Benj's Algo Trading Journal

[benjcal]

Ok, this is my first post with the goal of keeping myself accountable!

For the last couple of months I've been just building the tools to do "market research", like, I have observed some interesting behavior in ES premarket, where trends seem to tend to continue since there's not a lot happening, like, if it is slowing going up, it continues to do so until either 8:30 when news are usually released or 9:30 when market opens... that's the hypotheses at least.

The challenge that I have is that I'm a programmer and the ability to build stuff has taken me in a multi-month journey of building the perfect tool to "query" the market and explore algo ideas.

At the moment though I'm very close to finishing what is a very cozy algorithm platform that works just the way I want it to, where I can research ideas quickly, and then I can implement them in NT8 and forward test them with real money (with MES of course!)

So, that's where I am. I'm hopping to post, at least once a week progress report but by goal is that before the end of this month I should have finished the framework and have at least one algo running.

An incredible idea I got from Adam Grimes in his Hudson Sessions is to print charts in paper and just analyze them by hand. I've printed about 200 charts of the time frame and specific hours I'm interested in and it's been incredibly insightful. That's a practice that I'm hoping to continue, although I might change to using a Remarkable tablet or something, because that's A LOT of paper!

Anyways, best of luck to everybody here and stay green!

 

[mikkom]

Hello Ben,

If you are a programmer and not interested to build your own backtester (which I have done many times), it's worth looking at quantconnect, you can get decent data (I think they even have tick data) and runtime environment for a small price to easily test your hypotheses with good quality historical data. It's much cheaper than testing your hypotheses with real money :-D

Another way would be to get decent market data + implement your own fill methods etc but IMHO it's so deep rabbit hole that it's worth investing few bucks to buy quantconnect subscription to test your ideas.

 

[benjcal]

Hello Ben,

If you are a programmer and not interested to build your own backtester (which I have done many times), it's worth looking at quantconnect, you can get decent data (I think they even have tick data) and runtime environment for a small price to easily test your hypotheses with good quality historical data. It's much cheaper than testing your hypotheses with real money :-D

Another way would be to get decent market data + implement your own fill methods etc but IMHO it's so deep rabbit hole that it's worth investing few bucks to buy quantconnect subscription to test your ideas.

Thanks for the suggestion. I have used quantconnect and is not a bad platform but I wasn't the biggest fan of it. I guess now that you mention it, what I'm actually building is my own backtester. I'm using simple strategies to test that 1. the results of my backtest match the results of NT8 backtests and 2. that real life orders, if given the same data, are reproduced by my backtester.

 

[benjcal]

Today most of the framework is done. I'm not generating any statistics/performance data yet, I've been focused on seeing what my strategy is doing. This is primarily because my approach is very visual, as in, my ideas mostly come from things I see in the charts as opposed to some esoteric (at least for me) math.

So, the baseline is, I can make all the charts I want, I can add whatever indicators, I can plot buy/sell signals and I can generate individual charts of the times I am interested in.

That last one is probably one of my biggest peeves with other software, they would show me a continuous chart spanning for months, which was a pain when I'm interested in a strategy that trades only a few times a week!

In the attached chart, the signals are just test signals every 4th bar... I needed some data to plot

 

[benjcal]

I was reading The Art and Science of Technical Analysis by Adam Grimes, which is a fantastic book, btw!

Anyways, I've gone over that book two or three times now. This time though I read his concept of pivots, specifically how he explained first, second and third order pivots and how they establish market structure. He pretty much spell out the algorithm to compute those, which I had never seen before! So, I decided to explore that concept myself in my tool

The attached chart contains the high pivot points of the first and second order.

The reason why I think this concept very valuable is, as a friend put it, "how do you teach a computer to see levels", and I think this might be a way to accomplish that!