I had met a gentleman who really liked the idea of trying to determine intra-day market direction and taking a trade in that direction. He did, however, say, “I don’t care what you say, there’s no way in hell something like this is going to work on a lot of stocks.”
My idea and his idea of ‘working’ were probably very different at the time. I was still trying to develop ways to demonstrate The Intra-Day Momentum Method in real-time. My idea was to develop a way to demonstrate the research in real-time and do it on a VERY LARGE number of stocks/ETFs.
By the time we had met, I had already developed the software and application to test, simulate, and automate trading mathematical approaches, but ran into challenges with Prop firms when trying to implement them, more fraud. Not to mention, I would have been seriously underfunded as a trader. My money and time was spent developing the software and trying to develop solutions for transforming the ‘Art of Trading’ into something more scientific. The goal was not to develop a trading system per se, but a trading methodology based on mathematical patterns.
This methodology was not coded into the software. The software could test, simulate, and trade two dimensional patterns, such as breakouts and pullbacks. At the time, I had also developed other applications to help me in determining the significance of the levels in the Intra-Day Momentum Methodology as well as quite a few other similar approaches that traders had historically used.
I had sent him an email asking him if he would like me to explain what is wrong with most people’s analysis techniques. He insisted, NO, he wasn’t interested in that. Later on in our venture, I decided that I wasn’t going to ask, I was just going to let him know the flaws of traditional analysis techniques. The conversations I had with people who I had no idea knew what I was working on were very interesting to say the least. I was told by a man and lady: “You are going to have a whole lot of people that are not going to like you.” I asked “Why’s that?” Their response: “Because they are going to be jealous.” I responded with disbelief that anyone would care that I was trying to create something to help other people.
I also sent a video demonstrating the software. He did not watch it at that time. I sent it to him later on and he was like: “Why haven’t we been using this all along?”
The software did not have this methodology coded into it. I turned to this methodology after my attempts trading through prop firms turned into ‘bad ideas.’
His idea was to create a hedge fund and trade a similar method, but it would be computerized and optimized by a computer. I have absolutely ZERO interest in trading other people’s money. It is something I will not do.
He proposed an approach to where a popular trading software package was able to determine the levels that were to be traded. This would be done based on an optimization routine that would figure out what levels would be best to trade based on historical data. I did not disagree with the idea. I was incredibly skeptical of the idea of optimizing just for the sake of optimization vs. learning something and optimizing based on something we had learned. Letting the computer do it to me was like throwing something up in the air and believing that it would work, without an understanding of why it worked, when it did. Although it would have worked over the historical time period that was analyzed to create the levels in the first place. To me, learning something and optimizing based on something you had learned was far more powerful.
In back-testing, I recall that when the time frame was changed, so did the results. Something that should not happen when testing a mathematical approach. This is because traditional trading applications are based on traditional technical analysis. When using traditional technical analysis, you have to wait until the time period ends before the value of the indicator is known. This means that from my experience, the software will not place a trade within a price bar, before the price bar has closed. So, the higher the value of the time frame you were using, the more inaccurate the results would likely be. Also, when you change the time period that you are applying an indicator to, you are also changing the results, this should be obvious. I assume that this is what happens when you try to apply a mathematical approach as well. It is waiting for the time period to close before it can place a trade. That process is not math.
We worked on this for quite some time. The code was written by an expert in the software’s coding language. After a while I downloaded the software and began to assist in figuring out how this was working. What I found was that there seemed to be something wrong with how it was working, and I could not figure out what it was.
I became more confused than I ever have been because I still cannot tell you what the software was doing. However, it was working very well for a while. I think it was adding a price measurement to the end of a time period-based data point and it was creating a level there.
The simulation results were very good for quite some time. And then it had a relatively large drawdown in a week. This happened, if I recall correctly, when the market trend changed.
Due to my experience with traditional software, I haven’t used anything that is based on a chart to do any mathematical analysis since that experience. I do look at charts for ideas on designing patterns and to verify prices and patterns, but not for mathematical analysis. I just prefer to look at the data for analysis and not a chart.
My experience with traditional trading software was that it did not appear as though applying a mathematical approach was possible with these products. I am not certain that I had not experienced these issues in the past. I am certain that my software’s design was much more suited for any sort of mathematical pattern-based approach to analyzing data for patterns.