As I told previously, I’ve already polished the most important parts of trading (risk management, taking small losses, waiting for bigger profits, adding to winning positions, not to losing ones, patience and execution), now it’s time for the least important part (looking for entry and exit points, aka “trading system”).

I started with analyzing tick data, hoping to find a number of good trading ideas. I developed an algorythm to find similar price patterns in the past using Dynamic Time Warping, calculated their outcomes and built a dynamic tool to visualize real-time results.

I was inspired with its initial performance because it clearly gave me the right direction for the next few minutes. Not always right, but statistically significant. I was even able to trade it in my simulator at a breakeven level, at least covering spreads and comissions.

I started with just few days worth of analysis, then I decided to dig deeper and did some speed optimizations. At the end, I was able to analyze 3 years worth of tick data in 60 seconds, tick-by-tick. But, the more patterns I found, the closer their outcomes were to random distribution. At first I thought that my algorythm was wrong: while it gave me correct statistical parameters, patterns I found looked not the same or even close to each other.

I spent several hours doing various tests until I found the correct answer: the problem wasn’t in the algorythm itself, it was in the data. Tick data is just noise, there is no clear signal in it. My first good results were due to the fact that my algorythm always told me to act according to a current short-term trend. But, when I got to a higher level of analysis, I returned back to random results, as statistically expected.

Taking into account spreads and comissions, there is no way to trade profitably on tick data. Signal/noise ratio is too small to look for the statistical advantage.

Then I moved to a higher level, trying to do the same trick with 1-minute data, then with 5-minute data, etc. This time there were no suprise for me to get the same results. Yes, signal/noise ratio slightly increases as we go higher, but not in a way that gives a chance to build a robust statistical model. It seems to me that Forex daytraders are trying to trade pure noise. It’s the main reason why their win/lose ratio is close to 50%.

Then I went even higher, trying to find a timeframe where I can develop some kind of statistical advantage which will be able to, at least, cover risks. I finally found the one, but it’s not for daytraders. D1 is the lowest timeframe where you can build something statistically significant. If you don’t believe my, look first at H1 chart, then at D1. While H1 seems much better than M1, it’s still highly noisy.

I believe that there are some people who can trade profitably on H1, or even on M15, but it’s not the way I want to go. While trading less frequently sounds like earning less, it’s exactly the opposite. You can earn much more on a clear rare signal with high RR ratio, than on some frequent noise where you will always stay close to breakeven. If you are an expert in lower timeframes, just look at higher timeframes and imagine what may happen if you stay in some of your winning position not for few hours, but for several days, weeks or even months…

So, I clearly understood that daytrading isn’t for me. I want to trade on daily charts, entering the market 1-2 times a month for 1-3 weeks, at least. It perfectly fits good habits I acquired during the very first stage of my self-education: taking small losses and waiting for bigger profits. It also gives me plenty of free time to live my life and do some other work I enjoy, like researching the market, organizing and automating my work.

It doesn’t mean that I will stop looking at lower timeframes. I still need to polish entries and exits. I want to avoid huge stops associated with trading on higher timeframes. I feel that there is still plenty of room for experiments and automation.

As for trading results for this weeks, they are mixed:

I decided to start with a demo account before going live, so I entered perfectly into the market this Wednesday using my “Anti-Joe” approach. I spent just about $10 to build 0.1 lot position which gives me $10 for every 100 pips. At the moment I’m writing the report, I’m demo-positive about $50, $40 net. The target is about $120, or even more. Even if the price will finally go back, I will still be breakeven.

But, according to FOMO, I decided to enter into the same direction this Friday on my real account. It wasn’t a perfect entry in any way or form, just a hope to get the very first positive week as fast as possible. Yet another reminder to look for perfect entry setups, not for “good enough” ones 🙂

So, I’m $40 demo-positive and $30 real-negative at the same point 🙂

I will no longer try to enter the market in imperfect points, I will just sit and wait for the perfect setup. On the other side, taking into account much higher targets, I increased my risk per position to $30, or about 0.3% of my total risk capital. I also feel that I should find a way to add to winning positions smarter, not mechanically. It’s what I will try to polish on the next entry point. I expect it to happen in 1-3 weeks.

What’s next?

I will continue to develop my simulator, waiting for the next perfect setup on the daily chart. This time I will focus my tool on polishing long-term trading, not scalping. I hope to find a better way to add to winning positions before I enter the real market again.

I also want to look closer at commodities and stocks. If I decided to trade on daily charts, I can find much more opportinities there, than within these 2.5 highly corellated Forex currency pairs.

By the way, I’m not disappointed about switching from daytrading to longer-term trading. I feel that I can earn even more than I initially expected. The reason is simple: with much higher RR ratio and much less less volatility on my account balance I will be able to gradually increase risks and profits to a point, unreachable in daytrading.