Wow, this is a simple yet full of details book on system development. Before I continue, allow me to rate this book at 10/10. I think all traders need to have this book. Despite its simplicity, all necessary ingredients are presented. As such, it is a must read for beginners and it will be a good reference and refreshment for others too.
This is not a book showing exciting buy and sells signals. The whole concepts discussed are not direct. However, it provides proper guidelines and details explanation in a systematic manner. Having said that, please do not expect direct formulas and systems to make you rich in a quick manner. After all, system development is more than just buying and selling rules. This book at least helps traders to build up basic concept for further assessment into better stuff in long run. In short, grab a copy of this book and it certainly deserved a place in our book shelf.
Last but not least, listed below are some of the quotes that really "refreshed" me after more than 15 years in this industry. Great job by the authors!!!
If you make a bad trade, you have money management, you have a whole bunch of things that will come to your aid, and you're really not in so much trouble if you make a bad trade. But, if you miss a good trade there's really nowhere to turn. If you miss good trades with any regularity you're finished, you're doomed in this game. ~~~ Bill Eckhardt
The wider the sample size and the lower the number of variables, the better the estimation ~~~ Calculation of the degrees of freedom= whole date sample - rules and regulations - data consumed by rules and conditions.
The dangerous point for Monte Carlos analysis ~~~ you take a trades from the trading logic as you get them from your back-tests. But, what if this trading logic is only curved fitted and over optimized? The Monte Carlos cannot see this...
My concern is that people are using Monte Carlos as a certainty test. It isn't. It's a probability test.
Given enough complexity, it is always possible to fashion a rule that buys at every market low point and and sells at every market high point. This is a bad idea. Perfect timing on past data can only be the result of a rule that is contaminated with noise. In other words, perfect signals or anything approaching them almost certainly means the rule is, to a disturbing degree, a description of past random behavior (i.e. over-fitted).
If your trading system is simple enough but has a sound logic with valid rules, it is able to capture some parts of the recurrent predictable patterns but does not adjust itself to the market noise.
It is often better to use just one or two indicators, or only the price itself... Adding more and more rules will just increase the adaption of your trading system to past market data, but it will not increase its predictive power for real trading.
Be careful with short data samples regardless of timeframe or trading frequency. If your optimisation windows is short, it is more likely you will miss important data outside of the window.
If you think that systematic trading is a job for life, please do not overrate the importance of programming skill.
The more complex the system the more easily it will be over-optimised with too many variables or inputs. If you are not a sophisticated programmer this could be good luck because you will never run the risk of writing codes that are too lengthy.
No comments:
Post a Comment