up or down bias in back adjust continuous data?

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.
Post Reply
yoyo2000
Roundtable Fellow
Roundtable Fellow
Posts: 58
Joined: Fri Jan 30, 2004 10:37 pm

up or down bias in back adjust continuous data?

Post by yoyo2000 »

Hi,guys,recently I read an article about continuous data,which said, "When early contract prices in a concatenated set are significantly less than their real contract counterparts, they tend to produce a bias that in simulated trading would heavily favor the act of buying over selling. In addition, even if the early contract prices are not significantly different from their current-contract counterparts, inflation could play a role in influencing buying over selling when such a long series is introduced as representative of current pricing norms. This phenomenon should tell you that your results may be invalid and that applying in the present what you have learned by simulating the past can distort your trading algorithm. "
(from: http://www.csidata.com/cgi-bin/getManua ... =essay.htm)


but I can't catch that where these bias come from?

regards.
7432
Roundtable Knight
Roundtable Knight
Posts: 199
Joined: Wed Dec 31, 1969 7:00 pm
Location: sandy, ut

Post by 7432 »

the bias comes from adjusting a back-adjusted contract up or down to close the gap from one active month to the next. if the adjustment is always the same direction far out data can look cheap representing a bullish bias on a chart and in the data. in may of 1984 the june 30yr US bond was actually trading about 51 points, on a back-adjusted basis it was trading -9.

one of the nice things about csi is that although they limit the number of futures markets you can have(or charge you for more) they don't seem to have a limit on the number of different ways you can download each market.
so set up some market as back-adjusted, then perpetual, then nth nearest, then gann, then back-adjusted with the proportional box checked. roll through the charts and you can see the difference.

for crude, a perpetual contract trades around 20 from 1987 to 2002, a back adjusted contract was trading in the 30's during the late 80's and around 70 in 2002.
you might really question your system when you look at interest rates. back-adjusted bonds look like a smooth bull ride for 30 yrs, but perpetual contracts are choppy with some big swings in the short term rates.

you can download 4 different versions of the same contract for all the markets you trade, make 4 copies of trading blox, set up the futures dictionary in each trading blox to use the different types of data and run your systems to see how much difference it would make. did you pick up the same trades in each format? re-optimize using perpetual and see if your parameters change much.
then you can check trades by going back to csi and downloading the actual contract data from the contract month traded and see if it was really possible to catch that move in crude during the summer of 1990.
if your back-adjusted data shows you getting short the 30yr in feb of 1984 at -2.25, pull up the actual june 84 contract and see if your system would have shorted at 53 1/2 using the actual data.
yoyo2000
Roundtable Fellow
Roundtable Fellow
Posts: 58
Joined: Fri Jan 30, 2004 10:37 pm

Post by yoyo2000 »

Got it,thank you,7432.

The point is historical reality v.s. trading reality.
And I think,mybe it's reasonable to remove the bias from inflation when there are more than 10 years,but others bias is from the adjusting rules of continuous data,it's inherent.

regards.
Post Reply