Nelson Freeburg was one of the earliest quantitative researchers in systematic trading, yet most retail traders have never heard his name. He published a newsletter called Formula Research from the 1990s until his death in 2014, building systematic timing models for stocks, bonds, and commodities. His clients included Martin Zweig, Gerald Appel, and Paul Tudor Jones. He charged just $195 per year for the subscription.
In this episode, Linda Raschke, who knew Nelson for over 24 years, returns to BST to discuss his methods, his philosophy, and what modern systematic traders can learn from his approach to model development. This is episode 50, and it is dedicated to his work.
Watch the full episode below, then read on for the complete breakdown.
Who was Nelson Freeburg?
Nelson started his quantitative career modelling thermonuclear war scenarios as a PhD student in world politics. He moved into financial markets from there, eventually building one of the first serious financial databases that went back to the early twentieth century. With that data he built a library of systematic timing models for the stock market, bonds, gold, and currencies.
Linda describes him as one of the very first quants who subscribed to the principle that methods are deficient if they rely on subjective judgment. He never wrote a book. He never had a website beyond the newsletter. When he died suddenly, most of his work scattered across old issues of Formula Research with no central repository. That is partly why Linda felt compelled to dedicate this episode to him.
Paul Tudor Jones was a fan. The two of them grew up in Tennessee. Nelson once presented to the Tudor group how to turn $10,000 into $65 billion through composite models and compounding, over several decades. According to Linda, the Tudor traders were stunned.
The Russell 2000 growth versus value model
Linda walks through one of Nelson’s actual models as a concrete example of how he thought. It is called the Russell 2000 growth versus value model, and it is built around three conditions:
- Relative strength: Take rate-of-change readings of Russell 2000 growth and Russell 2000 value using lookback periods of 5, 15, 25, and 35 bars. Average those four readings. When growth closes above value for two consecutive bars, condition one is met.
- OEX/SPX ratio: The OEX divided by the SPX must cross above its 50-period simple moving average.
- NASDAQ/SPX ratio: The NASDAQ divided by the SPX must be above its 50-day moving average.
When all three conditions are positive, you are long the market. It takes all three turning negative to move to cash. The model generates just one to two signals per year. That simplicity was deliberate. Nelson believed most optimization and complexity was counterproductive.
Nelson’s core model building philosophy
Linda identifies several recurring patterns across Nelson’s work:
- Monetary filters: He consistently incorporated short-term interest rate variables, often using whether yields were above or below a 50-period moving average as a filter layer.
- Trend filter: Almost all of his best models included some variable that required the market to be above its 200-day moving average or equivalent long-term trend indicator.
- Voting systems: He used multi-component approaches where, say, four out of six indicators needed to be positive before triggering an entry. This reduced false signals without adding complexity to any single component.
- ETF switching: He was among the first systematic traders to use ETF rotation models, long before it became common.
- Drawdown as the primary constraint: His number one goal was to reduce drawdown, not maximize returns. His reasoning: reduce the drawdown enough, and you can increase leverage. From there, the return potential is exponential.
The sample size problem and Nelson’s answer
Nelson’s style of modeling creates a genuine tension. Using fundamental variables like sentiment, monetary conditions, and market breadth means you end up with far fewer signals than a pure price-based system. Where Linda might need 300 trades to feel confident in a result, Nelson’s models might produce 20 to 30 signals over a full test period.
His answer to this was that common sense matters alongside sample size. If a model makes logical, economic sense and holds up across different market environments, a smaller sample can still be instructive. He did not claim it was statistically definitive. He thought it was worth knowing about.
Linda challenges listeners to replicate his gold timing model, which was published in a Formula Research newsletter covered by Tom McClellan. The original test ran back to 1950 and was published in 2003. The next decade included an unprecedented global monetary expansion with no historical precedent. How did the model hold up? That is the kind of out-of-sample test that reveals whether the underlying logic survives.
What Nelson’s approach meant for risk
The goal was never to show the highest return. It was to show the most risk-adjusted path to a high return. Linda puts it plainly: any trend-following system producing 15 percent annual returns over time will likely carry a 50 percent drawdown at some point. At least. Nelson’s focus was on finding ways to reduce that drawdown through filters and overlays, then using leverage to amplify the reduced-risk base.
He once demonstrated this to Paul Tudor Jones using five composite timing models combined. The combined model dramatically outperformed any single component. The Tudor group was apparently astonished by the results, though obviously the mathematics involved decades of compounding.
Linda also describes how Nelson used overlay systems, sometimes pairing a longer-term trend following approach with a short-term mean reversion or volatility breakout that would temporarily flatten positions during dislocations. The mean-reversion component might hold for one and a half days on average, close out, and return the trader to their original longer-term position.
What we can actually learn from his work
Linda’s practical advice for anyone wanting to apply Nelson’s lessons:
- Stop optimizing for profit factor or average win. Look at drawdown characteristics: consecutive losses, how long the model spends recovering from peak equity, largest individual losers. Those are where the real risks hide.
- Build your own database and do your own testing. Never just trade what someone published. Even careful researchers make assumptions that do not hold in real markets. Verify everything yourself.
- Look at other people’s models as frameworks, not answers. Nelson took models from other researchers and improved them rather than trading them exactly as published. That approach is still valid today.
- Be skeptical of small sample studies. Linda notes that studies showing positive results over 17 out of 20 years will produce confidence levels around 65 percent, simply due to insufficient sample size. That does not make them worthless, but it means you cannot weight them heavily.
Where to find Nelson’s work
Because Nelson never compiled his work into a book or website, finding it requires some effort. Searching for “Formula Research” and “Tom McClellan” will surface two issues of Formula Research that Tom covered on his own site, including a detailed gold timing model. Searching for Nelson Freeburg with Larry Connors will find a two-part interview. INO markets has some older webinars. The Market Technicians Association in London also has recordings from a presentation he gave around 2012 or 2013.
The key takeaway from this episode is not any single model Nelson developed. It is the discipline he brought to the process: systematic, humble about what the data can prove, relentless about risk reduction, and always willing to learn from what other researchers had done before him.
Related episodes
- Use indicators like a trading champion with Larry Williams
- A new approach to trading volatility with Rob Hanna
- Stop losses with Cesar Alvarez
Get the show notes & transcript
Want to study the foundations of quantitative model development? Subscribe to the Better System Trader podcast for weekly interviews with the world’s top systematic traders and quantitative researchers.
