Most traders backtest their strategies as if market data is a homogeneous set. Lawrence Chan thinks that’s a fundamental mistake. In episode 204 of Better System Trader, I spoke with Lawrence, a Toronto-based trader and algorithmic system developer who started as a floor trader three decades ago, about how understanding structural market biases can turn a mediocre strategy into a genuinely robust one.
Lawrence has traded S&P 500 futures, e-mini contracts, Forex, bonds, and a range of other futures markets. He’s built first-generation market-making algorithms, institutional buy-side and sell-side systems, and now runs his own proprietary trading firm. That depth of experience across different market contexts, spanning the open-outcry era through electronic trading and into the current retail-driven volatility environment, gives him a perspective on market behavior that’s hard to find elsewhere.
Watch the full episode below, then read on for the complete breakdown.
Why your backtest is lying to you
Lawrence’s central argument is that backtesting across a full historical data set as if conditions were uniform throughout produces misleading results. Markets go through distinct phases, and the behavior within each phase is different enough that treating them as one data set amounts to averaging out information you should be acting on.
He describes the market cycle in terms of economic booms and busts, each lasting nine to ten years, with smaller cycles nested inside them. At the start of a bull market cycle, price behavior operates one way; in the late stages of the same cycle, it operates differently. A strategy optimized across the full period may capture neither phase well, while a strategy designed with awareness of where you are in the cycle can perform significantly better.
The dotcom bubble of 1999-2000 changed market dynamics temporarily. The GFC of 2008 did the same. And in the period Lawrence is speaking from, post-COVID retail participation and the proliferation of commission-free trading apps are creating dynamics that traders who learned markets in the 2010s would find unfamiliar. “For people who have traded for several years and got used to the rhythm from a few years ago, it will drive them crazy.”
The atomic strategy concept
One of the most practically useful ideas in this conversation is what Lawrence calls an “atomic strategy.” An atomic strategy is stripped back to a single, clear behavior: if it goes long, it knows exactly when to go long and exactly when not to engage. The failure mode he’s addressing is over-complexity: strategies that have so many conditions, filters, and parameters that they become difficult to reason about and prone to overfitting.
The other half of the atomic concept is knowing when to sit out. “The important thing is that you need to know, for your particular strategy, when will it not work.” A simple moving-average crossover system will produce 40% to 60% winners almost regardless of what you do with it. But if you know the market conditions under which it performs worst, say, the final stage of a bull market with elevated retail participation and extreme volatility, you can simply not trade it during those periods.
The improvement is not theoretical. Lawrence describes winning rates swinging from 40-60% in neutral conditions to 60-80% during the middle phase of a bull market, purely by filtering for market context. That’s not a marginal edge improvement; it’s the difference between a struggling strategy and a profitable one.
Understanding structural bias in markets
Lawrence distinguishes between several layers of structural bias that affect price behavior. At the longest scale, there are decade-long economic cycles driven by interest rate regimes and capital allocation trends. In the current environment, a prolonged period of low interest rates has pushed all major asset classes upward, creating a pervasive long bias that shows up in almost any backtest across the past decade. “Over the last twelve to thirteen years, it’s relatively difficult to get a good model that doesn’t have a long bias.”
At shorter scales, there are recurring calendar-based biases: monthly effects, options expiration dynamics, end-of-quarter rebalancing, and seasonal patterns tied to economic cycles. Lawrence’s firm, DayTradingBias, focuses specifically on identifying and exploiting these recurring structural patterns in futures markets.
The key insight is that retail traders can exploit some of these biases as freely as institutions can, particularly the shorter-term calendar effects, because they’re structural features of how markets are organized rather than informational advantages. You don’t need better data or faster execution to benefit from options expiration dynamics; you need to understand when they occur and what they tend to do to price behavior in the instruments you trade.
How retail participation changes the game
Lawrence frames the explosion of retail trading in 2020-2021, driven by commission-free apps, stimulus checks, and time at home, as a recurring pattern rather than a unique event. He draws direct parallels to 1999-2000, when the first generation of internet-based retail trading created similar dynamics. “After so many years of retail not participating, you have a whole new generation of younger investors, traders, whatever you want to name them, but they’re new to the market.”
The practical effect is increased short-term volatility, driven by retail traders making extremely fast, low-conviction decisions. “They will press the button. Five seconds later, they see something feels bad, they will press another button to get out.” This adds volatility that has nothing to do with the underlying fundamental or technical picture. Strategies that were calibrated to a lower-volatility environment can break simply because the noise level has increased, not because the underlying edge has disappeared.
The analogy he uses is poker: “You need to know your opponents.” If the table is full of inexperienced players making impulsive decisions, that’s actually an exploitable situation, but only if you’ve adjusted your approach to match the conditions, rather than playing the same way you would against experienced opponents.
The most toxic period to backtest
Lawrence identifies a specific period as particularly dangerous to include in backtesting without careful segmentation: the post-2008 environment of sustained quantitative easing and zero interest rates. The monetary conditions during this period were historically unprecedented, and the market behaviors they produced were correspondingly unusual. Strategies tested primarily on post-2008 data may have very high performance metrics that are entirely artifacts of that specific monetary regime.
His approach is to segment data by market regime before backtesting and to treat performance in each segment separately. A strategy that works well across all regimes, both the QE era and the pre-2008 period, is genuinely robust. A strategy that only works in one of them is making a bet on that regime continuing.
Simple tools and thinking in concepts first
Lawrence’s approach to building strategies starts not with indicators or code but with a hypothesis about market behavior. “Think in concepts first.” What is the structural reason this pattern should exist? Why do buyers and sellers behave in this particular way at this particular time? Only after answering those questions does he go to the mechanics of testing the idea.
This reverses the process most retail traders use, which starts with backtesting readily available indicators and works backward to try to understand why they might be working. Lawrence’s concern with that approach is that it tends to produce systems that are data-fit rather than logic-driven, and those systems tend to stop working when conditions change.
Related episodes
- How to Find Statistical Edges in Trading – Scott Andrews
- How to Avoid Curve Fitting Using Independent Testing – Jeff Swanson
- Seasonality – Jay Kaeppel (Episode 34)
- Game Theory in Trading – Brian Miller (Episode 118)
Want to build strategies that adapt to changing market conditions? Subscribe to the Better System Trader podcast for weekly interviews with the world’s top systematic traders and quantitative researchers.
