Marsten Parker: TraderLion Conference 2024 - An Introduction into Systematic Trading
Market Wizard, Marsten Parker walks through systematic trading strategies
About Marsten
Marsten Parker is the systematic trader featured in Unknown Market Wizards by Jack Schwager. He has been "trading for a living" since 1998 and 100% systematic since 2000. Prior to 1998, Marsten had a 13-year career as a coder working for several Boston-area software companies. Before that, he attended the Mannes College of Music in NYC and earned a degree in violin performance. Marsten built and uses his own backtesting software - RealTest - which has been offered to the public since 2021 and is now used by hundreds of people including some other well-known technical and systematic traders.
What is "Systematic Trading"?
It's also known as “Algorithmic Trading” or “Algo Trading”
All trade entries/exits/ size changes are determined by computer code
No human discretion at the individual trade level
The human's role is to design and deploy the trading system
A “system” can be a single strategy or a collection of strategies that run together
Multiple strategies can share a single equity pool (combined compounding)
Advantages of…
Discretionary Trading:
A good discretionary trader knows "what is in play" and can focus on those opportunities. A system might be able to guess "in play” from e.g. unusual volume but won't be as accurate about it.
Similarly “situational awareness” about the overall market or individual names can only be roughly approximated by systematic rules.
Such awareness (assuming it's usually accurate due to trader experience) can also influence position sizing in a meaningful way.
A discretionary trader could still be mostly “rules-based" while using these advantages
Systematic Trading:
Forces exact specification of your trading rules
Allows you to know exactly what past performance of a set of rules would have been (past performance does ensure future performance but neither does past non-performance and I'd rather run something that would have worked recently than something that would not have)
Allows the actual daily trading process to require very little time and to not be influenced by personal circumstances, emotional state, etc.
Systematic Expectations
Think of this as a long-term approach that, if done right, could allow you to slightly beat passive investing
It does not compare to e.g. Minervini-style growth-stock trading or various other discretionary approaches (for those who are able to master them) in earning potential
He mentions that he just barely qualified for Jack's “Market Wizard” criteria and only because of the length of my trading history
If trying to decide, consider how you want to spend your time each day for the next however many years (or do both!)
How Marsten Develops and Tests Systems
Good clean data is the #1 requirement - he uses and recommends Norgate
Currently, all of the work is with daily bars, though he used intraday in the past and may do so in the future
Obsessive emphasis on realistic backtests
A backtest run in the future must match today's live trades
Use an iterative and interactive process of think, test, review, repeat while doing research
How Marsten Trades the System
Use RealTest to generate orders for each trading day for the strategies being used, accounting for actual current account balance and open positions
Use OrderClerk (a companion program that is available to RealTest users via the forum) to place orders (via IB TWS API), obtain fills, and maintain the round-trip trade history
Trading itself takes less than 5 minutes per day (it could be 0 but prefers to require some direct involvement for risk management)
Continually monitor strategy performance, tweaking or replacing them as seems fit (this is his discretionary layer)
RealTest Software
First coded in the late 1990s as a way to test alternative exit rules for a list of discretionary entries
Gradually evolved over the following 20 years with a few other traders using it here and there
A major rewrite was started in 2017 to support multiple strategies, better formula syntax, a full set of helper functions, etc.
Development is ongoing with frequent software updates and an active forum community helping me know what to focus on
See RealTest for more information
Examples of Strategies:
The 3 example strategies below contain:
the script (at least the part that could fit)
the test result summary row
the equity curve graph (often with SPY benchmark for comparison)
Interesting Strategy Examples:
Uses the IBD Relative Strength formula as its entry and exit triggers.
Once a month, you take all current constituents of the NASDAQ100 and rank them on the relative strength factor and will hold the top 5 ranked stocks.
Combines a variety of strategies in order to diversify risk profiles
The huge decline in the blue line occurred when the Mean Reversion Short Strategy shorted GME 0.00%↑ in 2021 and made one of the worst short trades possible.
Marsten includes this example because there is a hidden tail risk in mean reversion that you always have to keep an eye out for. He also uses this example as a way to explain how a systematic trader could turn a strategy on and off based on the equity curve of a particular strategy (ie. if the equity curve drops below its moving average the strategy would be turned off)
Summary Thoughts
Systematic trading is not a path to quick wealth, but can be a way to ensure long-term consistency in your trading process and (maybe) its outcome
Unless you enjoy the activity of strategy research more than you enjoy hunting for specific trade opportunities, it's probably better to stay discretionary
Remember to focus on discovering what doesn't work (and why), not just on looking for what does seem to work
Q&A Section:
How do you come up with ideas for different strategies that you want to eventually test and use?
Like many successful traders, he used strategies that already existed and tweaked them to fit his needs. Remember, it’s not about reinventing the wheel but making small improvements on systems already known to most traders.
When designing a system, what metrics do you optimize for?
For some systematic traders, there will be complex metrics that are used. For Marsten the most important metric is Net Profit. His rule of thumb is to have a strategy or combination of strategies that have a 20% annualized return with no more than a 20% drawdown. In other words, an MAR ratio of 1 or more.
A MAR ratio is calculated by dividing the compound annual growth rate (CAGR) of a fund or strategy since its inception by its most significant drawdown. The higher the ratio, the better the risk-adjusted returns.
How do you limit the overfitting of a strategy on the data you are using?
The most likely way to overfit a strategy is to develop the strategy for a single symbol and use a lot of parameters and run a brute force optimization. For example, I could take SPY 0.00%↑ and throw a ton of parameters at it, and find something that would have been spectacular but utterly meaningless about the future.
The first level of defense against over-optimization is trading a portfolio with the same parameters. For Marsten, he throws out the notion that individual stocks have individual personalities which for most discretionary trade (including myself) would say is probably not true. Marsten acknowledges this but states it's impossible to capture individual personalities systematically. Using Walk forward optimization is a method to help determine if the parameters you set are stable.
Walk forward optimisation is a process for testing a trading strategy by finding its optimal trading parameters in a certain time period (called the in-sample or training data) and checking the performance of those parameters in the following time period (called the out-of-sample or testing data).
What are your favorite data visualization techniques and use cases to improve strategies when backtesting?
Drawdown % Graph - You would be able to determine what years drawdowns have occurred during a backtest and can dig deeper into why that is the case.
Consecutive Days in Drawdown Graph - This graph shows you how many consecutive days you would be in a drawdown during the strategy implementation. You can imagine if you were in a drawdown for more than a year, you could be questioning if the strategy is working properly.
Expectancy - For shorter-term mean reversion strategies, Marsten makes sure expectancy is high enough for the strategy to be worthwhile.
In designing different systems do you think buy-side rules, sell-side rules, or vehicle selection are the most important aspects of a strategy?
They all go together in some sense. Vehicle selection in systematic trading is more about the technical concept of your trading system rather than anything else. A mistake some systematic traders make is they will hand-select a bunch of symbols that have done great recently and then backtest them in the past. Would you have selected them five years ago? Probably not, as you would have taken whatever was hot or trending at that time. And that’s one of the key advantages of discretionary trading. You can have situational awareness that lets you know what's in play or not. Whereas in systematic trading it is a tough concept to capture in code.
For entries and exits, it depends on the strategy you are trying to implement. It’s a good idea to use ATR multiple to set stops for more or less volatile stocks.
It’s also important to keep look-back lengths consistent with how you want to trade. If you're trading a strategy that holds for one to three days, it doesn't really matter what the 50 or 200-day average (generally speaking)
How often do strategies lose their edge? How long does a typical strategy seem to work?
The more parameters your system has the higher the chance that it will stop working because it’s past performance depends on what the parameter settings were.
Sometimes the underlying concept of a strategy that used to work doesn’t work anymore. For example, Marsten used to trade an IPO strategy but because there are new methods of how a company can enter the public market, the edge for that strategy has diminished.
How important do you think it is for systematic traders to understand discretionary styles? Do you think people should trade discretionally before they dive into becoming a systematic trader?
I would say so. At least for some amount of time or at least spend some time watching the markets and having an interest in the markets in general. I think everybody should try to day trade at some point, preferably with a small amount of money. It's important to get a feeling for how price moves. I’m not a big fan of paper accounts other than checking for bugs in automation systems.
Some discretionary traders have tried to systematize what they do and it often doesn't work because I believe that situational awareness is the number one factor for discretionary trading success. It’s more about when to use the setup than what the setup is. Because even if you backtest a setup you're backtesting it on every possible instance which for a systematic trader will yield unfavorable results. When a setup requires some form of visual judgment, coding that judgment would be extremely difficult as well. In those cases, you would look for a proxy, where you can formulate a rule that would encapsulate the vast majority of cases where the visual judgment confirms the trade.
Index of TraderLion Trading Conference 2024:
Day 1:
Marios Stamatoudis: The Two Layers for Building Trading Mastery
Marsten Parker: A Survey of Systematic Trading Strategies
Dr. Steenbarger: Turning Personal Strengths into Trading Strengths
Pradeep Bonde: Swing Trading Catalysts and Momentum Bursts
John Burns: How to Tap into Your Intuition – Turning Your Mental Game into a Trading Edge
Roy & Wes Mattox: Tracking Market Trends and Market Environments
John Pocorobba: Building Confidence Trading Earnings Gaps