Throughout history there have been many cases of unfair/false allegations, no regard for evidence, and coincident with a background of mass hysteria. Three examples immediately come to mind, the first being the witchcraft trials in the 17th century, and then McCarthyism in the 1950s.
The third that comes to mind is the practice of trading system optimization. OK, so it doesn't quite rank up there with burning witches and flushing out communists, but hopefully you get the analogy. Some would like to instill the fear of God in us with the belief that optimization is an evil practice.
My post today is to assure trading system designers worldwide that this is not the case. There are legitimate reasons for optimizing your design, so don't let fear or someone else's agenda interfere with what you SHOULD be doing, unless of course you are a meta-human with clairvoyant powers or have access to NZT, in which case you really don't care.
For most of us mere mortal traders, we are confronted with a barage of financial information on a daily basis... fundamental factors, technical indicators, trading signals, ... This isn't so bad when you are a trading system junkie like myself. But a basic dilemma arises when one has multiple pieces of conflicting information which is more often than not the case. Don't have this problem? I would be very surprised if this were the case, but clearly you don't need to read this post as you are one step up on the rest of the trading world.
We have a choice: we can (1) do nothing and let the cards fall as they may, perhaps flip a coin, or just turn off the computer and have a snooze; or we can (2) take our best stab at coming up with a single signal from all that conflicting information. Given that we have all the resources to perform the latter at Portfolio123, the choice shouldn't be all that difficult to make. Portfolio123 has historical fundamental and price data coming out of the ying yang, and development tools that trading system wannabees have wet dreams over. So it should be possible to come up with a single signal that puts some order to all that confusing information.
The real problem isn't optimization unto itself, but that it is an art and not a science. You can fail in this task as easy as you can fail at any other design system task, such as estimating parameters for a Dividend Discount Model (DDM). Both may lead to losses. The more experience and wisdom you apply the better your results should be.
With all the power that Portfolio123 affords, just make sure you don't get silly and start using non-relevant factors, and make sure you target time periods that you believe are relevant to the current economic environment. Also, re-optimize on a regular basis. Markets change and so should your relevant inputs, weights, etc. A good rule of thumb is to use a backtest period of 4x the forward usage. So if you use four years of backtest data, then consider re-optimizing your system within one year. Or when there is a noticeable change to market behaviour.
Another thing to be wary of is that optimization is a double-edged sword. Provide too much informational input then you will simply end up memorizing historical data. This is a bad thing. Minimization of input data is a god thing and should be a design goal. It may result in discovery of data relationships that you may not have been aware of.
So there you go. If you are a closet optimizer, you can come out now. There is no shame in doing so. When someone tells you that optimization is bad, give that person an ear full.