Essence of Trading

Just writing a very simple sampling program to test various trading strategies on actual stock data: this seems the most productive way forward at the moment - characterising the "nature" of stock movements is a bit to complex for me at the moment!!

Ideas stripped down look like this.

Trading is a game which involves stepping between two side-by-side conveyor belts, which constantly change their rate and direction.

It is simpler to consider one (money) as static and the other (stock) as moving (though this makes the difference a more complex function). We step from money into stock, travel a bit and then step back.

The object of the game is to get as far up the money escalator (to the right) as possible as this escalator is next to all the other escalators (can buy anything with it).

The optimum solution (to be tested) is to be in Cash as the stock escalator passes to the left and then to step onto the escalator as it changes direction and starts to move right. We leave a doughnut or something to mark where we were. We stay on the stock escalator until another hypothetical point H (the High) passes the doughnut and we step back onto cash escalator. This way we have travelled up the cash escalator.

Superficially I'm noting this is like a Turing machine. Turning machines have States which are like the stock and they travel up and down the data-tape or cash "register" ;-) Whether this similarity is of any use only time will tell.

It seems a game we can't lose on. However there is the problem that we have to pay to change escalator .. in other words we have to take a step down the cash escalator every time we step off or back to the cash escalator. It means that we have to be able to think head just a little bit to ensure that we will recover that initial step.

To ensure that we always move right we need to determine these H/L points and whether they are far enough apart to make shifting escalators worth while. That is all there is between being broke and making a fortune...

But the escalators do not move randomly. There are only so many people allowed on stock escalators and to get on someone must step off. We have to decide who to swap places with. If we think that someone is standing near a L then we are more likely to swap places with them so the escalator gets dragged up so we can cross over. They in turn get to step onto the cash escalator higher than where they were. The opposite happens also dragging the escalator down.

So the actual movement of the escalator is caused by these differences of trade. There is a strong feedback process where the response of players to previous movements is actually causing the current movement. This is most evidential in the fractal patterns in the market.

Fractal systems are iterative systems where a simple input data is fed through a system repeatedly. It means that the basic "motif" or pattern is very simple but by repeating it at different scales a very complex structure of repeating units is created. This is the stock price trace (SPT - my term) the price plotted against time. IN essence it is simple but the simple patterns are repeated in complex structures. I have lost it now but I had a wonderful plot of the gold price I think at 5 day, 3 month and 1 year and the three plots could be virtually over laid. There is certainly no way of telling from an SPT what the duration is... except perhaps the macroeconomic features like inflation and economic growth.

Boxes

To try and characterise SPTs I'm using an idea of Boxes. They are essentially the standard "candles". A box contains the SPT between two dates. How the stock gets from the entry point to the exit point is irrelevant (a black box). We only know its max and min prices (top and bottom of the box) .

The problem then is to replace an SPT with a succession of boxes the only variable between the width of each box. We want each box to not contain too much information about the SPT preferable a straight line between entrance and exit is best, but most importantly a box should start and end at an extreme or turning point. This way each box represents an optimal trading window or period, and that is all we are interested in.

Now because the charts are fractal we have to chose a scale, or trading scale. Apparently short trades are traditionally of period of months to a year and long trades many years. Day traders trade between days and weeks. From experience as a private investor with access to delayed chart data and high costs of entry and exit to the market I've only succeeded a few times in day trading. The advantage is however that through compound interest a small percentage each day compounds to huge gains a year. IgIndex and the like is perhaps a better type of gambling for this scale. Really I'm aiming at the short scale: months.

Anyway that is progress so far and now back to coding... end of Xmas holiday tonight so this may get stuck for a while...