• Systems don’t need to be changed. The trick is for a trader to develop a system with which he is compatible.
• My original system was very simple with hard-and-fast rules that didn’t allow for any deviations. I found it
  difficult to stay with the system while disregarding my own feelings. I kept jumping on and off—often at just the wrong time. I thought I knew better than the system.
• Also, it seemed a waste of my intellect and MIT education to just sit there and not try to figure out the markets.
• Eventually, as I became more confident of trading with the trend, and more able to ignore the news, I became more comfortable with the approach. Also, as I continued to incorporate more “expert trader rules,” my system became more compatible with my trading style.
• As I keep trading and learning, my system (that is the mechanical computer version of what I do) keeps evolving.
• Over time, I have become more mechanical, since (1) I have become more trusting of trend trading, and (2)
  my mechanical programs have factored in more and more “tricks of the trade.” I still go through periods of thinking I can outperform my own system, but such excursions are often self-correcting through the process of losing money.
• I don’t think traders can follow rules for very long unless they reflect their own trading style. Eventually, a breaking point is reached and the trader has to quit or change, or find a new set of rules he can follow. This seems to be part of the process of evolution and growth of a trader.
• A trading system is an agreement you make between yourself and the markets.

Statistics

• Win rate
• Average winner / Average loser
• Profit factor (gross winnings / gross losses)
• Number of consecutive losers (needs to match with your psychological ability to handle)
• Expectancy (i.e. P/L of an average trade)
• Maximum drawdown
• Annual return
• t-statistic > 2

Robustness

• Markets: The system should be tested across all market environments, e.g. bull, bear, choppy, etc. Performance should preferably be consistent across all environments.
• Outliers: Is a significant chunk of the performance attributable to a just few trades or a particular market? or is the P/L across trades fairly consistent?
• Costs: Slippage and commission deducted?
• Logic: Are all signals immediately executable? e.g. limit moves might not allow execution

System Details

• How many rules and variables are there in the system? Beware of over-optimization.