Howard Bandy has been working on quantitative trading systems for longer than most traders have been alive. His background spans mathematics, physics, engineering, and computer science, with graduate work in artificial intelligence and pattern recognition at a time when those fields were just getting started. He spent decades as a research analyst for a CTA firm, managing real money through systematic models, and then wrote five books on quantitative trading systems.
In episode 6, Howard laid out a frank assessment of where the markets are now and what that means for system developers. His view: trading is getting harder, the easy inefficiencies are gone, and most retail traders are competing in an arena where they’re seriously outgunned. But there’s a path through it, and it starts with understanding why systems go out of sync with the markets they were built to trade.
Watch the full episode below, then read on for the complete breakdown.
The disappearance of large market inefficiencies
Howard opens with a point that a lot of system developers don’t want to hear: the large inefficiencies that made systematic trading so profitable in the 1970s and 80s are mostly gone.
The Turtle traders are the clearest example. Their rules are now public knowledge, published by former members. If you test those rules on data from when they were actively trading, they work well. Test them from the mid-1990s forward, and the performance disappears. Howard’s interpretation is straightforward: once the rules became widely known, enough traders applied them to arbitrage out the inefficiency. Each winning trade removed some of the edge the system was designed to capture.
This is an ongoing process. As computers became cheaper, data more available, and trading platforms easier to use, more participants entered the market looking for the same patterns. The barriers to entry are low and the rewards for being right are high, which means competition for any discoverable edge is intense and growing.
The practical implication: what’s left are smaller, shorter-lived inefficiencies. Where the old turtle-style trades might have captured 10 to 30% moves over weeks or months, the opportunities Howard sees now are more like half a percent to one percent over days. That’s not impossible to exploit, but it requires a different approach and much tighter execution.
Why your trading system goes in and out of sync with the market
This is the central concept of the conversation, and Howard explains it clearly.
A trading system is a model. It contains logic designed to identify specific patterns or conditions in price data. Once built, that logic is relatively fixed. The market, however, is not fixed. It changes constantly, driven by economic conditions, fundamental news, shifting participation patterns, regulatory changes, and dozens of other factors.
Most analysis techniques assume the data is stationary, meaning the statistical properties of the data don’t change over time. Financial data is notoriously non-stationary. The patterns that worked in one regime don’t necessarily work in the next.
Howard compares it to a pendulum. When your system and the market fall out of sync, the question is whether this is temporary, meaning the market will swing back toward conditions your system handles well, or permanent, meaning the regime has genuinely shifted. In the temporary case, you wait. In the permanent case, you stop trading that system.
The problem is that you usually don’t know which it is when you’re in the middle of it. That’s what makes system management so difficult.
Why large inefficiencies won’t come back
Howard makes a point worth sitting with. When a trading firm brings on a new quant, one of the early training exercises is often to implement historical systems, including the well-documented ones, and test whether they’re still profitable. If anything is found to be working, it gets traded immediately, which removes whatever inefficiency remained.
This means that even if a formerly profitable pattern temporarily returns, it’s likely to be traded out again quickly. The combination of institutional resources, systematic monitoring, and low execution costs means the window to exploit any rediscovered edge is very short.
Howard’s phrase for what’s left is “abandon financial astrology.” Stop relying on indicators and patterns that became widely known years ago. The edge has been extracted. What matters now is finding something genuinely different, validating it rigorously, and accepting that it probably won’t last as long as the old approaches did.
The role of in-sample and out-of-sample testing
Given that financial data is non-stationary, Howard is emphatic about the importance of out-of-sample validation. This isn’t just good practice; it’s the only honest way to know whether you’ve found a real signal or just fit a model to noise.
The process he describes:
- Identify an objective function that defines what a good trade looks like for you specifically
- Use in-sample data to search for patterns that satisfy that objective
- Validate on out-of-sample data the model never touched during development
- Treat the out-of-sample results as your best estimate of what live trading will look like
He’s careful to note that even good out-of-sample performance isn’t a guarantee. The market regime can shift between when you validated the model and when you deploy it. This is why ongoing monitoring matters as much as the initial development.
Identifying when a system is out of sync: practical signals
Howard describes the problem of knowing when a system’s drawdown is noise within a normal range versus a signal that the system is no longer aligned with current market conditions. This is genuinely hard because drawdowns are normal, and overreacting to them is a common mistake.
His approach centres on defining acceptable performance boundaries before you start trading, not after the drawdown has already happened. If you’ve done proper out-of-sample testing, you have a realistic picture of what the system’s drawdown characteristics look like. You set thresholds based on that picture and monitor whether live performance stays within those thresholds.
When performance falls outside those bounds for long enough, that’s a signal to reduce position size or stop trading the system while you evaluate whether the edge is gone or just temporarily absent.
The two most important skills in system development
Howard names two capabilities that he sees as genuinely differentiating among system developers.
The first is statistical rigor. The ability to design a proper test, avoid data snooping, set aside out-of-sample data before you begin, and evaluate results honestly rather than searching for confirmation of what you want to believe. This sounds basic, but most published trading research doesn’t meet this standard, and most retail system developers work to a lower bar still.
The second is the ability to model risk accurately. Howard’s work on Monte Carlo analysis for trading systems reflects this directly. Knowing what your system’s drawdown profile looks like under different market conditions, and sizing positions accordingly, is what separates traders who survive bad stretches from those who blow up during them.
His book Quantitative Technical Analysis, published in January 2015 at the time of this episode, covers risk normalization and a technique for adjusting position size trade-by-trade based on recent system performance. The idea is to scale down when the system is performing poorly relative to its historical baseline, and scale up when it’s in a good period. This is a systematic way to respond to the in-sync versus out-of-sync problem.
Why trading is harder than most people think
Howard uses a sports analogy that I think is worth repeating. When a beginning golfer or tennis player enters a tournament, they can enter events matched to their skill level. There are handicaps, challenger circuits, categories that let people compete against others at similar levels.
In trading, there’s no such structure. When you press buy, you’re on the other side of a trade from someone who might be running a well-funded, highly sophisticated institutional desk with deep analytical resources and much better execution. There are no challenger tournaments. No handicaps. No mulligans.
Howard isn’t saying individual traders can’t compete. He is saying that going in with the assumption that it should be easy, or that yesterday’s approaches still work, is a path to getting taken apart by people who are much better equipped. The starting point has to be an honest assessment of the competition you’re entering.
Get the show notes & transcript
Related episodes
- Episode 5: Kevin Davey on trading system development, backtesting, and building winning systems
- Episode 3: Cesar Alvarez on trading ideas, backtesting, stops and market timing
- Episode 42: Murray Ruggiero on system premise and robust strategy development
Want to learn more about building better trading systems? Subscribe to the Better System Trader podcast for weekly interviews with the world’s top systematic traders and quantitative researchers.
