November 6th, 2016

A Stock Trading System – Part I

In my last article: A Tradable Plan – Part I, it was expressed that only 3 numbers, three portfolio metrics are sufficient to summarize all the trading activity and trading history of any stock portfolio over any duration.

Those numbers were: n, the number of trades, u, the trading unit used (bet size), and PT, the average percent profit per trade, profit margin, edge, or, whatever you might like to call it.

It is simple: three metrics giving: n*(u*PT), average percent profit/trade multiplied by the size of the trade unit, as a dollar amount, which is multiplied by how many times it is, was, or will be done. That's it! That you trade in a discretionary fashion, or have your computer do the job. Total profit(loss) generated by any stock trading strategy = n*(u*PT).

It does not matter what trading strategy you used or intend to use, it will end up with those 3 metrics when you tally the final results, or any result in between. Three very simple numbers, nothing esoteric, no emotions, no psychology, no preset notion, just plain numbers.

It reduces all of what was, is, or will be accomplished to bean counting. It can be used in the making of estimates of what is to come.

It is not very complicated: three numbers, two of which you can set yourself, leaving you with one "unknown" that you can approximate based on "realistic" portfolio level simulations. Meaning you do not mold the results to your liking by peeking ahead, selecting only data that works, over-optimizing, over-fitting, or rearranging trading decisions after the fact to suit a data set. Whatever you do, any trading strategy where you cheat in some way, or to some degree, has for result that you are really only cheating yourself by giving you a distorted version of reality.

Should you ever want to apply such a trading strategy in real life trading, you will see that the future wasn't based on your modelled trading strategy. You will simply blow up your trading account, or call it quits before it is completely depleted.

The solution is simple there too, you don't "prearrange" your backtests to make them look good, even for yourself. Show they can work with real data. Not on everything, you don't build a portfolio with everything in it, you will remain selective in real life. Those trading rules can be programmed just as the stock selection process can. It is not because you finally do something on automated mode that you lose your insights or your common sense.

If only three portfolio metrics can say at any time how much profits were made, it in fact becomes your portfolio profit status. It is the what you have achieved with all the efforts (n trades), with all your trade entry decisions (n*u), and with all your trade exit decisions (n*u*PT). This, over the life time of your portfolio.

Those 3 metrics don't really care that you are talking about a simulation over some past data, or considering the life of a portfolio over the next 50 or 100 years for that matter. The way to tabulate the end results will be the same: always the same metrics, for any portfolio, no exception.

Therefore, if you fix the trading unit to something of your liking, say $2k, $5k, ..., $100k+, you solved one third of the problem instantly. No prediction needed there. Your bet size is your trading unit, the how much you are ready to put on a single trade when you consider to execute on a trade opportunity.

You can fix another of the metrics by giving a profit objective to your trading unit. As simply as saying: I want 10% profit on my trades (u*PT). So, depending on your bet size, you are looking for profits of: $200, $500, ..., and, $1,000+ on their respective trading unit bets.

This sets two of the metrics not as unknown, but known before the fact, before you even start trading.

That is what you want to do, that is what you are ready to accept, that is what you will program your machine to do. If your system takes a trade, it will wait for its profit, then take it. Evidently, that can be done a lot of times, a really large number of times. And that is where n comes in. You wanted n to be large anyway, so you build your trading decisions based on the trade opportunities you can capture that can generate your percent profit PT.

How many plays of 10% are there in the stock universe you intend to play with? You could look at the past and get an estimate, there is more than sufficient data to do so. Let me give you a spoiler: you won't be able to trade them all for the simple reason, you won't have sufficient funds to do so. And furthermore, if you could get them all, you would "be" the market. Now, that is big. All you can touch is a minuscule minuscule fraction of what is available. So, I don't think you will miss anything by making your estimate of n based on your ongoing equity since you already know the size of your trade unit.

Should I even ask the question: will there be 10% stock price plays in the future?


November 6th, 2016 © Guy R. Fleury. All rights reserved.  

Alpha Project