Saturday, 28 February 2009

Small Makeover for Gold Setup

My trading setup for gold based on the Commitments of Traders reports has gotten a small makeover. The new specs and results are all posted on my latest signals table. The change is that I've tweaked the parameter values for the large spec total open interest setup. The upper signal line has gone from 0.2 standard deviations above the moving average to 0.25. The lower signal line went from 0.6 standard deviations below the average to 0.65. Not a big different, but those values produced slightly better results in my walk-around testing - i.e., my check on the backtested performance of neighbouring setups that have slightly altered parameter values. A robust setup should have neighbouring setups that also performed well. Taking a look at them is an important way to reduce the risk of curve-fitting and improve the odds of good performance in actual trading. Right now, the new setup is in cash (like the old one). Now that I've finalized my testing on the S&P 500 and gold - at long last! - I'm planning to get to work on new setups for a couple of markets that aren't strongly correlated to the first two: the Nikkei and natural gas.

7 comments:

Unknown said...

Thnx for the new info. Was curious if you have tried to model a COT setup for gold stocks that uses not only gold but the S&P as well. Given gold stocks seem to be correlated with either gold at one moment or the general equity market the next, I wonder if there is any way to tease out a robust setup.

Alex Roslin said...

Hi Ian,

Thanks for your interesting suggestion. I did long ago find some potentially useful setups for the gold COT data against the HUI Gold Bugs Index and TSX Gold Index, but there were very large drawdowns and the robustness wasn't as high as for the COT gold data against bullion itself. Of course, I could trade the signal with a gold stocks ETF, but I find that occasionally - like when there's a big market selloff - the gold stocks will go down while bullion goes up. I find the more closely related the instrument traded is to the actual data in the COT report, the better the results of the setups. Mind you, I haven't returned to check a gold stocks setup in a while with my currently improved testing, so who knows what could be found out there!

On your other point, gold actually has only a 0.28 correlation with the S&P 500 between 1995 and 2007.

Regards,
Alex

alysomji said...

Hi Alex,

I was wondering what software you use to identify and test trading setups?

Thanks

alysomji said...

Hi Alex,

I was wondering what software you use to backtest your trade ideas?

Thanks

Anonymous said...

Hi Alex,

I was researching how to use the COT after I have read about it from a book of fundamental basics of forex trading. I came to witness your blog, CFTC and Cotchartprice.

After I have studied your info on your blog, I found that it is well researched and produces results. I am very new in forex trading though I understanding technical analysis and fundamentals.

To support these technics I wanted something even better that can generate robust results.
I trade basically the euordollar pair. Is it possible that you assist me to draw a setup for this pair that would generate signals as with your gold and s&p?

I am more than willing to collect whatever data that would be needed to fulfil this setup.
I have looked on the DIY page, but some info is rather confusing to me since i'm new in this field.
I also came to see http://www.cotpricecharts.com/commitmentscurrent/ but I saw they are using the Futures only data for the eurodollar. Therefore I was not sure whether that is good data to use for this setup.
Would you help?

Regards,
Tobela

Alex Roslin said...

Hi Alysomji,

I use a combination of an app developed by a very clever and generous reader and a bunch of Excel spreadsheets. I think the testing could be automated in Excel by someone with enough dedication and/or expertise.

Regards,
Alex

Alex Roslin said...

Hi Tobela,

Thanks for your message. I've been asked lots of times to develop a setup for the Euro-USD pair, but the problem is it hasn't had enough history and thus enough data to generate robust setups or to include varying market conditions. Perhaps with the recent ups and downs of the pair, it will be worthwhile searching for a setup when I start my planned round of setup revisions with data as of mid-2009.

I've also found that using the futures and options data tends to lead to more robust and performing setups than the futures-only data.

Regards,
Alex