Dynamic Position Sizing: How ATR and the Kelly Criterion Protect Your Portfolio from Volatility
Position sizing is arguably the single most important variable in portfolio management — and yet most retail investors handle it intuitively. "I'll put in 50,000 CZK" regardless of whether the stock is calm like ČEZ or wild like NVDA. This article shows how to connect volatility measurement (ATR), stop derivation, and the mathematics of the Kelly Criterion into a concrete, reproducible process.
---
What Is ATR and Why Does It Matter
Average True Range (ATR) is an indicator developed by Welles Wilder that measures the average true daily volatility of an asset over a chosen period (typically 14 days).
"True Range" (TR) is the maximum of three values:
- High − Low of the current day
- |High − previous Close|
- |Low − previous Close|
ATR then averages these values. Key property:
ATR is expressed in absolute price units (CZK / $), not percentages — for a stock trading at 1,000 CZK with ATR = 30 CZK, we know it typically moves ±30 CZK per day.
ATR vs. Standard Deviation
| Property | ATR | Std. Deviation of Returns |
|---|
| Unit | CZK / $ | % or decimal |
| Sensitivity to gaps | Yes (includes overnight) | Depends on calculation |
| Intuitive for stops | High — directly in price | Requires conversion |
| Use for stop placement | Direct (n × ATR) | Needs recalculation |
For position sizing, ATR is more practical — stops are set as a multiple of ATR (e.g., 2× ATR below entry price).
---
Volatility-Based Position Sizing: The Formula
The core idea: risk a fixed CZK amount (or % of portfolio) on every trade, and adjust position size to match the stop width.
Formula
Number of shares = Risk in CZK / Stop in CZK per share
where:
Stop in CZK per share = n × ATR(14)
Risk in CZK = Portfolio × % risk per trade
A common setting is n = 2 (stop is 2× ATR below entry) and % risk = 1–2% of portfolio per trade.
---
Concrete Example: ČEZ vs. NVDA
Assume a portfolio of 1,000,000 CZK and a rule: risk max 1% of portfolio per trade = 10,000 CZK.
| Parameter | ČEZ (calm) | NVDA (volatile) |
|---|
| Current price | 850 CZK | 3,100 CZK* |
| ATR(14) | 18 CZK | 185 CZK |
| Stop (2× ATR) | 36 CZK | 370 CZK |
| Risk per trade | 10,000 CZK | 10,000 CZK |
| Number of shares | 10,000 / 36 ≈ 277 shares | 10,000 / 370 ≈ 27 shares |
| Total position value | ~235,450 CZK | ~83,700 CZK |
\* Prices are illustrative for calculation purposes only, not a recommendation.
Result: Same CZK risk — completely different share counts and capital allocation. With the calm stock, the strategy naturally enters a larger position; with the volatile one, a smaller position. Without this approach, an investor buying 100,000 CZK in each would be risking ~5× more in NVDA than in ČEZ.
---
The Kelly Criterion: Optimal Bet Size
The Kelly Criterion (John L. Kelly Jr., 1956) answers the question: what fraction of capital maximizes long-term geometric growth?
Formula
f* = (p × b − q) / b
where:
- f* = optimal fraction of capital per trade
- p = probability of profit (win rate)
- q = probability of loss = 1 − p
- b = average gain / average loss (reward/risk ratio)
Example
A strategy with a win rate of 55% and average R/R of 1.5:1:
f* = (0.55 × 1.5 − 0.45) / 1.5
f* = (0.825 − 0.45) / 1.5
f* = 0.375 / 1.5
f* = 0.25 → 25% of capital per trade
Why Fractional Kelly (¼ Kelly) Is Used in Practice
Full Kelly is mathematically optimal, but in the real world it is extremely aggressive and highly sensitive to the accuracy of input estimates:
- Win rate and R/R are estimates — small errors significantly change the result
- Full Kelly produces enormous drawdowns (even 50%+) on the path to the optimum
- An investor using full Kelly will psychologically fail before the math "pays off"
¼ Kelly (f
/ 4) dramatically reduces drawdown while preserving most of the long-term return. In practice, the fractional Kelly result is used as an upper bound* on risk per trade, while ATR-based sizing provides the concrete share count.
| Approach | Risk per Trade | Drawdown | Psychological Sustainability |
|---|
| Full Kelly (25%) | Very high | ~50–60% | Low |
| ½ Kelly (12.5%) | High | ~30–40% | Medium |
| ¼ Kelly (6.25%) | Medium-low | ~15–20% | High |
| Fixed 1–2% | Conservative | <10% | Very high |
Most professional systematic traders use either ¼ Kelly or fixed 1–2% as a practical compromise.
---
The Relationship Between ATR and Stop Placement
A stop-loss is not an arbitrary number — it should respect the market's natural "noise." A stop placed too close to price will be triggered by normal intraday volatility. A stop placed too far creates an unacceptably large loss per trade.
Empirical guidelines:
- 1.5–2× ATR below entry for swing trading (3–10 days)
- 2.5–3× ATR for positional trading (weeks to months)
- Never a stop tighter than 0.5× ATR — statistically very likely to be hit prematurely
When volatility changes (e.g., NVDA pre-earnings vs. post-earnings), position size should be recalculated — ATR changes significantly around such events.
---
How to Do This in QMA
QMA provides several tools that directly support this approach:
/stocks/[symbol] — Stock Detail Page
On the detail page of a specific stock (e.g.,
/stocks/NVDA or
/stocks/CEZ), technical indicators including ATR(14) are available. This is where the current ATR value can be monitored and compared to its historical average — a high ATR signals elevated volatility and the need to reduce position size.
/portfolio — Portfolio Management
The portfolio section displays total exposure and position breakdown. It allows a visual check of whether the portfolio is diversified not only by sector but also by
volatility — a balance between high-volatility and low-volatility positions is important for overall risk management.
/screener — Stock Filtering
The screener allows filtering stocks by various parameters including volatility metrics. A strategy targeting low volatility for conservative sizing can screen for stocks with ATR% below a defined threshold.
/strategy — Rules Backtesting
In the
/strategy section, it is possible to backtest rules including volatility-based sizing — the strategy tool evaluates historical results of different stop-loss and position-sizing approaches, enabling an empirical comparison of full Kelly vs. ¼ Kelly on historical data.
/smart-money — Institutional Flow
The
/smart-money section displays institutional flows and unusual volume. High institutional interest in a stock with a low ATR may indicate stable accumulation — contextual information relevant to sizing decisions.
QMA Confidence Score
QMA assigns stocks a
confidence score that aggregates fundamental, technical, and sentiment factors. This score
does not constitute investment advice, but can serve as one input when deciding whether to open a position at all — historically, a higher score has correlated with a lower probability of adverse outcomes, which directly influences the win rate estimate used in the Kelly formula.
---
Summary: Step-by-Step Process
- Find ATR(14) for the stock at
/stocks/[symbol] - Set the stop at 2× ATR below entry price
- Determine risk in CZK = portfolio × % risk (typically 1–2%, or the ¼ Kelly result)
- Divide risk in CZK / stop in CZK = number of shares
- Check total position value relative to portfolio — it should not exceed a reasonable limit (e.g., 20–25% of the portfolio in a single position)
- Recalculate whenever ATR changes significantly (earnings, macro events)
This process turns the decision of "how much to buy" into a
mathematical result, not a guess.
---
Disclaimer
QMA is an analytical and educational tool, not investment advice. Past performance is not a guarantee of future results. All investment decisions are your own responsibility.