Monthly Archives: March 2015

Small post, filter tool box

Yes I am still alive! Hah. I am planning on getting back to being able to post some stuff to share in maybe another month or so. The next project will take me a while to get to, and I haven’t been able to properly have time to work through it.

I worked on a small “mini-project” today and the idea popped into my head. I won’t really elaborate on the thought process behind it too much, but I do want to share it because it may be helpful to others. I also won’t explain what exactly the data is detailing, because it’s not necessary (nor ground breaking).

When testing a signal or set up, we have some options when trying to improve the result. For someone like me who isn’t quite there but it close, it really only does take 1 small thing to completely blow something up; to make it either the foundation for an edge or something not worth touching. The initial batch of re-works and filters, and their subsequent +/- %’s to the success rate are important. The first method is more or less working forwards. Taking the result, glancing over it, and say, “Let me try to add this, or subtract this, etc”. An example of this is to try to filter out certain time periods. Perhaps the original idea planned to trade in healthy and liquid market states, and potential triggers that occur in the dead zones could be hurting the trading win percentage. This is a bit like brute forcing it, which is something that I think should be avoided.

It seems much better to try to work backwards – to go back through all the failures and see which hours of the day the signal works better or worse. BUT, and the point of this post, is that we must not forget other small additions we can make to really make the difference between the two stand out.

accum

The left side shows the failures by hour of the day, and the right show the winners. Alone, or separated, there doesn’t seem to be much difference. As researchers, we need to find good ways to compare both aspects in a way that is as least bias as possible. In this case, I chose to use a ratio of failure/winner. The standard assumptions show to be true. Trading very early in the session (asian open) seem to cause a lot of losses, while trading during the LO and throughout the active periods during the day seem to be a good strategy.

Possible next steps

I haven’t been doing much work recently because I’ve been trading! It’s a bit shocking for me, as it’s something I haven’t been this active about it in over a year. I’ve done my best to recall all the little things I’ve done since I started this blog and put them to use as I enter the market. My original plan when I started was to have a core probability for price to reach specific points at any point in time. Surprisingly, what I have now isn’t that far off! Although, given the nature of modeling, we can never be too sure that our probabilities are accurately forecasting the next move. I’ve put together what use I have been able to gather from SB (not much unfortunately, still working on that..), transient bars, transient waves, common daily and weekly play outs, weekly gaps, etc. I’ve been pretty satisfied with out the trades turned out, although I’m only on my 3rd week so I will have to wait and see.

Moving on, there are a few major hurdles I’d still like to solve, and I’ve begun pondering something that I never thought I would: Elliott waves. I absolutely loathed the damn hindsight painting, but I thinking understanding where it is wrong and right can provide clues for improving my own wave work. I had the pleasure of having some correspondence with an elliottician, and I must say it is interesting to see the way people have made it work. It’s currently in theory phase so I may not get the project off the ground (like many before it) but I think it would be quite a unique idea that I don’t think many have thought of.