/contract
- Then Units = (E × r) ÷ D
Multiply Units × entry price to get notional exposure.
Example (equities/crypto):
- Equity E = ₹10,00,000
- Risk r = 1% → ₹10,000
- You buy BTC at ₹50,00,000 with a stop 8% lower → D = 0.08 × 50,00,000 = ₹4,00,000 per BTC
- Units = ₹10,000 ÷ ₹4,00,000 = 0.025 BTC
- Notional = 0.025 × ₹50,00,000 = ₹1,25,000
If stopped, loss ≈ ₹10,000 (ignoring slippage/fees).
Why it works
- Scales down after losses (self-defensive) and up after gains (compounds).
- Meshes cleanly with ATR-based or structure-based stops.
Caveats
- Correlated positions can stack risk. If you put on five highly correlated trades, your effective r% is closer to 5× one trade’s risk.
- Stop quality matters. A too-tight or arbitrary stop invalidates the math.
Strategy 2: Kelly—then “Kelly-lite”
What it is: Kelly sizing maximizes long-run geometric growth (expected log wealth). In practice it’s often too aggressive because real markets have fat tails, regime shifts, slippage, and model error. That’s why professionals typically use fractional Kelly—often ½-Kelly (“Kelly-lite”) or less.
Discrete-outcome Kelly (win/loss form)
Let p = probability of a win
Let q = 1 − p
Let b = win/loss payoff ratio (average win divided by average loss)
Kelly fraction:
f∗=bp−qbf^* = \frac{bp – q}{b}f∗=bbp−q
(fraction of equity to wager; if negative, don’t trade)
Example: hit rate p = 0.45, b = 2 (average win twice average loss)
f∗=2⋅0.45−0.552=0.90−0.552=0.175=17.5%f^* = \frac{2 \cdot 0.45 – 0.55}{2} = \frac{0.90 – 0.55}{2} = 0.175 = 17.5\%f∗=22⋅0.45−0.55=20.90−0.55=0.175=17.5%
Full-Kelly says risk ~17.5% of equity—far too high for real-world trading.
Kelly-lite (½-Kelly) → 8.75%; ¼-Kelly → 4.375%. Many practitioners still find ¼-Kelly punchy and go even lower.
Continuous-return Kelly (mean/variance form)
For a strategy with expected excess return μ and variance σ² per period:
f∗≈μσ2f^* \approx \frac{\mu}{\sigma^2}f∗≈σ2μ
This version generalizes to multiple assets with covariance matrix Σ (risk-parity-like vector proportional to Σ⁻¹ μ), but again fraction it down in practice.
Why Kelly-lite works
It preserves Kelly’s growth logic while de-risking model error.
Smaller fractions sharply cut drawdowns and risk of ruin at modest cost to growth.
Caveats
Inputs (p, b, μ, σ) are unstable. Estimate errors make full-Kelly dangerous.
Tail events and slippage are not fully captured by your backtest’s averages.
Strategy 3: Max drawdown caps (equity-curve control)
What it is: A rule-set that forces you to cut exposure as losses accumulate, regardless of signal quality. Drawdown caps protect the portfolio from a strategy going out of tune or a regime change.
Common implementations
Step-downs:
0–5% DD: 100% normal size
5–10% DD: 50% size
10–12% DD: 25% size
12% DD: flatten to cash / halt until recovery to −8%
Linear throttle:
Scale = max(0, 1 − DD / DD_max).
If DD (drawdown) reaches DD_max, exposure goes to zero.
Time-based lockouts:
If daily loss > −2%, stop trading for the rest of the day.
If weekly loss > −4%, size at 50% the next week.
Why it works
Limits the worst-case path (behavioral + capital preservation).
Adds a regime-switch circuit breaker for model risk.
Caveats
Can mute rebounds if the system recovers quickly.
Needs clarity on when to re-enable size (e.g., equity recovers above a moving high-water mark or a fixed level).
Volatility targeting (a useful add-on)
Not a full sizing regime by itself, but powerful when combined with the above. The idea: scale positions so your portfolio volatility stays near a target (e.g., 8%–12% annualized).
Single asset shortcut
Weight≈Target VolRealized Vol\text{Weight} \approx \frac{\text{Target Vol}}{\text{Realized Vol}}Weight≈Realized VolTarget Vol
Use daily realized vol (e.g., 20-day) and annualize via σann≈σdaily×252\sigma_\text{ann} \approx \sigma_\text{daily} \times \sqrt{252}σann≈σdaily×252.
Example: If your asset’s annualized vol ≈ 20% and your target is 10%, hold ~50% of capital in it. If vol spikes to 40%, cut weight to ~25%.
Why it helps
Smooths the equity curve.
Reduces the chance that a single high-vol asset dominates portfolio risk.
Putting it together: a robust, real-world playbook
Here’s how traders blend these position sizing strategies:
Choose a conservative r% (fixed-fractional) for per-trade sizing—e.g., 0.5%–1.0%.
Overlay Kelly-lite at the strategy level to allocate capital among systems: allocate more capital to higher-edge, lower-vol systems using ¼-Kelly or less (based on μ/σ²), then cap with risk budgets.
Throttle by volatility to keep portfolio vol inside your comfort zone.
Enforce drawdown caps (step-down or linear) to stop bad regimes from spiraling.
This stack lets you (a) size each trade sanely, (b) tilt capital toward better edges, (c) avoid excessive portfolio swings, and (d) survive model error.
Worked examples (complete arithmetic)
Fixed-fractional with ATR stop
Equity E = ₹20,00,000
Risk per trade r = 0.75% → ₹15,000
Stock at ₹500; 14-day ATR = ₹12; stop at 2×ATR = ₹24 → D = ₹24 per share
Units = ₹15,000 ÷ ₹24 ≈ 625 shares
Notional ≈ 625 × ₹500 = ₹3,12,500
If stopped, loss ≈ ₹15,000 (ex-slippage/fees)
Kelly-lite to allocate across 3 strategies
Assume monthly stats (excess returns over cash):
| Strategy | μ (avg mthly) | σ (mthly) | μ/σ² (Kelly) |
| Trend Futures | 1.2% | 4.0% | 0.012 / 0.0016 = 7.50 |
| Mean-Rev Equities | 0.8% | 3.0% | 0.008 / 0.0009 = 8.89 |
| Crypto Carry | 1.8% | 9.0% | 0.018 / 0.0081 = 2.22 |
The raw Kelly ratios say mean-rev is strongest, then trend, then carry. But full Kelly is unrealistic.
¼-Kelly weights (proportional): 1.875 : 2.222 : 0.555. Normalize to 100%:
Sum = 1.875 + 2.222 + 0.555 = 4.652
Weights ≈ 40.3%, 47.8%, 11.9% (then check portfolio vol and apply drawdown caps).
Max drawdown cap step-down
You set: DD_max = −12% on total equity.
Rules:
DD 0–6% → 100% size; 6–9% → 50% size; 9–12% → 25% size; >12% → flat.
Account hits −7% → next day all new trades sized at 50% of normal fixed-fractional r%.
Equity recovers to −4% → restore full size (or step up gradually).
Advanced touches that pay off
Correlation-aware caps
Define a portfolio heat metric: sum of per-trade r% multiplied by a correlation factor. If multiple trades share >0.8 correlation (same sector/coin beta), count them at 1.5× their risk toward your limit.
Gap/slippage buffer
Multiply D (stop distance) by 1.1–1.2 to account for gaps, especially in crypto or small caps.
Vol targeting per position
For each instrument i, set weight wi∝1σ^iw_i \propto \frac{1}{\hat\sigma_i}wi∝σ^i1 so each contributes similar vol. Then cap each position by your fixed-fractional risk.
Equity-curve filters
If a strategy’s rolling Sharpe < 0 over 3 months, halve size until it recovers.
Capacity awareness
If your avg daily $ traded in a name is > 5–10% of its ADV, cut size aggressively; liquidity risk rises non-linearly.
Risk of ruin (intuitive take)
Exact “ruin” math depends on distributional assumptions, but rules of thumb:
If you risk ≤1% per trade and have a positive edge, ruin probability becomes very small unless you stack too many correlated bets.
Doubling r% roughly quadruples blow-up risk when tails dominate.
Kelly-lite lowers ruin risk dramatically versus full Kelly; combine with drawdown caps for path control.
Operational checklist (use this before every trade)
Define the stop (structure/ATR), then compute D.
Decide r% (e.g., 0.5%–1.0%).
Compute Units = (E × r) ÷ D. Round down to lot size.
Check portfolio heat: sum of active r% (correlation-adjusted). If above limit, skip or cut.
Apply vol targeting (if used) and instrument caps (e.g., no single name >12% of equity).
Enforce drawdown step if equity curve is below thresholds.
Document the numbers (pre-trade log). If you can’t write it, don’t size it.
Common mistakes (and simple fixes)
Mistake: Sizing off conviction, not risk.
Fix: Always compute units from stop distance; conviction can adjust whether you take the trade, not the math.
Mistake: Ignoring correlation.
Fix: Treat sector/coin clusters like one big trade when summing risk.
Mistake: Moving stops to fit size.
Fix: Set stops from structure/volatility first; size comes after.
Mistake: Using full Kelly from a backtest.
Fix: Use ¼-Kelly or less and apply drawdown caps.
Mistake: No slippage buffer.
Fix: Inflate D by 10–20% for sizing, or reduce r% accordingly.
Quick “formulas at a glance”
Fixed-Fractional Units:
Units=E×rD\text{Units} = \frac{E \times r}{D}Units=DE×r
Discrete Kelly:
f∗=bp−(1−p)bf^* = \frac{bp – (1-p)}{b}f∗=bbp−(1−p)
Kelly-lite:
fuse=c×f∗,c∈[0.1,0.5]f_{\text{use}} = c \times f^*, \quad c \in [0.1, 0.5]fuse=c×f∗,c∈[0.1,0.5]
Vol Target Weight (single asset):
w≈Target VolRealized Volw \approx \frac{\text{Target Vol}}{\text{Realized Vol}}w≈Realized VolTarget Vol
Linear Drawdown Throttle:
Scale=max (0,1−DDDDmax)\text{Scale} = \max\!\left(0, 1 – \frac{\text{DD}}{\text{DD}_{\max}}\right)Scale=max(0,1−DDmaxDD)
A simple, robust template you can adopt today
Set budgets: r% = 0.75%, DD_max = −12%, day loss stop −2%.
Per trade: ATR stop at 2×ATR (or structural swing-low/high). Size with fixed-fractional.
Per strategy: Allocate capital with ¼-Kelly based on rolling μ/σ², rebalance monthly, cap any one strategy at 40% of portfolio.
Vol targeting: Aim for 10% annualized portfolio vol; scale aggregate size weekly to target.
DD control: Step-down size at −6% and −9%; flat at −12% until recovery.
This stack covers 90% of the edge-preservation problem with 10% of the complexity.
Final word
Edges are fragile; risk is permanent. The best position sizing strategies—fixed-fractional for per-trade discipline, Kelly-lite for growth-aware allocation, and drawdown caps for survival—work because they respect what markets can do to you when you’re wrong and how rarely you’ll be perfectly right. Put the math in front of your conviction, automate the checks, and let compounding do the rest.
Disclaimer: This article is for education only and is not investment advice. Trading involves risk, including the possible loss of principal. Backtests and probability estimates are uncertain and may not reflect future conditions.
