These "must-watch" videos are tagetted at algo traders who are serious about investing their time and research into their own success. They were produced by Trade Like A Machine's Martyn Tinsley for Darwinex.
Martyn Tinsley
This first video of the series embarks on the start of a journey to challenge the backtesting status quo that prevails among algorithmic traders. The video looks at just how difficult it is to distinguish a genuine edge from stochastic (random) results.
Martyn Tinsley
This tutorial looks at how the required sample size can be determined to ensure both a robust optimization, and the effective extraction of the parameters with the best possible edge, using a technique called ‘Statistical Power Analysis’.
Martyn Tinsley
Over-fitting in optimizations happens for two primary reasons: Over-fitting to noise in the price action, and over-fitting to market news events. By attempting to model the overall price action too closely with a large number of parameters means that the model is forced to over-fit.
Martyn Tinsley
This episode starts to look at 6 practical steps that can be taken to improve the reliability of your own trading system optimizations. This starts by improving statistical significance by increasing sample size (the number of uncorrelated trades), and the use of multi-asset and multi-timeframe testing.
Martyn Tinsley
This video looks at practical ways that you can avoid the effects of over-fitting due to Economic News Events. Over-fitting here results in large losing or winning trades that can skew your results significantly. Solutions include enhancing your performance metric and avoiding news events by exiting trades before significant market news is released.
Martyn Tinsley
The video considers how over-fitting to the noise component of price action can destroy a trading system’s ability to work in live trading. It shows how the ‘degrees of freedom’ (or number of parameters being optimized) in a trading optimization process, can cause over-fitting to noise in the price action. A real life illustration of this damaging effect is shown.
Martyn Tinsley
Determining the ratio between the duration of the 'optimization phase' to that of the 'walk forward phase' is critical to a successful optimization process. This episode explains how getting that ratio right, is essential to ensure both robust optimal parameter extraction and robust validation of results. The importance of a ‘Pre-Live’ optimization is also explained, in order to ensure your parameters are in tune with current market dynamics/regimes.
Martyn Tinsley
Simply selecting the best performing set of parameters in an algorithmic trading optimization is often not the best approach. Just because a set of parameter values appeared to perform best in the optimization does not necessarily mean they will perform best in live trading. This episode shows you how to avoid 3 major pitfalls in the selection process by using optimization profiles to best effect.
Martyn Tinsley
Choosing suitable performance metrics/criteria for effective selection of algorithmic trading parameters is essential for backtesting and trading optimizations. Because the choice of performance criteria will impact which parameters you choose to trade with, getting this right is almost as important as the rules of your trading system itself.
Martyn Tinsley
This video covers three techniques to measure system performance: 1. CAGR / ”MEAN” DD (or ‘average’ drawdown), 2. Pearson's Correlation Coefficient, and 3. The Coefficient of Determination. Furthermore, the importance of matching the position sizing methodology (‘scaled’ or 'fixed') to each of these is also fully explained.
Martyn Tinsley
This video uncovers common issues with the use of price data models used in the backtesting process, which can cause significant differences between the results you see in your backtests compared with what you obtain in live trading. The solution to the problems is also explained.
Martyn Tinsley
This video uses two examples of algorithmic trading systems - the first based on a price action strategy and the second based on an indicator system - to illustrate how you can improve your optimization results by choosing a logical range of parameters.
Martyn Tinsley
Ever wondered which metrics work best for measuring algorithmic trading performance in an optimization? If so then these research results will help to answer that question. In scope are the Sharpe Ratio, Profit Factor, Expected Payoff, Recovery Factor, CAGR/Max Drawdown, CAGR/Mean Drawdown, r and R-Squared.
Martyn Tinsley
This research concludes that the optimal way to ascertain the best performance criterion in trading system optimizations is to compare the metric with the Correlation Coefficient of the walk forward equity curve.
More comments taken from the YouTube Darwinex channel:
"That's really awesome content!... Very, VERY useful to see if there's an edge or it is just because of luck and/or overfitting. He's a very good source of valuable knowledge."
"Thanks Martyn. Very informative. Looking forward to more content on this."
"These series of videos are really interesting for two reasons: 1) they address a real and serious problem for algorithmic trading. 2) they are absolutely clear (which is not a common thing). I am eager to see the next videos of this series."
"These videos are purely gold"
"This is fantastic. Finally, some real value being added to the retail space. Keep it up"
"This is amazing, and although I understood this subconsciously, its the first time I've heard it academically."
"Martyn your whole series is excellent. thank u!"
"Thank you very much for sharing your knowledge, you're awesome man!"
"Thank you for your hard work and dedication of putting this series together for the Algo Trading community!"
Thank you everyone for your kind comments. It makes it all worthwhile, Martyn Tinsley
Martyn is a quantitative researcher and algorithmic strategist at the forefront of trading strategy validation. As the pioneer of the Walk Forward Correlation (WFC) technique, he has dedicated his work to solving the industry's most persistent challenge: distinguishing genuine alpha from over-optimized noise.
Driven by a rigorous analytical framework, Martyn specializes in deconstructing market complexities to develop proprietary diagnostic tools and evidence-based insights. His mission is to bridge the gap between research and actionable execution, providing the global trading community with the blueprints and resources necessary to build systems with mathematical rigour and a verifiable edge.
"Wow. Amazingly enlightening. I have read so much on proper ways to backtest and [these videos] blew my mind. Thank you so much."
"Martyn, you should seriously consider writing a book... Your content is top notch."
"What a unique level of great quality, useful content"
"This is great!... You are a talented educator Martyn!"
See more comments on the YouTube playlist or at the bottom of this page
About Us
Specialists in algorithmic trading for over a decade, we trade our own accounts using a blend of traditional algos and machine learning models.
We are on a mission to help other algo traders succeed by promoting best-practice trading system development and optimization.
Popular Resources
Useful Links
Contact
Built in the UK.
Proudly serving Europe and the World.
