June 6, 2011

I started my Alpha Power implementation phase around mid-March. It took a long time to get there. It seemed that I was always sidetracked by something or other. I first wanted to prove to myself mathematically that the concept worked. After all, it worked in my randomly generated stock price series. I would at times hit a mathematical wall so to speak; not being able to express in mathematical form what I had in mind. For those that have read my papers, you simply don’t get up in the morning saying: what you need is a function matrix of stochastic differential equations.

You first need to get used to those things. Nonetheless, the math is there to corroborate that the Alpha Power trading philosophy is based on a serious and solid foundation.

In plain text, the method advocates only a few simple concepts: buy and hold for the long term, but do it progressively on the way up. If you have a handsome short-term profit, take it and reinvest the proceeds in your next trade to accumulate more shares. This way you will trade market cycles to your advantage over your long-term holding objective; you have for objective to hold for the long term anyway. Use your paper profits (the equity buildup) to acquire even more shares on the way up (which you can also sell for a profit and reacquire more shares). It took over three years to say those things in math and demonstrate that these procedures would increase overall return way over a simple Buy & Hold strategy.

The results of the first draft did just that; they were presented April 21st with a 47% compounded return over the 5.83 years test period (1500 bars). The next day, just to make the point, I put out another group of 43 stocks with an annualized return of 48%. Both tests were based on the Gyro Trend Checker script found on the old WL 4 site. Naturally, I modified extensively the script not only to include my own trading methods but to change the trend definition to better suit my purpose. You intend to accumulate shares on the way up; you need something that says the trend is up. It does not need to be accurate, only that somehow a stand is made; an uptrend declaration is given. Added note: the old Wealth-Lab 4 website has been taken down 14/04/2018).

The results of both tests way outperformed the Buy & Hold to such an extent that I don’t think any of the over 1800 scripts on the old WL site could even come close. The modified script was intended to have full market exposure as in time it accumulated more shares than the Buy & Hold. And then, it traded market swings over its accumulative process.

Alpha points are very expensive and very hard to get. Most of the current literature on portfolio management can demonstrate mathematically that alpha points if there are any, will tend to zero over the long term. And yet, there they were. Alpha points gained using trading skills and a stock holding function.

It took only a few days of added modifications to push performance higher. At these levels, alpha points are even harder to get. But nonetheless, on the same two data sets respectively, performance rose to 55% and 54%.

From this level of performance, I tried to modify other scripts, even with loose trend definitions. One that showed promise was the Neo Master version 2 script. After many modifications, I did release one performance chart (IMAX) which operated at over 100% compounded return. The script was put aside, as a not ready to show. But it raised the bar anyway.

My next step was to show that the pre-set functions could be regulated in an attempt to extract performance. In early May, a test on RIMM using trading levels was presented. You wanted more performance, you reach (pre-program) for a higher level. RIMM was not the best of candidates, over its almost 6-year test, the price went from about $140 down to $45 at the time (even lower today). But still, performance levels could be pre-set as shown in the Jensen Modified Sharpe paper.

In the last few days of May, I converted QQQ and QID Trader script to my trading philosophy. Just as with the New Master version script, performance levels were way high. The first display was in at a 91% annual rate of return for the first presented data set. Same data set using a different trend definition system and with full utilization of the excess equity buildup.

Then on June 1st, I presented this great idea after seeing on the old WL 4 site someone displaying a chart with the Livermore Master Key script. Issue a challenge to all to improve the script to a tradable level, all starting from the same point. So, using the first data set in the series, I presented the performance results using the script as is. The performance was dismal, barely made money, a mere 0.21% over the 5.83 years of the test.

At first view, the method had no value; its trend definition was unusable or was a totally worthless script. This would be fun for a few weeks at least, it’s a complex script, has a legendary trader’s trading method as a backdrop so everyone should be interested. Sure…

Well, that was a short challenge. Only a few hours later, my own modifications to the original script had pushed performance to 86% per year. At which point, it became useless to continue the challenge as already the bar was much too high for anyone on the board.

I was in modification mode, so I continued to improve not only on the model itself, but I also jumped to level 1 where I knew performance would increase further. Thereafter, I presented the 3 datasets in succession with 109%, 117%, and 103% respectively. Even after a 5-week market decline, all 3 datasets exceeded 100% compounded return.

An astonishing performance even if I have to say so myself.