Group Co.
Things You Learn After 1 Year of Day Trading for a Living
Traders Magazine Online News, December 17, 2018
I’ve made 20% ROI in 6 months and lost it all in a single month. Then I lost 30% in 10 trades the following month. I’ve earned and lost thousands of dollars, almost blew up my entire account twice, got dozens of margin calls, tried multiple Machine Learning techniques, traded multiple markets, time-frames and instruments. I’ve made any possible mistake, but somehow I survived and learned a lot.
After 4 years in the Software Engineering industry, I realized my path was too predictable. I would always deal with Data Science related projects. Working in a small company, enterprise and a startup shaped my industry perspective but nothing was quite satisfying. My good old passion for Algorithmic Trading would never leave me alone. I wanted something else, so I decided to quit my Data Science career and pursue day trading for a living. Here is summary of mistakes I’ve made and how you can avoid them.
Machine Learning Hype
We have seen Machine Learning applications everywhere. The media loves it, people don’t understand it and investors call it “buzz words” because they will never get it. DeepMind’s AlphaGo progress is mind-blowing, and it’s only the beginning. I was always telling my colleagues that someday we will sit and think about Java classes as they appear in Eclipse in front of our eyes, and during our sleep time we will be injected with some books “read” like in The Matrix.
The problem with Machine Learning is that it’s very tough to apply in trading. It’s more of a filtering method rather than a decision making tool. Most of the paper trading tests will be awesome and will fail in real trading because they over-fit. You will fight it with cross validation and cherry pick the best models that performed best on out of sample, thinking you are safe, in a way adding bias and leaking data. This is not the way to do that.
Avoid over-fitting by carefully averaging and evaluating on different assets, time frames or periods. Use non-conventional train/test splits and add random noise to evaluate your generalization power. Always be extremely careful, because you don’t know what you don’t know. You don’t know what’s going to happen, but Monte-Carlo simulation is your best friend, as you can at least simulate a lot of scenarios.
Risk
Multiple times during my trading I was feeling safe and thought I have nailed it. I felt like there is nothing that can surprise me, and time after time I was slapped in my face by mister market. Single trade going wrong and wiping out your previous 10 profitable trades, volatility spikes and your stops are breached like paper by knifes and free liquidity turns into a killing strangle on your portfolio, following a nasty margin call and your broker fixing all of your positions before it’s a total loss. Believe me, unless you saw that, you think it’s just war stories and imagination. Folks, this is reality, there is no free money out there. Everything involves risk, it’s a matter of how good you are at understanding your odds and your probabilities of loss.
For more information on related topics, visit the following channels:
Comments (0)