Most traders treat every trade the same. Same risk percentage, same number of contracts, same exposure – regardless of what the market conditions look like that day or what the statistical context of that specific trade is. This is a mistake, and it’s one I made for a long time before developing a better approach.
Dynamic position sizing (DPS) is a concept we developed privately within our hedge fund, and it’s become what I consider our second core edge – the first being our rigorous robustness testing framework. The idea is straightforward: the probability of success for any given trade is not constant. It fluctuates. If you can identify that fluctuation, you can vary your position size accordingly and extract a meaningful additional edge from data you already have.
Why Trade Probabilities Are Never Constant
If your strategy has an average win rate of 40%, that 40% is a long-run average across thousands of trades. It doesn’t mean that every individual trade has a 40% probability of winning. Some trades within that same strategy will have 30% probability. Others will have 60%. The average is 40%, but it fluctuates constantly with market conditions, time of day, day of week, volatility regime, and dozens of other variables.
This is important because it means treating all trades equally is mathematically suboptimal. If you know – based on statistical evidence – that certain conditions produce a 60% win rate within your system while others produce only 30%, why would you risk the same amount on both?
In our hedge fund, risk per trade fluctuates between zero (we skip the trade entirely) and 3.5% of capital. Zero means we’ve assessed the specific conditions of that trade and the probability is too low to justify any position. 3.5% means conditions strongly favour entry and we want maximum allocation. Every trade falls somewhere on that spectrum based on the algorithm’s probability assessment.
A Basic Example: Day of Week as a Probability Factor
Day of week is a concept popularised by Larry Williams, and it’s a clean example of how to think about DPS in practice. For many index strategies, there are statistically meaningful differences in the probability of profitable long or short trades by day of week. This pattern holds up across 15-20+ years of data with reasonable statistical significance.
The naive application would be: only trade long on Monday, only trade short on Tuesday. That approach doesn’t work well. It reduces sample size, cuts off trades you’d want to take, and the overall strategy quality usually deteriorates.
The DPS application is different. Say Monday statistically shows slightly higher probability for long trades. Rather than only trading long on Monday, you take two contracts for long trades on Monday and one contract for short trades. You don’t eliminate the short trades – you just express more confidence in the statistically stronger setup by sizing up, and less confidence in the weaker setup by sizing down.
When we tested this simple concept on basic day-of-week patterns, the equity curve got smoother and the net profit to drawdown ratio improved. That’s the foundation of the concept – and it scales from there into much more sophisticated applications.
How to Apply DPS: From Simple to Advanced
Starting with Fixed Fraction Adjustments
The simplest implementation: take your base risk per trade – say 1% of capital – and adjust it slightly based on probability assessments. A trade with average conditions stays at 1%. A trade where conditions suggest lower probability might drop to 0.7-0.8%. A trade where conditions suggest higher probability might increase to 1.2-1.3%.
These are small adjustments, but applied consistently across hundreds or thousands of trades, they compound into meaningful performance improvements.
Adjusting Number of Contracts
Alternatively, you can vary the number of contracts directly rather than the risk percentage. The logic is the same: more contracts when probability is higher, fewer (or zero) when probability is lower. This is more intuitive for many traders and produces the same mathematical outcome.
Adjusting Breakout Level Distance
A more advanced application – one we discuss in depth elsewhere – involves changing the distance of the breakout entry level itself based on probability. Higher-probability setups can justify tighter entry criteria. Lower-probability setups might require more confirmation before entry. This modifies both when you enter and how much you risk in a single adjustment.
Addressing the Overfitting Concern
The obvious question: isn’t this just overfitting? If you test enough conditions and find ones that happen to show higher win rates historically, aren’t you just curve-fitting to past data?
The risk of overfitting is always present. It never disappears. The mitigation is the same as for any strategy component: rigorous robustness testing.
In our hedge fund, we only apply DPS to strategies that have already passed Robustness Level 3 – the most demanding level of our validation framework. These are strategies where the core logic has been stress-tested extensively and where live performance has matched back-test performance closely over time. Layering DPS onto a poorly tested strategy amplifies curve fitting. Layering it onto a robust, validated strategy amplifies genuine edge.
At the time of recording this episode, our first CTA programme was approaching one full year of live trading with DPS active. Live performance was matching back-test performance closely across that period. That’s the validation you need to have confidence in the concept.
What DPS Is Not
DPS is not Martingale. Martingale means increasing position size after losses – the assumption being that a losing streak must end, so you bet bigger after each loss. That’s mathematically dangerous. A long enough losing streak bankrupts you before the mean reversion arrives.
DPS is not about the recent sequence of wins and losses at all. It’s about the statistical properties of each specific trade opportunity, assessed fresh before each entry. A trade after three losses isn’t treated differently than a trade after three wins – what matters is the conditions that specific trade is entering in.
DPS is also not a substitute for a good strategy. A weak strategy with dynamic sizing is still a weak strategy. The concept amplifies what already works – it doesn’t rescue what doesn’t.
Why Static Position Sizing Leaves Performance on the Table
Once you internalise that trade probabilities fluctuate, static sizing looks like an obvious inefficiency. You’re essentially saying: “I don’t care whether the statistical odds on this trade are 30% or 60% – I’m risking the same amount on both.”
That’s a choice many systematic traders make, and it’s defensible as a conservative default. But it’s not optimal. Trading is a probability game. The foundation of any good systematic approach is identifying when the odds are in your favour and acting accordingly. DPS extends that logic into position sizing, where most traders have never applied it.
I genuinely can’t imagine going back to static sizing at this point. Once you’ve seen the improvement in equity smoothness and net profit to drawdown ratio that DPS produces on robust strategies, trading without it feels like leaving a material amount of performance on the table.
Does DPS Work Across Different Trading Styles?
Yes. The concept is universal. It works for futures traders, forex traders, stock traders, CFD traders. Day trading strategies and swing strategies. The implementation details vary – what constitutes a relevant probability factor differs by market and timeframe – but the core logic applies everywhere.
The key requirement is that you’re trading a systematic strategy with enough historical data to identify meaningful probability variations within the strategy’s trade sample. Without that data foundation, you’re just guessing. With it, you have a genuine basis for differentiated sizing.
For more on building the systematic foundation you need before DPS makes sense, see our Masterclass series starting with Part 1, and our guide on algo trading for non-programmers.
The Core Principle
Trading is always a probability game. The goal is never certainty – markets don’t offer certainty. The goal is to identify when the odds are better than average and act more aggressively, and when the odds are worse than average and pull back or step aside entirely.
Most systematic traders apply this logic to entries and exits. Dynamic position sizing extends it to the one variable most traders leave static: how much they risk on each trade. That extension – when applied to robust strategies with sufficient historical data – produces consistent, measurable improvements in portfolio performance.
Want to apply Dynamic Position Sizing across your strategies without building the framework from scratch? Check out Breakout Trading PRO. It’s a structured, visual strategy builder inside BreakoutOS, powered by the proven Mr. Breakouts Formula used by thousands of traders.