So are u admitting , all your random entry systems boasting on forums is BS?
How many different systems did you test and over how many years?
How did it perform without M A E, but using all other improvements?
How different is it from using a dynamic stop loss i.e stop placed at previous low or high on bar?
O D T
I'll try to clarify a few points. I split my time between trading manual and automated strategies. The vast majority of my income comes from trading a manually executed strategy, its basically a micro swing strategy based on oscillator divergence at support & resistance.
In addition, Ive been working on automated strategies for a while, but only seriously since around 2005, I probably devote around 3-4 months of full time research each year to this, and even when I'm not involved full time, I still probably spend most weekends and as much as 5-6 hours a day working on this stuff because in the longer term, thats what I intend to do. I dont want to be watching a screen for 5 hours a day when I'm 70 !
I'm not about to discuss anything of value in a zoo such as T2W but I will say that one of the biggest challenges that designers of automated trading systems face, is distingusishing the results that they achieve from those obtained from random chance.
I therefore tend to use quite a lot of random sampling, and this applies to entries. Most of my systems work precisiely as I've described above, I'll start with a system based on random parameters, and then incrementally add systems with additional filters. I am 100% serious when I advise people to approach system design from the perspective of random trades, then compare the distibution in gains and losses of any "improvement" against that random benchmark.
I dont generally spend a great deal of time backtesting "systems" as such, I'm more interested in the component elements of a system. As a simple example, lets say I want to assess the suitability of an indicator to assess "trend", then testing a system to determine if the system made a profit is a bit pointless, as it doesnt really tell you if the indicator is correctly assessing trend, I trust you see the difference.
You ask how these things perform without MAE, or with alternative stops, and thats a reasonable question. In general, most of my automated strategies are based on the assumption that markets trend, and that there'll be kurtosis in the distribution of gains and losses. So as a general rule, I'll want to cut losses short, and let profits run. An MAE based
approach is therefore sensible (dont get the idea I use something as simple as MAE cos I dont !)
The high or low of a bar doesnt really mean much to me, a bar is an artifical construct drawing a box around a unit of arbitarily selected time, if it means anything at all to anyone anywhere, then the statistics will reveal that.
If you trade a bunch of random systems (and by random I mean systems that are either deliberately designed to be random, or are random simply because the ignorant monkeys who designed them know no better) the you'll find a distrubution in the equity curves generated by those systems. Some do amazingly well, some do amazingly badly, and on average, a group of random junk is going to break even minus costs.
Thats why forums where trading and systems are discuseed are a zoo, you have a bunch of people scattered across that distribution, some are doing well and think they have it cracked, some are doing poorely and cant understand why, but in reality, they are the clueless inhabitants of a zoo stumbling in confusion through a hall of mirrors. Random systems are the key to setting them free
The same distribution in equity curves also occur with "better" systems, and ironically, some designers far more clued up than I often argue that the "better" the system (by which they mean "more profitable"), the greater the vaience in the distribution of returns. So as you've worked out for yourself, diversification is an essential component of automated trading. So to answer your question, I've had years where totally random systems outperformed some very clever algorithms. My comments regarding random entry are anything but BS, they are an extremely important component in my overall strategy, I'd be lost without the information these provide.
It comes down to Talibs argument about the dentist living all possible lives that a dentist might live !
As a concept, MAE makes sense as a worst case stop, but its not an edge in itself, and of course, there are set ups that offer a much better reward : risk should you wish to take them.
I might actually post some detailed analysis data from one of the systems at one of the trading forums, and if your really lucky, I might just post some stuff about the development methodology I use for doing this stuff, if I can work out ways to stop reverse engineering.
