In Part 1 of this conversation, orthopedic surgeon Al walked us through how he built 75 highly robust trading strategies across seven markets in under two months, while running a busy medical practice. In this episode, we go deeper – into the specific techniques he used to upgrade his strategies, how he applied dynamic position sizing, and what he discovered that completely surprised him.
If you haven’t listened to Part 1 yet, start there first. What follows will make a lot more sense.
The Three Problems Al Was Solving
At the point of Part 2, Al had 75 strategies at Robustness Level 2 or above, with 52 at RL3. His immediate goals before going live were:
- Improve average net trade size on strategies where it was below the threshold needed to cover commissions and slippage with margin to spare
- Convert remaining RL2 strategies to RL3 using independent filters (the Smashing False Breakouts approach)
- Apply dynamic position sizing to enhance the strategies that were already working
All three goals are connected. They’re about taking a robust foundation – strategies that have already proven themselves through demanding validation – and extracting more performance from them without compromising that robustness.
The Average Trade Threshold and Why It Matters
Larry Williams’ guideline on average trade sizing says it should be approximately two to three times the cost of commissions and slippage to be robust to execution friction (internet delays, order processing, partial fills). Al found that several of his strategies were producing strong equity curves but with average trades below that threshold.
This is a common situation. A strategy can look excellent in back-testing and still be difficult to trade profitably in live conditions if the average profit per trade is too thin. Any deviation from ideal execution – a slightly late fill, a tick of slippage more than expected – eats into margins that were already too small.
The solution Al applied was using Smashing False Breakouts techniques as independent filters – adding conditions derived from different analytical frameworks than the ones used to develop the original strategy. Applied correctly, these filters don’t just improve average trade size; they improve it without degrading robustness, because they’re targeting genuine market behaviours rather than data-fitting to past results.
Using Independent Filters to Convert RL2 Strategies to RL3
The concept behind Smashing False Breakouts techniques is identifying and filtering out the trades within a strategy where the probability is lower – the “noise” trades that drag down average performance while not contributing meaningfully to the overall edge. By removing these with independent filters (conditions that weren’t part of the original strategy design), you improve signal quality without curve-fitting to the original data.
Al’s approach: split market conditions into quadrants – trending vs non-trending, volatile vs non-volatile – and run the Smashing techniques across each quadrant independently. His test case was a Lean Hogs strategy that looked terrible on its own: flat equity, commissions and slippage eating most of the profit, equity curve that looked like ocean waves.
He applied the quadrant analysis, found the conditions where the strategy’s edge was concentrated, filtered out the rest. What came out the other end was a viable strategy with an equity curve that looked completely different from what he started with.
His first response was: “Congratulations, Al, you’ve managed to curve fit yet another strategy.” So he ran it on a full year of out-of-sample data he’d deliberately held back. The equity slope didn’t change at all. The trades continued performing right up to the present day. A strategy that “shouldn’t have worked at all” became viable – not by chance, but because it was capturing something real about how that market behaves.
Dynamic Position Sizing: The Blackjack Analogy
Al came to the conversation having read about Ed Thorp – the mathematician who developed the first statistical analysis of blackjack in 1962 and built a strategy framework around it. Thorp’s key insight: if you have a hand that’s statistically likely to lose, surrender. If you have a pair of aces or eights – a statistical edge over the house – split them and essentially double your position. When you have a statistical edge on doubling down on 11, double down.
The parallel to trading is exact. Dynamic position sizing applies the same logic: add contracts when you have a statistical edge, surrender (skip or reduce) when the statistics suggest the conditions are unfavourable. This isn’t gut feel – it’s systematic analysis of when your strategy’s win probability is above and below average.
For Al’s Crude Oil strategy, Wednesdays and Thursdays showed statistically stronger performance. The reasoning is fundamental: oil inventory reports come out on those days, creating directional movement that breakout strategies can capture. Al applied dynamic position sizing on those days, sizing up when the underlying market logic supported higher probability of a winning trade.
The result: a strategy that was at RL2 converted to RL3. Not by tweaking the entry or exit rules, but by varying how much risk was taken on each trade based on statistically justified probability differences.
A Note on the Risk of Overfitting
The overfitting concern comes up every time dynamic position sizing is discussed. Al addressed it directly: “In doing that, it actually converted the Crude strategy into an RL3. And because not every contract and not every trade is being doubled, you’re not doubling your drawdown and doubling your margin requirements. You’re actually, surgically, going in and removing what you need to, and enhancing what should be enhanced – and there’s an underlying reason in the market that’s justifiable for doing that.”
That last point is critical. The oil inventory report dynamic is a real, persistent market mechanism. Using it as a basis for sizing decisions is not the same as finding a random pattern in historical data and fitting a rule to it. The presence of a fundamental, logical explanation for why the pattern exists is one of the key checks against overfitting.
The other check is robustness testing. You can only justify DPS on top of strategies that have already passed demanding validation. Applying it to poorly tested strategies amplifies problems. Applied to robust, validated strategies with a logical underpinning for the probability factor, it amplifies genuine edge.
The Insight About RL2 Strategies in a Portfolio
Al’s goal had been to convert all his RL2 strategies to RL3. This prompted an interesting conversation about whether that’s actually the right objective.
In our hedge fund’s strategy database – which runs to hundreds and hundreds of strategies – we’ve consistently found that RL2 strategies as a group perform very well. While RL3 strategies individually have higher success rates, a portfolio that includes one or two RL2 strategies isn’t weakened by them. In fact, if an RL2 strategy has low correlation to the rest of your portfolio – if it zigs when everything else zags – it can be a more valuable portfolio component than a higher-rated strategy in a correlated market.
Al had already discovered this empirically. He described putting a flat, uninteresting strategy into a portfolio simulation out of curiosity. In the areas where it generated modest positive returns, none of the other strategies were doing anything. The combined portfolio looked better with the weak strategy included than without it. The strategy that looked useless in isolation was actually the best correlation hedge in his portfolio.
This is the core lesson of portfolio construction: the unit of performance is the portfolio, not the individual strategy. Your job is to assemble a set of strategies whose imperfections cancel each other out – not to find a set of individually perfect strategies.
Building Probability-Based DPS Within a Trade
Al took the dynamic position sizing concept one level further in a way that genuinely surprised me. He described using DPS within a single trade – adding contracts during the trade based on real-time assessment of the trade’s probability.
The example: a strategy has a 75% historical probability of a $10 win, with a $5 stop. That’s a 2:1 reward to risk ratio. If the trade moves in your favour to a point where a clear pivot low forms with increased volume – indicating buyer support – the remaining risk is now only $2. Al’s approach: at that point, double the contracts. You’ve reduced your effective dollar risk (two contracts at $2 stop vs. one contract at $5 stop) while increasing your position size and your exposure to the remaining profit potential.
When coded, this converted a linear equity curve into an exponential one. It’s a legitimate application of the core DPS principle – varying exposure based on real-time probability assessment – applied at the intra-trade level rather than just the pre-trade level.
Al’s 216-Filter Framework
Al mentioned that he started with 88 independent filter conditions derived from Smashing False Breakouts concepts. This expanded to 112 for an article he wrote. At the time of this conversation, he was working with 216 individual filters that he could run against any strategy to immediately identify where the strongest signals were concentrated.
This is where automation and a systematic approach compound. Each filter takes minutes to code and test once you have the framework. Running 216 filters against a strategy produces a clear map of where the strategy’s edge is strongest and where it’s being dragged down by noise. That map tells you exactly where to apply DPS – size up where the signal is strongest, size down or skip where it’s weakest.
His production rate before the framework: one or two strategies per month, each requiring hundreds of hours. With the framework, 75 robust strategies in under two months. That’s the multiplicative impact of having the right tools and process.
Advice for Other Traders
Al’s closing thoughts, condensed:
- Be sceptical – test everything, including what you read here
- Don’t give up when it gets hard; the breakthroughs come after you’ve pushed through the discouraging periods
- The community around a good framework is as valuable as the framework itself; sharing what you know compounds returns for everyone
- Small, incremental improvements across multiple dimensions – entries, exits, position sizing, robustness testing – add up synergistically
- Get involved, share what you learn, and watch what comes back to you
Al’s timeline to paper trading (simulated live): two weeks from the recording date. He had the strategies. He had the portfolio structure. He was ready to see how it performed in real market conditions.
For the foundational concepts Al referenced throughout this conversation, see our trend filter guide and Masterclass Part 2 on validating strategies.
Want to build a live trading portfolio and manage it with confidence under real pressure? 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.