>Better statistics

>I’ve broken apart the time frames that I’m studying, and so I now have a better idea of the sort and variety of results I can expect from the random-entry system. First I divided the data into 2 parts, and then I divided it into 6. For both studies, I performed the same simulation and gathered the same data on profitability and risk. The results are below.

Two 5-year periods

All numbers are counts.

Criteria 1st 5-year period 2nd 5-year period
Ended with positive number of pips 9841 9795
Simulations that made money by the end 9839 9920
Experienced absolute drawdown 9823 9839
More losing than winning trades 10000 10000
Positive R-expectancy 9880 9939

6 sub-periods of 10000 hours each (Except the last one)

All numbers are counts (Out of 10,000 simulations), except for the last row, which is the R-expectancy for all the 10,000 simulations combined. 10,000 hours here does not include the time taken for weekends (When the markets are off). So each 10,000 hours is approximately 1.5 years.

Criteria 1st period 2nd period 3rd period 4th period 5th period 6th period
Ended with positive number of pips 7876 9735 8165 9391 9129 7980
Simulations that made money by the end 7631 9595 8484 9730 8934 8256
Experienced absolute drawdown 9838 9148 9607 9136 9254 9146
More losing than winning trades 9999 10000 10000 10000 10000 10000
Positive R-expectancy 7829 9658 8683 9771 9048 8380
R-expectancy 0.0321 0.0801 0.0456 0.0892 0.0558 0.0439


Now for the good stuff. Lets see what these numbers tell us.

Bad news

Lets begin with the bad news, since its all uphill from there. The bad news is that the numbers are not consistent. During some times, the system does work less well (Than other times). I’m only going to address the 6 sub-periods study since it gives more fine-grained control.

The bad news is that for some 1.5 year periods, only just more than 75% (7631) of the simulations make money. The next best is just above 80% (8256).

Good news

The good news is that on average, even for the worst scenario, this system makes money! Even when only 75% of the simulations make money, the average return per trade (For all 10,000 simulations (Even the ones that lose money at the end of the run)), shown by the R-expectancy, is 3%! Lets put that in real-world terms.

If 10,000 people ran this system independently over this time frame, and used the same position-sizing rules that I did, and they agreed to share all their profits (Positive and negative) evenly, they would have made the equivalent of a 3% increase of the amount they risked per trade, for every trade that they make.

How would you like to be mathematically guaranteed a 3% return on the amount of money you risk, per trade?

Next steps

So now I have a benchmark of the results I would like from a trading system. This is a good starting point; an R-expectancy of 0.03. The next goal is to start practising making neural nets that can figure out better entry points. I am comfortable with the volatility trailing stop as an exit rule (For now).

Thanks for reading,

Ravi Desai

Leave a Reply