How Auto Pilot Day Trading optimizes the Strategies

Optimizing strategies is a science and an art. It is altogether too easy to over-optimize a strategy for a time period, and then have it fall on its rear as it goes forward taking real trades. There is a good book to read on this subject: "Design, Testing, and Optimization of Trading Systems" by Robert Prado. We use the back data TradeStation provides, converted to continuous contract data, to optimize our strategies. We generally optimize over a time period of a few months, beginning a month ago and going back, say, 5 months. Then we let it "walk forward" to today, to see how well the optimization holds up. If it continues to do well up to today, we may then begin to trade that strategy on our own accounts.

We always take into account how a trade will "feel" to the trader. We try not to choose parameters that will make the trader too worried about a trade, such as an adverse risk/reward ratio on any individual trade, for example, a small profit target and an unusually large stop loss. This is part of the "art" of optimizing — no matter what past performance on past data shows in a hypothetical set of trades, if the trader doesn't like the trade, it is not a good one.

These are the sorts of parameters we optimize:

• Entry orders
  • Parameters for the Alchemy functions
  • Number of ticks from the entry signal to place the order
  • Should we use MACD and if so, with what parameters
  • Should we enter at market if the price is moving quickly
• Targets
• Stops
• Trailing stops
• Profit target for the day
• Loss maximum for the day
• Start and end times
• How many ticks against the trade do we let it go before changing the target to damage control
• Should we take a trade in the reverse direction if price goes against us
• And some others