Monday, 1 December 2008

New S&P 500 Setup Bearish, All Five Commodities Bullish

Interesting new data from today's holiday-delayed Commitments of Traders report. My trading setups based on the reports are giving me these new signals that would have been effective on today's open: bearish for my new S&P 500 trading setup (more on it below) and bullish for crude oil. Earlier signals from past weeks had also given me bullish signals for gold and the BKX Bank Index today. The two new commodity signals mean all five of my five highly correlated commodities are now bullish. See my latest signals page for all the details.

Please note, however, my post of last Friday, where I talk about my decision to temporarily combine my COT signals with technical analysis for trading. (The reason is I want to test my setups on detrended price data and apply Monte Carlo testing or some other randomization test to reduce the risk of having found my setups by pure luck.) I didn't enter any trades today owing to the breakdowns on the charts. (There were very bearish pullbacks from resistance on the 60-minute charts as defined by an interesting system developed by Tom DeMark, about which you can read more in this interview with DeMark here and in this excellent primer from a trader who uses it to great profit, Stephen Vita.) One promising result of my testing so far is that I came up with the same trading setup for the S&P 500 using raw price data as I did using detrended price data. (Detrending the data is a way to remove a market's net bullish or bearish bias. This is vital in order to ensure a trading strategy isn't just benefitting from parameters that are slanted to take advantage of this bias. Such a setup might not work as well in other market conditions.)

Speaking of the S&P 500, I've chosen a new setup I'll be following for this market. I've put up those results on my latest signals page. I chose it based on a slight re-weighting of the measures I use to score my best setups, with a little more weight given to my walk-around tests and use of what I think are more robust out-of-sample measures. As I said, I thought it was a good sign that it scored the best out of the setups I looked at on both raw and detrended price data. Still need to do a Monte Carlo test, but I'm working on developing that. You can see all the specs for the setup in the "notes" on my latest signals table. Or if you want to really get your hands dirty, check out my DIY page and download it for yourself. I welcome your thoughts.

Note (added Dec. 4): I should clarify that this setup is temporary until I've finished my Monte Carlo testing.

TAGS: gold, BKX, Bank Index, crude oil, S&P 500, SPX, COT, Commitments of Traders, derivatives, Black Swans, market timing, trading system development, CFTC, Commodity Futures Trading Commission, COTs Timer, out-of-sample testing, walk-around testing


Anonymous said...

Comment: in the PDF download, the column "Date of COTs Report That...." is too narrow. Can't see the year for many of the signals.

Question: When do you update drawdown? Looking at the British pound which went bullish in Dec 07 (maybe 06? the year is clipped) the pound is down way more than the 14% shown.

Alex Roslin said...

Thanks for pointing that out. I've widened the columns a little. The drawdown is the past entry-to-intratrade loss during the backtested period. It's updated when I update my backtesting with new data starting in spring to summer 2009.

The update should probably not be done until about 18 months to three years after my last update, based on the size of the dataset (approx. 13 years since the start of the COT data). But I might start a little early because I'm eager to get the recent data from the market crisis into the dataset and see how that shakes things up.