20 Jan 2009 @ 11:47 PM 
 

I’m an apalling blogger, I admit that. Having received several encouraging emails from people interested in Hurst’s theories, I realised that I haven’t posted anything here for a while. In fact I have been working away, burning the midnight oil trying to get the software to really do what I set out to have it do: analyse the market the way Hurst would have done. I’m pleased to say I’ve made some progress. Now I am working on updating the screensaver aspect of the software so that it performs exactly as the main software works. On the way I have encountered many interesting problems, and I’m going to start sharing those on this site, which I guess is what blogging is all about! Today I’m going to write about the whole screensaver idea.

The screensaver

I have been asked why on earth I’m building a screensaver as a part of the software package. The reason is very simple. At the moment when loading data and creating a chart for the first time the software takes about 4 minutes to perform the phasing analysis on a chart with 15 years of history (on my computer, which is fairly fast, but not a super-computer by any means). Then every day it updates the analysis, a process that takes about 30 seconds. Twice in every trading cycle (and I am mostly looking at the 20 and 40 day cycles at the moment) the software does a complete re-evaluation of its phasing analysis, a 3-4 minute process. It’s all very well stepping the software through its paces every day, and seeing how it performs, but that doesn’t help one to know that in fact the software is “working” – in other words giving good advice about when to buy and when to sell. The only way to do that is to look back at a history of several years worth of the software doing its thing. To generate that history it is necessary to step the software through the data day-by-day, which is a process that really ties up your computer, and isn’t very entertaining. Which is why I am building the screensaver. From the software you allocate a chart you are working on to the screensaver, and then when you take a break the screensaver starts building the history. It’s a process that seems to work fairly well. The screensaver also builds a “trading journal” so that you can go back and see why the software bought and sold when it did. For me this is the key to making the software really work. Of course I can only expect the software to work as well as an expert in Hurst’s theories would work … and although there is some evidence that the theory works, the real test will be the trading history generated by the software, and the ongoing trading the software does.

Backtesting

Backtesting has earned itself a very bad reputation in trading circles, mostly because it is possible to optimize the parameters a system uses so as to perfectly fit the historical data, and produce a fabulous trading history that simply doesn’t work when you use the same parameters on new data. So is there a danger that my screensaver is going to fall prey to this dreaded problem (which has been the downfall of many trading systems – and is related to the problems Rich Swannell encountered with his Refined Elliott Trader because he presented hypothetical results rather than real ones)?

The simple answer is no! But I should explain that. There are two reasons why I don’t think optimized backtesting is going to distort the results generated by the screensaver:

  •  There are no parameters involved in what my software is doing. I am not optimizing any parameter values. The software performs phasing analysis in a way very similar to the way Hurst prescribed – by using a visual “pattern recognition” approach rather than a mathematical analysis, and so there are no parameters to optimize. The software won’t be changing what it does (or the way it does it) because of the data it encounters. I will write soon about how the software performs phasing analysis because that is key to what it does.
  • Secondly the screen saver actually works with truncated data – even though the data for the “future” is available to it when backtesting, it truncates that data and works only with data up to the day it is testing. This means that any mathematical calculations (such as moving average calculations) cannot use data that would not be available if you were running the software on the date in question.

There are two ways in which the screensaver’s results will not match the real performance of the software:

  • I am not using “tick” data, and so the software doesn’t detect situations where a market moves too fast for you to get in or out at a particular price. If the market traded at a certain price (it is included between the high and low price of the day) then the software assumes you traded at that price. Of course when actually trading there is often “slippage” – failure to trade exactly at a particular price.
  • The other difference is that there is no human involved in the testing process. As much as I have great respect for humans, they can make some really foolish trading decisions – and when back testing the software doesn’t have a human filter saying “Buy? What nonsense … there’s obviously something wrong with this stupid software. I’m certainly not buying now.” Right before the biggest rise in the history of the stock you’re trading. I know it happens – I’m just such a human!

OK, so that’s it – why I’m creating a screensaver. The other thing about it is I think it looks cool! Here is a sample:

screensaver in action 

screensaver in action

You’ll see the phasing along the bottom of the chart, as well as the buy and sell zones (those are my creations, they aren’t strictly Hurst). And the dominant envelope, VTL’s and FLD’s. The beige area is a pause zone (PZ). Volatility is given at the top left, and that about covers it!
Tags Categories: Project Progress Posted By: admin
Last Edit: 20 Jan 2009 @ 11 49 PM

EmailPermalinkComments (0)
\/ More Options ...
Change Theme...
  • Users » 1
  • Posts/Pages » 26
  • Comments » 19
Change Theme...
  • VoidVoid « Default
  • LifeLife
  • EarthEarth
  • WindWind
  • WaterWater
  • FireFire
  • LightLight
Orient didn't invent the power-reserve indicator, Jaeger-LeCoultre gets those bona fides; however, starting in the rolex replica 1950s, Orient did manage to put them into affordable watches in a way that really none other has achieved. Watches with power-reserve indicators often rolex replica cost over $1,000. While Miyota has developed a cheaper mechanism with a power-reserve indicator, Orient's is a true in-house rolex replica sale movement: built on its base 46N45 caliber that's been well refined over the years, offering 40 hours of hublot replica uk charge (as all the models in this review do, in fact). The power-reserve indicator on the Orient FFD0F004W0 here makes it a true bargain at just over $450. Placed unexpectedly at the top of the dial, it both rolex replica sale enhances the Orient FFD0F004W0's elegant symmetry, and also provides functionality not always present in more traditional dress rolex replica uk options. The 40mm case diameter looks contemporary on the replica watches sale wrist, while stopping short of being too big for a dress watch.