A listener named Paul sent in a question after our episode on Dynamic Position Sizing that turned out to be far more common than I expected. He’d been trading for two years, was familiar with Ed Thorp’s card-counting principle of “when your edge increases, bet more,” and figured DPS was just a formalised version of what he was already doing intuitively.
He was wrong – but not in a way that should embarrass him. The confusion between Dynamic Position Sizing and Martingale-style betting is widespread, even among experienced systematic traders. This episode is my attempt to draw a sharp, clear line between the two concepts, because conflating them is genuinely dangerous to your capital.
Paul’s question also raised a second issue worth unpacking: the temptation to think that “knowing when to break your rules” is a legitimate trading edge. It isn’t. And understanding why gets to the heart of what makes algorithmic trading different from discretionary trading.
What Paul Was Actually Describing: Martingale Position Sizing
Paul described varying his position size based on a confluence of factors – a buy signal, a 10-trade losing streak, strong support levels, the S&P near all-time highs, and so on. In his mind, these factors combined to signal an “increased edge,” so he’d bet bigger.
What he’s describing is Martingale-influenced thinking: increasing your position size after losses or when you perceive conditions to be favourable. The classic Martingale approach doubles your bet after every loss, with the idea that eventually you’ll win and recover everything. It comes from casino gambling – specifically roulette – and its history in trading is catastrophic.
I’ve seen a major hedge fund manager – someone with billions under management, genuinely respected in the industry – lose everything twice because of Martingale. He admitted it post-mortem. His strategies were solid. His risk management wasn’t. The second time, he lost even more than the first because he’d rebuilt the fund and had more to lose.
The core problem is this: in card games, the deck is finite. You can calculate remaining probabilities with reasonable accuracy. In markets, there is no fixed deck. Your back-test might show a worst-case losing streak of 10 trades. Monte Carlo analysis might push that to 16. But when you go live, 16 can become the new normal – and if you’ve been increasing your size expecting reversion, two more losses can wipe you out.
Why “Breaking the Rules” Is a Red Flag
Paul’s framing of his position sizing as “breaking the rules when it makes sense” tells me something important about where he is in his development as a trader. I say this directly, not to be harsh, but because it matters: rule-breaking is not a feature. It is not sophistication. It is the opposite of what makes algorithmic trading work.
Mature trading means having rules that are so thoroughly back-tested and statistically validated that you never need to break them. If you find yourself overriding your system based on gut feel or a confluence of discretionary signals, that’s not edge – that’s noise dressed up as insight.
Anti-Martingale is the more sensible alternative for traders who want to size dynamically. Rather than doubling down after losses, Anti-Martingale increases size after wins and reduces it after losses. It’s still not what we teach as Dynamic Position Sizing at BTA, but it’s at least not systematically destroying your capital during drawdowns.
What Dynamic Position Sizing Actually Is
Here’s the concept explained simply.
When you back-test a strategy over 1,000 trades, you might find that, on average, 60% of trades are winners. That 60% is an average. It is not constant. At different points in time – depending on market regime, volatility, instrument behaviour – that probability might be running at 30% or pushing toward 70% or 80%. It fluctuates.
The key insight behind DPS is this: rather than applying a filter to remove the low-probability periods (which reduces your sample size and increases curve-fitting risk), you research why the win rate fluctuates. You identify the specific, statistically validated conditions that correlate with the strategy performing above or below its average win rate. Then you build an algorithm – not a discretionary override – that adjusts position size accordingly.
When conditions indicate the strategy is likely in a 70%+ win-rate regime, you increase size. When conditions indicate it’s likely in a 30% regime, you decrease size. This is done through exact rules, back-tested with robust methodology, with no guessing and no exceptions.
Why Not Just Filter Out Low-Probability Trades?
This is a question that came from another student in the forum: if you can identify low-probability periods, why not just skip those trades entirely?
Two reasons.
First, adding a hard filter to skip trades increases the complexity of your system, which means higher overfitting and curve-fitting risk. Every additional rule you add is another opportunity for the back-test to “learn” patterns that don’t persist out of sample.
Second, sample size. I am extremely firm on the importance of maintaining large, statistically valid trade samples. When you start layering in filters to remove trades, your effective sample size drops dramatically. At that point, your statistical confidence in the strategy’s edge is undermined. The strategy looks better on paper – because you’ve removed losing trades – but it’s far more likely to fail in live trading.
Dynamic Position Sizing sidesteps both problems. You keep all the trades, preserving sample size. And instead of a binary in/out decision, you scale size continuously based on statistically valid inputs. That’s how we approach it in the Breakout Masterclass.
Dependent vs. Independent Inputs: The Overfitting Safeguard
One of the more advanced concepts inside the DPS programme is the distinction between dependent and independent inputs when building your position-sizing rules.
This matters because if your position-sizing inputs are too closely correlated with your base strategy’s entry logic, you’re essentially just re-optimising the strategy under a different label. That creates the illusion of improvement without genuine additional edge. Independent inputs – factors that aren’t already baked into your strategy’s logic – are what give DPS its real statistical validity.
Getting this right is what separates genuine Dynamic Position Sizing from sophisticated-sounding Martingale. If you’re in the Masterclass programme, this is covered in depth in the DPS workshop.
The Monte Carlo Reality Check
Here’s something I’ve seen repeatedly across years of back-testing and live trading: Monte Carlo’s worst-case scenarios have a tendency to become the new normal once you go live.
Static back-test says your worst losing streak is 10 trades. Monte Carlo pushes that to 16. You go live thinking 16 is the extreme outlier you’ll never see. Then 16 becomes routine. You’ve experienced it twice and it barely registers.
This isn’t a bug in your strategy – it’s just the nature of live market conditions versus historical data. Markets evolve. The point is that any position-sizing approach that assumes historical worst-case is the actual worst-case is built on a false premise. Martingale exploits that false premise in the most dangerous way possible.
Dynamic Position Sizing is not built on that premise. It’s built on the observed, statistically valid fluctuation of win probability – and it adjusts in both directions, scaling down when conditions are unfavourable, scaling up when they’re favourable, all within a firmly defined algorithmic framework.
Where Paul Should Start
If you’re reading this and you recognise yourself in Paul’s description – trading discretionally, adjusting size based on feel, thinking of it as “knowing when to break the rules” – the first step is simple: move from Martingale thinking to Anti-Martingale thinking.
That means reducing size after losses, not increasing it. It means getting defensive, not aggressive, when your system is in a losing streak. It means respecting that a back-test losing streak of 10 might become 16 or 20 in live trading, and your risk management needs to account for that.
From there, if you want to explore genuine Dynamic Position Sizing, start by studying the Breakout Masterclass framework for building robust strategies. DPS is a layer that sits on top of that foundation. It doesn’t substitute for robust strategy development – it enhances it.
The distinction matters enormously. One approach has put hedge funds out of business. The other is a statistically grounded, rule-based system for extracting more from the genuine edge you’ve already identified in your strategies.
Want to see Dynamic Position Sizing applied to real breakout strategies – not just theory? 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.