October 22nd, 2015 

A Case Study

Over the last week or so, I've had some discussions on my trading methodology. One of which centered around a demonstration of what it could do. In reply to a friend's statement, I said: "my program would have done that too on that stock", which was to buy shares during the last price decline about a month and a half ago or so. I realized afterward that it was very easy to say.

You make a claim, don't provide anything to corroborate it, and voilà, it should be accepted at face value. Well, there should be some skepticism thrown in somewhere, understandably.

The solution was: run the program on his stock. It will be easy to see if, yes or no, it would have done the same thing. I got the 12.88 years of NDAQ price data and ran the program. And show how the strategy behaved over the period in question, among other things.

It resulted in the following chart:

     NDAQ October 14 test
NDAQ Oct 14

And yes, the program did buy shares during the last price decline. I would add, it did not break down and would have provided more than decent results.

The DEVX6 program was last modified in June 2014 over a year ago as printed on the chart. So all the trade data you see should be considered as a walk forward, or at the very least, an out of sample test. The program had never seen any NDAQ data before and is, therefore, dealing with 16 months of what should be considered its future. It is the same as if you had the trading strategy on automated mode since then, with no intervention whatsoever, facing its unknown future.

The trade placement on that chart says a lot. It looks like all trades are sold at a profit! Yes, they are. The red and green line seems to separate trade clusters. Yes. However, those lines are programming afterthoughts. They are not used. I wanted to have a visualization of where the trade clusters were. So, I added those lines. I can say that now that they are there, they could be used in the trading process itself since they were pretty good at isolating most of the clusters. The why the lines are not used is simple: the trading logic was designed prior to adding them, and the program did not need them. All trade entries are answering to conditional random-like functions.

The strategy (DEVX6) is basically a variation on Buy & Hold theme. I view it more as a Buy & weak Hold.

                            The program stands ready to let go of some shares for a profit, otherwise it holds. For the shares sold, the intention is to buy them back, at a discount, later on.

As simple as that, but still 2,028 lines of Wealth-Lab code to make it happen, and that is a lot considering.

The following stats, some of which are printed on the chart, were no surprise, at least to me, as they were a direct consequence of the trading procedures used:

Time: 12.88 years (all the available NDAQ data EOD)
Type of orders: all market orders, at the open, the next day, whatever the price

Number of trades: 8,715
Number of closed trades: 7,036
Still opened positions: 1,679

Number of profitable trades: 8,715
Profitable still opened positions: 1,679

Maximum consecutive winning trades: 8,715
Number of losing trades: 0
Maximum consecutive losing trades: 0

Winning percentage: 100%
Trade unit: $5k
Trade size: from 100 to 900 shares

Initial capital: $2M
Long term CAGR: 25.04%

On drawdowns? You are mostly in a Buy & Hold scenario, so you should expect about the same as the market average, something in the range of 50%. Even Mr. Buffett has suffered 50% drawdowns on 4 occasions during his career. It is part of the game. However, in the chart presented, the largest monetary drawdown observed was 21%. But that is easy to evaluate using the formula: DD*Cap*(1+CAGR)t. In this case: 0.21*2M*(1+0.2504)12.88.

Notice that whatever the outcome, the largest drawdown will always be somewhere in your more distant future and will be highly dependent on your future portfolio size. You are playing a CAGR game after all. Also, if you put in the time, it will get bigger than the initial capital, at times many times over. For instance, using the above numbers, it would be about a $7M drawdown (21%). And if you put in more years, it will be larger still. During the last financial crisis, Mr. Buffett's drawdown was in excess of $100B, now that is a drawdown! He stayed the course, bought some stuff. With time, got it all back, and made new highs.

For me, the chart is a resounding success. Over the 12.88 years, the strategy accumulated over 240,000 shares now valued at near their historical high. Using this strategy, it appears that any stock at their historical highs would have this 100% winning trades! And it is within the nature of the market to make new highs.

The ending cash reserve was about $20M so the program could keep on accumulating with ease. If you liquidated all shares in inventory, you would get the $33M profit printed on the chart. You ended with 60% in cash. Even a 50% drop in the stock price would represent only a 20% drawdown of the stock's weight in the portfolio!

The following graph might be more explicit at showing the equity:

   NDAQ – Equity Curve (12.88 years)
NDAQ Equity $5k trading unit

Being totally scalable, you could double the fun by putting in $4M and using $10k as trading unit. It would result in a chart like this: 

   NDAQ – Equity Curve (12.88 years)  Capital: $4M. Trading unit: $10k
NDAQ Equity $10k trading unit

It becomes a matter of taste and available trading capital. One could also do less naturally, divide by 2 and go for the $2.5k trading unit and get about half of the first profit chart, or a quarter of the second.

The strategy starts trading slow, $5k at a time, distributes these bets on its timeline. Over the 3,344 trading days investment period, it took 8,715 positions. It also cashed in 7,036 of those positions, at a profit, (84% of total) while holding the rest waiting for a still better price. The unsold shares are all at a profit as it is, every single one of those 1,679 positions. There is even $20M in cash reserves waiting to be put to good use (10 times the original stake).

You could quit the game at that very instant, liquidate all remaining positions, be profitable on 100% of all the trades taken over the past 12.88 years. And have in your account what is printed on the chart: a $33M profit.

It is a trading system where each time a stock makes new historical highs will be profitable on 100% of its trades. All that it requires is the answer to one question: can this stock, that you want to trade, make new highs in the next 5 to 10 years? The stock's story and its fundamentals should help answer that question.

If the answer is yes, then each time you will have new highs, you will be 100% profitable. It is also the best time to say: I quit, and thank you for all the pennies. But I would recommend staying on. Stocks that make new highs tend to make more than one.

If the answer is no, then you should short the thing, or ignore that stock. If you can't see the stock price going up in the next few years, then why buy in the first place? There are a lot of stocks where you can answer: yes. Do your research. This trading strategy could be performed on hundreds of stocks, even on hundreds traded at a time, with just that one question to answer, the same for all of them.

All this led to what will follow...

Created... October 22nd, 2015 © Guy R. Fleury. All rights reserved. 

August 24th, 2015 

Alpha Project