Marsten Parker traded full-time for 15 years, building an enviable equity curve, before watching it all nearly collapse in a 45% drawdown between June 2015 and January 2016. He almost quit. He reconsidered the past two decades of his career. Then he came back, built a more conservative approach, and added it to the unknown market wizards roster in Jack Schwager’s 2020 book. This episode is about how he did it, and what every systematic trader can learn from that process.
Marsten has been trading US common stocks exclusively since 1998, holding positions for a few days at a time, typically running around 1,000 round-trip trades per year. He also built his own backtesting software, which he now markets publicly. Over 20-plus years he has traded breakout strategies, mean reversion, and short-term trend following, and adapted his approach each time the market changed around him.
Watch the full episode below, then read on for the complete breakdown.
The 45% drawdown and what it actually feels like
The equity curve shown in Marsten’s Schwager chapter looks smooth from a distance. But that smooth curve includes a 45% drawdown that is jarring up close. It took seven months to play out, and it did not drop in a straight line. There were partial recoveries followed by further deterioration, which made it harder to cut.
The cause was a concentration decision. Marsten had traded mean reversion in parallel with his original breakout strategy for all of 2014. The mean reversion results were so consistent that he dropped the breakouts and went all-in on mean reversion in 2015. He also added some leverage. When both the long and short sides hit bad periods simultaneously from June 2015 onward, there was no buffer.
The psychological impact was significant. He had left his career as a software engineer in 1997. By the time of the drawdown he was in his mid-50s and had not worked for nearly 20 years. He explored going back to software work and found it impractical. He then did what many traders do after a major loss: he stepped away entirely, cancelled data subscriptions, moved most of his capital to a savings account, and waited.
What brought him back was not a specific catalyst. A few months passed. He kept thinking about systems and trading. He opened the backtesting software again, started experimenting, and found the underlying mean reversion strategies still worked. He had just traded them too aggressively.
System stops: the rule that protects the account
Marsten had always had a 20% system stop in the back of his mind: a level of portfolio drawdown that forces a pause and a reassessment. He had observed this rule in the past. During the 2015-2016 drawdown, he did not follow it. He kept expecting recovery, and the drawdown compounded.
When he came back, he made the stop more conservative. The new rule: stop trading if the portfolio loses 10% from inception, or 15% from any future peak. He also wrote a formal trading plan, something he had never done in 20 years of trading. Having a written plan, knowing the specific number at which he would stop and reassess, reduced the emotional uncertainty of being in a drawdown.
Since returning, that 15% from peak stop has come close but has not been triggered. Having the threshold defined in advance, rather than relying on willpower in the moment, is what makes the difference.
How Marsten’s trading styles evolved over 20 years
Three constants ran through all 20 years: US common stocks only, both long and short, short-term swing holding periods.
The style evolution went through four phases:
- Breakout entries at next open (1998-2004): Scanning for setups after the close, placing orders for the next open. The classic William O’Neill framework applied systematically.
- Breakout entries near prior close (2004-2005): Marsten found that entering before the close rather than waiting for the next open captured more of the move. This worked until it stopped working, which he traced to mutual funds shifting from block trades executed by human traders to algorithmic execution around 2005.
- Intraday breakout entries (2005-2013): Entering during the breakout day rather than waiting for the close or next open. This required real-time data monitoring across many symbols. Worked well until the short side broke in 2012 as buy-the-dip behavior became dominant.
- Mean reversion plus diversified systems (2013-present): Added mean reversion after the breakout short side died. Currently running six strategies: mean reversion long, mean reversion short, newer versions of short-term trend following, and two mean reversion day trading strategies added in mid-2020.
Why mean reversion shorts are dangerous
A common mistake Marsten sees is traders who build mean reversion long strategies using an uptrend filter (e.g., price above the 200-day moving average) and then simply invert the logic for shorts, requiring a downtrend. The problem: stocks in downtrends attract value investors and dip buyers. A short-term rally in a weak stock is a buy signal for certain traders, not a short signal.
What actually works for mean reversion shorting is targeting stocks in strong uptrends making blow-off tops, the opposite of the downtrend approach. But this creates its own risk. The strategy works consistently most of the time, then catastrophically fails when a short-term squeeze or gamma squeeze hits a heavily shorted high-momentum name. His backtesting software includes an example mean reversion short strategy that would have been destroyed by the GameStop situation in January 2021.
The mitigation: keep position sizing modest on the short side. Do not size short mean reversion the same way you size long mean reversion.
Diagnosing a broken strategy: what actually signals it
Marsten is honest that there is no clean answer. He has never had a fully systematic rule for declaring a strategy broken, and some of the clearest cases he has encountered still produced uncertainty.
The signals he looks for:
- A drawdown exceeding the worst drawdown in the backtest. Not definitive on its own, but a serious red flag.
- Multiple stats moving in the same direction over an extended period, win rate, average trade, profit factor, all degrading simultaneously rather than just one metric.
- The environment the strategy was built to exploit has visibly changed. His 2012 short strategy breakdown coincided with a clear, identifiable shift in market behavior: institutional buy-the-dip activity becoming dominant.
He also distinguishes between strategies that are broken and strategies that are out of regime. His example: a breakout strategy he stopped running between 2010 and 2019 because it only made 3% annually with 17% max drawdown. He rejected it. Then 2020 happened, and the same strategy performed strongly. A mediocre 10-year backtest followed by one exceptional year creates exactly the ambiguity that makes this decision hard. Is it a regime change, or a temporary blip?
Equity curve trading as a risk filter
One of the most practical tools from this episode is using a moving average of a strategy’s own equity curve as an on/off switch. Marsten applies a 300-day moving average to the equity curve of each strategy. When the equity curve falls below its 300-day average, the strategy is turned off.
He tested this on a mean reversion short strategy spanning 1995 to the present. The raw strategy had two terrible periods: a 50% drawdown in 1999, and a severe loss in early 2021. With the 300-day equity curve filter:
- The 1999 drawdown was reduced by approximately half
- The 2021 GameStop-related drawdown was avoided entirely
The specific lookback of 300 days is not precise. Marsten tested similar values and found the results were comparable across a range. The concept is more important than the exact number. Using the strategy’s own performance history as a regime filter prevents you from running a broken or out-of-regime strategy indefinitely.
How Marsten uses optimization
He builds backtesting software with a full optimizer built in, and almost never uses it in the traditional sense. He does not run parameter sweeps looking for the optimal value. Instead, he uses the optimizer to check robustness: he picks a key parameter, runs it across a range, and confirms there is a broad plateau of acceptable results rather than a narrow peak.
In practice he often just manually types three or four parameter values and reruns the test. His software runs fast enough that this interactive approach works better for him than formal optimization. The goal is to stay out of a flow-state-disrupting wait and to avoid the overfitting that comes from chasing peak parameter values.
The pace of markets speeding up
One insight from Schwager’s chapter on Marsten: he noticed the pace of markets accelerating. His evidence was direct and specific. Entry timing that worked entering at the next open stopped working. He moved to entering near the prior close. That edge eroded. He moved to entering earlier in the breakout day. That worked until about 2005, which coincidentally aligned with when mutual funds made the transition from human block traders to algorithmic execution.
He sees the same dynamic continuing. 2020 and 2021 brought extremely fast, high-amplitude short-duration moves. Adding shorter-term day trading strategies to his portfolio in mid-2020 proved timely.
Related episodes
- How to detect a failing trading strategy with Kevin Davey
- How to detect a failing trading strategy (updated) with Kevin Davey
- Combining algos using state-based market design with Richard Metzger
Want to build more resilient trading systems?
Subscribe to the Better System Trader podcast for weekly interviews with the world’s top systematic traders and quantitative researchers.
