Toby Crabel's methodology

Discussions about the testing and simulation of mechanical trading systems using historical data and other methods. Trading Blox Customers should post Trading Blox specific questions in the Customer Support forum.
blueberrycake
Roundtable Knight
Posts: 125
Joined: Mon Apr 21, 2003 11:04 pm
Location: California

Toby Crabel's methodology

I was recently rereading Toby Crabel's book, and here's how I see his methodology.

1) Identify a relatively common trade setup with a positive expectation (in his case Opening Range Breakout)

2) Analyze the preceeding price action to see if some price patterns are more likely to result in the primary setup being successful.

So a system based on his methodology would have a trade entry, that is taken only if one of a particular set of price patterns has previously taken place.

The obvious problem here is that while the main trade setup may occur quite often, some of the preceeding patterns that he identifies occur quite rarely, making them very hard to validate statistically.

I was wondering if anyone else has explored this methodology, and has had success with it, in out of sample testing or in real trading.

-bbc

richard
Roundtable Fellow
Posts: 94
Joined: Tue May 18, 2004 10:17 pm
I can't afford his book

They are asking \$500 or more for it

stancramer
Roundtable Fellow
Posts: 60
Joined: Sat Feb 07, 2004 10:34 am
Location: Washington DC
I'm experimenting with a system that fits your 1) 2) methodology above, but it isn't very Toby-Crabel-esque since it doesn't enter intraday.

The system is approximately the Congestion Phase system of Eugene Nofri, as described in Perry Kaufman's book Trading Systems and Methods. In the 1987 edition it's on page 217 and in the 1998 edition it's on page 292.

The price pattern beforehand, your item 2), is to observe that price has been contained within a congestion channel for a period of time. The trade entry trigger, your item 1), is the occurrence of a number of consecutive closes in the same direction. When item 2) happens and then item 1) happens, the system enters a trade.

The system has been extended and expanded upon by Robert Barnes, in Chapter 8 of his book Trading in Choppy Markets. He programmed it and ran some backtests across a range of parameters. So far, anyway, it appears promising.

richard
Roundtable Fellow
Posts: 94
Joined: Tue May 18, 2004 10:17 pm
it sounds similar to an approach discussed by Larry Connors and Linda Raschke, with some twists credited to someone named Gipson.

Take the ratio of 6 day to 100 day historical volatility. If it falls below 50%, and today is either an inside range day or the lowest volatility day of the last 4 days, enter the next day. There are variations that look at a high ADX to indicate trend, and that use DMI to decide whether to go long or short.

This approach relies upon the fact that volatility oscillates, so it finds low volatility times that look like they will breakout.

In one variation, you enter, then do a stop-and-reverse if the underlying or future goes the other way. I have found this pretty useful actually, at least so far.

(I use mechanical signals to enter but exit discretionarily)

Dean Hoffman
Roundtable Fellow
Posts: 87
Joined: Wed Jul 02, 2003 3:04 pm

Toby Crabels Methods

A lot of the ideas in Tobys book are avilable for \$27 at:

You realize pretty quick that this stuff isnt instant gold. You need to do A LOT of work with it.

Dean Hoffman

blueberrycake
Roundtable Knight
Posts: 125
Joined: Mon Apr 21, 2003 11:04 pm
Location: California
One observation I've made while testing patterns, is that when you dont have a very large sample size to work with, it's sometimes better to simplify your data rather than looking at raw trade results.

For instance, initially I spent some time algorithmically searching for patterns on one sample set, and then if the results looked good (win rate, profit factor, avg win / avg loss, etc) verifying them on an out of sample time period. The performance correlation was mediocre.

However, recently, I made the following change. Instead of looking at raw performance, I transformed my data by (1) throwing out all the small winners/losers (2) changing all big losers to a value of -1 and changing all big winner to a value of +1. Then when I searched for patterns, I was simply looking for ones that have a consistently higher ratio of these big winners to the big losers. While I wouldnt claim this to be the wholy grail of pattern validation, I have seen some more promising results using this approach. If anyone else is doing something similar, I'd love to discuss this in more detail.

-bbc

flex
Contributing Member
Posts: 7
Joined: Wed Apr 14, 2004 4:04 am
Firstly, to 'Richard' - You don't need to pay \$500 for the Crabel book. Check out the Stocks & Commodities web site. For just a few dollars you can purchase a PDF that goes through the Opening Range Breakout method.

To the original question posed by Blueberrycake.
I have done some testing of Crabel's work, I even programmed an indicator that gives me the 10 day avg 'Stretch' he describes.
When analysing Crables work u need to remember that it was written before the 24hr mkt was devised so in that regard I think it has less relevence than it once did (i.e. the open doesn't have the same 'heat' it once did). However, there are a few mkts that still only trade their own hours even in this computer age (I'll let u work out which mkts they are) & the work I've done on these mkts shows that the ORB stuff of Crabel's still works.
So hit your Tradestation & code in the Stretch & the preconditions required (maybe as a Paint Bar) & go from there.
The signals don't pop up often (say 30-40 times/year) but that's often enough.

Murray Ruggiero
Senior Member
Posts: 41
Joined: Mon Oct 11, 2004 3:35 pm

Crabel and Futures Magazine

I am currently writing a series of article on pattern based trading, I discussed and tested the Crabel patterns in upcoming futures article, three part series. The series is running in the Oct,Nov,Dec issues this year I think. I did my analysis based on some measure of range not fixed points.

RKM
Contributor
Posts: 4
Joined: Thu Sep 16, 2004 3:14 pm
I have done some testing on ORB based on Crabel and it shows validity. They key question is variation of bet sizes..

RKM