Other than the setup, the stop loss level is the next most important key to trading success. By ending doomed trades earlier, I can improve on my risk reward ratio. A tighter stop will also mean I can enter a larger position while keeping the dollar amount of risk constant.

My stop loss so far has been quite mechanical. No matter which setup, my initial stop loss will always be on the other end of the trigger bar. I may tighten my stops or shift to break even when price reaches key level. That is about it for the discretion part of my stop loss.

Similar to part 2 of the review, I will be focusing on the 25 counter-trend/whipsaw outside bar and pinbar trades. I will revisit the remaining with trend and double inside bar setup when the sample size is big enough.

The first step towards finding the optimal stop loss level is to find out how far did the trade go against you before you take profit or hit your stop.

I will illustrate the maximum loss with the following histograms.

Firstly, this is the histogram showing the maximum loss of the 25 trades. The categories are grouped into -1000 or less (maximum loss is $1000 or more), -800, -600, -400, -200, more (maximum loss is less than $200).


 Next this is the histogram for the 12 outside bar setups. All the 12 trades are profitable eventually.

Finally, this is the histogram for the 13 pinbar setups. The 4 trades in the '-1000', '-800' and '-400' are losing trades with the remaining being profitable trades.

From these set of graphs many meaningful insights can be drawn. The good news is that more than half of the trades have a maximum loss of less than $200. The best trades usually give a profit at the start and never look back.

The next thing I observed is the vacuum at the '-600' category. This shows that there are no trades with a maximum loss of -$800 to -$600. With this my hypothesis is once the loss reach the $600 mark, the chance of survival is close to zero. In fact, once it reaches the '-400' category the odds are bad enough though one of the trades that reached a loss of $500 did turned profitable eventually.

The right thing to do is to set the dollar amount for all the trades to be just below that of the maximum of all the winning trades, which is less than $600 among the sample. By doing so, the winning trades will still be winning trades, but the losing trades will be capped to a smaller loss. This is a naive method that overfits the model. And as I get more trades, there will eventually be trades with more than $600 maximum loss and still makes a profit in the end. The magical cut off just does not exist.

To complicate stuff, if you want to keep your absolute dollar risk constant, reducing your stop loss will mean that you get to trade a bigger position. In this case, it may be possible to set an even smaller stop loss. Even if some of the existing winning trades turned into losing trades as the stop is tighten, the additional gains from the other trades with a bigger position may be more than enough to offset the losses. This is a mathematical problem that can be solved, but not without a larger sample size. On a minimum, I need another full year of data to perform that analysis.

For a start, I am proposing a stop loss at two third of the current stop loss. It is slightly wider than the $600 mark. And more importantly, I will be trading at 1.5 times my current position.

In the findings from part 2 of the blog, the 25 trades above would net me a total of 11.7k blog. By reducing my stop loss and increasing my trade size, I would have made another 5.8k more.

Along with the with trend trades and inside bar trades, my total earnings from the 31 trades in the first half of 2017 would be just slightly above 20k. I certainly hope that I am not over fitting the model, but I would only know after some out of sample testing.

In part 4 of the review, I would be tying up some of the loose ends including the use of time stops and alternative exit strategies.