July 20, 2014

As a summary, up to now, 4 long-term trading strategies have been analyzed. All four started as nonproductive, meaning that they could not even beat the Buy & Hold over the long term (read 20+ years). The trading strategies' original versions have been in public view on the legacy Wealth-Lab site for 8 to 12 years. Each strategy was modified to gain a long-term perspective with, for a backdrop, the accumulation of shares.

To increase performance, one would trade over the accumulation process producing cash that could be recycled into buying more shares. Thereby producing a positive feedback loop. Their governing equation was like a dollar-cost averaging formula but with a twist. You were forced to sell what you were averaging in, in order to create the feedback loop. While everyone is looking at stop loss for protection in order to reduce risk, I was looking at holding and stop profit functions. Profits that would end up in the re-investment bin. The controlling equation:

A(t) = A(0) + Σ(H(1 + r + g + T)t.*ΔP)          with   r > 0,   g > 0, and  T > 0

You want to increase the reinvestment policy g by increasing T the trading activity's contribution to the mix. There is a relation to the number of trades you do. The feedback loop should be considered the heart of the system. It comes from the increased profitable trading activity. The more you can or could trade profitably, the more you could feed back to the system.

But you wanted minimal risk, so you opted to spread out your bets over time (time averaging) and over many stocks (diversification) using small bets (position sizing). This way you could not only increase your profitability, you could also reduce your overall long-term portfolio volatility.

Each trading strategy analyzed used very different trading methods to achieve their respective primary goals. Their payoff matrices could be easily compared using the following:

Σ(H(TS n=(1,2,3,4).*ΔP) >> Σ(H(B&H).*ΔP)

The payoff matrix of all 4 trading strategies (TS) generated much more profits (>>) than the Buy & Hold. They accumulated more shares and more cash than the Buy & Hold. Note that you had to be in the game for those 25 years to collect the rewards. Extrapolating (just to get a ballpark figure) using their respective CAGR, for where each strategy could or might be in, say, 5, 10, or 15 more years down the road, you get (1 + CAGR ( n(1,2,3,4) ) )^(25 or 30 or 35 or 40). With an average doubling time of less than 4 years, you would get... depending on the number of years, so and so. Note that extrapolating does not mean you know the future.

Their objectives were the same. Progressively, slowly, $5k bets at a time, these programs would accumulate shares over the long term and, to accelerate the process, would trade over it. Thereby providing a cash feedback loop to help the trading account grow faster. To put the $5k bets in perspective: each bet at the beginning represents 0.167% of the portfolio. And as the portfolio grows, this percentage will gradually decrease down to below 0.01% and then still further. Each $5k bet represents less and less as a percentage of the whole portfolio.

Now, if such a trading methodology had any value at all, it would be easy to show in a backtest to see what would be the outcome. This is exactly what has been done with all 4 of these trading strategies.

The results were positive, and by a wide margin which kind of validates the approach, a proof of concept if you will. At least, at the very minimum, all 4 trading strategies could have handled their past. Not going back only a few months, but going back 25 years... Not on a single stock but on 30 stocks... Not on pre-selected stocks, but on the 30 DOW stocks. Not just doing a few trades here and there but doing thousands, even hundreds of thousands of trades over the 25-year trading period.

Yet, all I see is a variation on the Buy & Hold theme with a weak hold.

Summary

Just to resume the current state of this project, here is a recall of the tested trading strategies.

The first one was based on the Livermore Market Key, published in the 1940s and coded in 2005, which I modified in 2011. It is currently in third place.

The second one was based on the Bull and Bear Power Balance System, published in 2003. My modified version is also dated 2011. It is currently in first place. It has an impressive trade clustering algorithm.

Next came the XDev trading script. This one was published in 2002. It has the distinction of having gained the number 2 spot. I like this one, it is more my style.

Coming in 4th position was the Support Resistance Trader, published by Fundtimer in 2006. This one was modified almost live as I was writing what you will find in that link. It was relatively easy to make it outperform. My modifications to this trading strategy greatly changed the original author's intent. But this can be said of all 4 tested trading strategies.

The above trading strategies have for origin code that was simply given away that you could modify to do whatever you wanted them to do. The original program code is still available for free on the Wealth-Lab legacy site. I thank each of the authors for having the generosity of sharing their work so graciously. Just as I thank the Wealth-Lab people for having kept their legacy site operational for the last 6 years (no longer available).

None of my trading strategies had been tested over the past 25 years of market data before. Their previous testing interval was limited to 6 years, and on data sets other than the DOW, (only 2 stocks out of the 30 have seen these strategies). So these strategies have to consider 25 years of unseen data, except for 2 stocks having seen 6 years of data. These 2 stocks will have 19 years of out-of-sample data, 3 at the back end, and 19 at the front end of their respective price series to contend with. Therefore, their influence in curve fitting or over-optimizing would be minimal, to say the least.

I think that the most important element in designing profitable long-term trading strategies is the person behind the keyboard. The tools don't matter that much as long as you, as a strategy designer, can do what you want them to do. There is no need to re-design the wheel if you can buy one at the local hardware store, just as there is no need to start from scratch when ready-made tools are available.

I would even dare to add that if I can do it, then anybody can.


Created... July 20, 2014,    © Guy R. Fleury. All rights reserved.