I can think of three ideas that made a huge difference in my performance (so far).
1. Learning to become a pig
I've been trading the SP for a long time. It's no secret that the opportunities for profit vary from day to day. Years ago I looked at all the fat tail trades for a couple of years. Once I identified when they happened, I went and looked at the intraday charts. What I found was just about all the big day moves were on trend days. From then on I've use a pyramid on suspected trend days to have my largest position on when the market makes it's largest move.
2. Learning what a edge was
After reading the Toby Crabel book I played around with his ideas. Using his methods I found lots of profitable setups. I wrote programs and tested them and found most did poorly. I couldn't understand why something worked well one period and did badly on another. I had read other trading books that talked about robustness but I found no matter how much data I backtested with, the forward tests were disappointing. I decided to figure out why the methods failed. I talked with some stats gurus and they told me about their tools. One guy questioned whether any of the ideas were better than random. I decided to write a program to pull out random trades and then rank them against the ideas from Toby's stuff. It was like a light went on when I saw the profits were similar to the random trades. I found some trades that performed better than random over time. When I did in sample and out of sample tests I found the profitability tended to persist. Not all have lasted over the years but with the test versus random I've been able to see when a idea is dying well in advance of losing money.
3. Learning dynamic position sizing
I've always been uncomfortable with using something like risk 1% on a trade with the idea that each trade is independent and has the same potential value. I've experimented with different money management ideas over the years. When I think of risk I think of maximum drawdown. If my account is at a equity high risking 1% seems trivial. However if my equity is down 20% from peak then the 1% risk seems like it may be too high. I wanted a way to size my positions so that when I'm killing the market I can trade like a pig. Likewise when I'm being killed I can duck a massive drawdown. I did lots of tests and earlier this year I created a formula that changes the size on each trade based on my equity curve and the level of drawdown I'm comfortable with. What I found was the total profits went way up, the profit factor dropped, and the sharpe ratio went well above 2. I've tested it on many models and I'm now using it with excellent results.
Here's a example of how the equity curve looks when it's used with a crappy 1.4 profit factor model with a maximum drawdown preference of 20% or less. The actual max drawdown in the test was 11.9%. The test started with 125,000 in 1998 and grew to 1,776,440 at the end of 10/2003.
