Page 1 of 1

George Pruitt really needs to get a copy of Trading Blox!

Posted: Thu Apr 16, 2009 7:47 pm
by Mark Johnson
George Pruitt is Director of Research at Futures Truth magazine ( ). He wrote an article for issue #1-2009 of the magazine, entitled "The Prioritized Step Search", in which he bemoaned the fact that a full Stepped Parameter run of his system optimization in Tradestation "took over 5 hours" (p. 38, first paragraph).

George's optimization seemed pretty simple to me; its features were
  1. First Parameter: 4 steps
  2. Second Parameter: 4 steps
  3. Third Parameter: 19 steps
  4. Fourth Parameter: 19 steps
  5. Portfolio = only one futures market (Crude Oil)
  6. EOD data, 1/1/1986 to 11/30/2008 (roughly 5800 bars)
Whoop-de-do. 5776 tests (4 x 4 x 19 x 19) of 1 system on 1 market, for 23 years. For Blox users, a rather small problem. So I coded up George's trading system in Trading Blox, and ran the same optimization using Blox on my Windows XP machine. It took 296 sec (4 minutes + 56 seconds) to run, which is a factor of SIXTY faster than George Pruitt's run on Tradestation. Five minutes instead of five hours.

George really needs to get a copy of Trading Blox!

George's system is a Lee Gettess / Larry Williams style Volatility Breakout system using EOD data: buy at (Open + (param3 x volatility)), sell at (Open - (param4 x volatility)). I've attached my Blox source code, along with my parameter setup values (encapsulated in the Simulation Suite). George, and anybody else who's interested, can download the code and experiment with it in Trading Blox.

Edit: added "Volatility Breakout" discussion

Posted: Sat Apr 18, 2009 7:38 am
by AFJ Garner
I wonder whether it might be worth copying this Blox to the Blox Market Place?

Posted: Mon Apr 20, 2009 9:49 am
by Mark Johnson
I'm not so sure it was a tremendously valuable system as published; it seemed to me to be more an illustrative example than production workhorse. There was no positionsizing; it traded a single contract, regardless of equity level. There was no portfolio; it traded a single market. It was a pure reversal system, which reduced the number of trading rules to just two:
  • Reverse and go long today, if today's candlestick is white and unusually tall
  • Reverse and go short today, if today's candlestick is black and unusually tall
I'll see if I can find a simple way to tune it up to trade a portfolio using more realistic positionsizing. If so (no guarantees), I'll put the tricked-out version into the Blox Marketplace.

I notice that the unmodified, just-for-illustrative-purposes code above has been downloaded 38 times (and still counting). I wonder whether anyone who has downloaded and run that code, would like to express any opinions, favorable or otherwise, about the trading system itself?

Re: George Pruitt really needs to get a copy of Trading Blox

Posted: Fri Nov 06, 2009 4:13 pm
by LeviF
Mark Johnson wrote:George really needs to get a copy of Trading Blox!
Looks like he did:

Posted: Fri Nov 06, 2009 8:11 pm
by danZman
This makes no sense to me regarding speed. TS is very fast. I don't believe for a second such a simple backtest would take that long. TS lets you backtest using genetic algos.

The only problem with TS is no portfolio backtesting...which really sucks. There's a lot of free downloadable stuff to make TS kick butt. COT info is now a simple indicator

But when it comes to knowing the "truth", I'd take TB over TS.