November 26, 2011

Latest Simulations

What follows are some of my posts on LinkedIn presented in reverse order, starting with the most recent and going back in time. Therefore it might be preferable to read this from the bottom up by date.

November 26

Here are some of my observations concerning the last 9 charts (AAPL, AGQ, AMZN, DDS, IMAX PNRA, SHS, SINA, ULTA) presented.

All 9 charts used the same script. There are currently 47 competing trading procedures in that script making it a complex structure to manage; each procedure having its own mission at trying to improve on performance if it can. Not knowing the future, a position is taken based on a decision surrogate, and then managed with only two possible outcomes: a profit or a loss.

I estimate that some 30% of trades were randomly generated within their opened trading windows, helping to add to trade clustering. But this has been shown to limit performance, therefore an easy way to raise performance even higher would be to reduce this limitation.

I do like the clustering of trades on those charts. Finally something that trades intermediary tops and bottoms. This is a common behavior for all nine charts. So the sum of the trading procedures implemented have somewhat the same aggregated response to price movements. You still don't know precisely which decision surrogate triggers a trade; but you do know that it has been programmed in the system to do so. The system still does not know what a top or bottom is, or what trade clustering means but that is not its mission. The trading procedures compete for a limited resource: the available cash in the account.

Based on the volume of trades, the program needs to be automated. Some 34,311 trades were initiated in those 9 stocks over the almost 6 years testing interval and 27,711 of those positions were closed; most with a profit considering the ending cash on each chart.

I also like the equity curve (system profit histogram) at the bottom of each chart, all are relatively smooth and vary a lot less, percentage wise, than their respective stock prices. The reason for this is simple: the ending cash in the stock's respective accounts amounted on average to over 50% of generated profits. A better use of this excess cash reserve would be to trade more and produce even higher results. It would be like exchanging some of the cash reserves for more stock profit potential thereby increasing the number of shares to be held for the long term. I already know which parameter modifications will make this happen.

The system profit histogram also says that there is less, not more, overall portfolio volatility than market prices would suggest. And therefore, this would mean that an exponential alpha might help reduce the extent of draw downs as a by-product of the trading methodology used.

Since my trading procedures are scalable, to increase the output by 10 would simply require to multiply the input by 10 by providing 10 times more capital in order to increase the bet size by 10. This would represent in the AAPL case to go from about 50 shares per trade to 500 shares traded at a time. And if that would be considered too much to trade in a single order, then it is relatively easy to time or quantity slice trades during the trading day. This scalability feature of my trading strategies has been demonstrated before, so there no need to redo this test again. Overall, it would push the initial investment of 900k to 9M, and the total profits from 200M+ to 2B+ with ending cash reserves in excess of 1B+.

To me, the real problem with this script is that it is too big for me. I can not supply what it needs. I designed a long term investment and trading strategy that I can not survive or properly fund. So, at present, its destiny is to rest in a safe until ...

November 24

I “claimed” in my first paper in 2007 that one could trade using equations instead of indicators. And that even in a randomly generated portfolio using randomly generated prices, one could use what is described as equation 16 in that paper as an exponential alpha generating function. 

All the equations are in my papers. The mathematical foundation on which my trading methods are based are there for all to see. What you are asking for is: can I have the scripts generating those charts for free? Since it would really be the only way to show that in fact there are no errors in the programs: logical, mathematical, statistical or procedural for that matter. Well, I must respectfully decline.

My trading procedures have a Buffett style investment methodology, in general, they try to do three things at the same time. This latest  trading strategy is no different and has these 3 objectives to meet:

  1. to Buy & Hold as much as possible and for as long as possible,
  2. to accumulate shares to hold over the entire trading interval as much as possible, and
  3. to trade over the accumulative process within all the portfolio constraints (available capital and trying not to go bankrupt).

One can find explanations and governing equations in all my papers to accomplish this task.

I do think that anyone having scripts that produced the charts already shown would not give them away or even provide clues as to how to duplicate his/her work. But that is what you have in my papers and research notes. You can reverse engineer the equations provided to then restructure your own trading strategies and trading procedures that can accomplish similar or better results. But don't ask me to just give it away free, just like that, to anyone who is not even ready to consider the effort needed to accomplish these performance results.

My trading procedures have adopted a Buffett style method of trading. And in this regard, even if a stock within the portfolio would be an Enron, it would be discarded (with its losses) long before it would go bankrupt or have much of a negative impact on the portfolio.

That is why the methodology uses over-diversification in the first place, so that any one stock that goes negative would have a minimal impact over the long term portfolio objectives. And to answer Manuel, I played Enron on the way down and at some point accepted the loss with the firm intention of not doing that again.

Here are the few stocks that I have tried after answering Edward's post where I “claimed” that all that was required to increase performance was to increase the profit extraction rate. Testing conditions were the same: 1,500 trading days (5.83 years), $100k initial capital and same script.

(click to enlarge)

AAPL  11/23/2011 ::  Trade Clustering AGQ  11/23/2011 ::  Trade Clustering AMZN  11/23/2011 ::  Trade Clustering

DDS  11/23/2011 ::  Trade Clustering IMAX  11/23/2011 ::  Trade Clustering PNRA  11/23/2011 ::  Trade Clustering

SHS  11/23/2011 ::  Trade Clustering SINA  11/23/2011 ::  Trade Clustering ULTA  11/23/2011 ::  Trade Clustering

November 23

I concluded my last post with: I can push performance even higher simply by increasing the profit extraction rate. It is an easy statement to make when you don't have to back it up or prove it.

The only way is to show that in fact, it can be done. After making the changes to the script, it was ran once under the same testing conditions. Here is AAPL again, where the request for more profits is being answered:

AAPL Nov. 23 

November 23

As was said previously, the AAPL chart shows only for the last 220 trading days of a 1,500 day testing period; an interval (2006-11) which more than covers the financial crisis where even AAPL “dipped” some 60% in 2008-09.

This particular trading strategy has 3 objectives to meet. 1- to Buy & Hold, 2- to accumulate shares over the entire trading interval, and 3- to trade over the accumulative process.

If one only did the Buy & Hold by, being fully invested there is no way even on AAPL to push your profits over $12M on a starting capital of $100k; all you could do would be to reach about $400k. It is by executing 1, 2 and 3 that you can reach the $12M profit mark.

AAPL Nov. 22

To generate the profits, the strategy executed 3,756 trades of which 2,943 were closed, meaning that the accumulative process is still holding some 813 positions; a lot more than the 10 positions or so that could have been acquired with the initial $100k. And the trading procedures are ready to keep on accumulating shares having more than ample cash reserves to add another 300 positions if the trading equations triggered such trades.

The trading strategy might be seeking full market exposure but it does not unload its entire holdings except in special situations and in the case of AAPL, none of those were triggered that I am aware of.

What I find interesting in the AAPL chart is its ability to take a profit bite here and there and keep on doing it at almost every price swing. And that is where a good part of the $12M profits come from. And having a relatively high ending cash reserve only mean that I can push performance even higher simply by increasing the profit extraction rate.

It is not my best but then who is perfect. What is attempted is to trade on each price cycle. Naturally you want to do this with a profit, if not immediate then you are ready to wait a little. Overall, it does display some nice behavioral characteristics: buying dips even if that was not your intent and selling on the rise which is surely a way to put money in the account.

Trading over an accumulative process has some advantages as it can help increase performance without necessarily having to predict where the price is going.

Modified ... November 26, 2011 © Guy R. Fleury. All rights reserved.

Alpha Project