It's been a long time since I posted - and for good reason. One of the things I have been thinking about deeply in the last couple of weeks was to implement my strategy as an EA. However, through the process I had to answer the following question: Is my trading strategy a byproduct of randomness?
This question merits questioning because, if the answer is yes, then I am putting confidence in something that will burn me one day.
There were many ways to answer this question.
One way is to back-test the strategy long enough to made a valid claim. The trouble with this is - what will constitute to 'long enough'? We hear of EA systems that perform great over one year, only to flop another. There is a chance that the EA is has bias because of hindsight. That is, the trading system performed well because it has been tweaked to suit backtesting data. There is just no way to eliminate this bias in real life.
I now propose a better way - develop an EA that fits this criteria:
* trades randomly - no buy/sell signals, just random()
* trades a ton load of trades
* uses martingale
The last criteria is a unique one. Martingale is known in the betting community for one very good reason - it can bankrupt a person eventually if the expected value is zero. The only way to win at a martingale strategy is to have an infinite amount of cash (for which one would reason why on earth he's investing) and have the expected value greater than zero. In the case of a purely random (or worse) trading system - it has to go bankrupt. In other words, if the system is random (or worse), then a martingale should go bankrupt eventually.
Now, here's where it gets interesting: I deliberately setup my system as follows:
* (takeprofit * average trade size) - (stop loss * average trade size) + (takeprofit * average trade size * stoploss/takeprofit) = 0 : this takes care of the expectancy theoretically being zero, assuming that price is random.
* beginning capital: sufficiently large enough so that the trading system isn't suffering from cash-flow issues, but rather it is the trend that kills it
* trade every tick: if the stock is truly random, it shouldn't matter if I trade today, tommorow, or in the next year. Since the expectancy is zero, it will eventually bankrupt itself.
What are the results? As expected, most of my back-tested martingale trading systems go bankrupt quite early, which is expected. However, not all. I'll explain this phenomena in the next couple of posts. For now, I present you with Exhibit A and Exhibit B.
Exhibit A: A random Martingale that actually performed? Exhibit B: A crashed random Martingale system that's in profit 99.99+% of the time???
