Trading System Development 101 (Appendix A)

This year, I’ve been trying to get more organized by turning rough drafts into finished blog posts. Sometimes, I don’t even understand what I have written [long ago] in the drafts, but I am presenting them anyway on the off chance that someone out there can benefit. In that vein, I have a number of loose ends and notes regarding my mini-series Trading System Development 101 (concluded here) and related posts that will occupy two further entries.

—————————

When choosing fitness functions, we need to understand how they can possibly deceive. For example, a profit factor of 2.0 may be $5,000 (if it makes $10K and loses $5K) or $50,000 (if it makes $100K and loses $50K). Also, average trade is not per day; $1,000 for a trade held for five days is more attractive than 50 or 500 days.

With regard to my brief experience thus far testing algorithmic strategies, I’m shocked to discover almost nothing works! This is despite all those books with chapters on indicators, all the instructional webinars, and all the educational programs alleging to teach technical analysis. Hardly anything that claims to work is backed by supporting data, either.

With the exception of equities, I have gotten the impression that money is much easier lost than gained. Making money in non-equity markets seems to require a behemoth effort.* With equities, almost everything makes money when bought. Problematic are the occasional sudden, fast, hard corrections and bear markets that wipe out much of the gains in a short period of time. This is no big deal for long-term investors who don’t often look at the market and hold positions for years. For traders who try to profit consistently over the shorter term, this can pose major psychological challenges.

To reiterate a point made near the end of this post, finding a viable trading strategy is probably not about reading an article or chapter on a TA indicator and using it as prescribed. The answer is not to attend an online webinar and implement said strategy verbatim in my live account. Most things I will test will not work; it’s not nearly as easy as the presenters make it sound. The most important thing is a well-thought-out development process and boatloads of patience and motivation.

Going back to this blog mini-series, here’s a note on over-optimization (i.e. overfitting):

     > Though not specific to automated trading systems, traders
     > who employ backtesting techniques can create systems that
     > look great on paper and perform terribly in a live market.
     > Over-optimization refers to excessive curve fitting that
     > produces a trading plan unreliable in live trading. It is
     > possible, for example, to tweak a strategy to achieve
     > exceptional results on the historical data on which it was
     > tested. Traders sometimes incorrectly assume a trading
     > plan should have close to 100% profitable trades or
     > should never experience a drawdown to be a viable plan.
     > As such, parameters can be adjusted to create a “near
     > perfect” plan — that completely fails as soon as it is
     > applied to a live market.

I will conclude next time.

* — Most of my testing thus far has been of symmetric strategies: opposite rules for buy and sell short.