Testing on Electronic data v Pit 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
ladidalimey
Roundtable Knight
Roundtable Knight
Posts: 197
Joined: Thu Sep 21, 2006 6:21 am
Location: London

Testing on Electronic data v Pit data

Post by ladidalimey » Sun Dec 07, 2008 1:02 pm

Hi,

I'm trying to get some opinions on back testing and when pit data is valid (up until mid 2006?) and when to switch to electonic (mid 2006 onwards?). As more of the pits seem to dwindle I'm comparing my systems on both and feeling I should dump pit data altogether.

System results between 2007 to present seem reasonably similar, but if I go back much beyond that, the electronic behave quite differently. Is this because electronic trading was less used on some contracts?

Thanks for any ideas/advice

L.

Aviator
Contributing Member
Contributing Member
Posts: 8
Joined: Sat Nov 29, 2008 3:23 pm
Location: Jacksonville, FL

Post by Aviator » Wed Dec 31, 2008 9:48 am

I have just recently come across this issue in my testing -- largely because I have been working on a shorter-term trading model. I wish I had more to offer on the topic, as I am still learning myself, but one thing I have thought about is the following:

Many indicators/patterns/formulas etc. attribute significance to the close and in many cases the relationship of the close to the high/low of the day. The close in pit trading vs. electronic is theoretically very different as one occurs in a high liquidity environment (afternoon) and the other occurs in the evening. Moreover, the factors affecting price action could be different in the evening than during the day, perhaps containing more sentiment related to expectations for tomorrow, which in turn would give the close a different character as it were.

I was working with the idea of "weak" or "strong" closes on short term market swings, and realized that the way I was defining it might not be valid for a 'round the clock, continuous trading environment. I certainly agree that pit vs. electronic raises a lot of questions, I would love to hear more views on the subject.

sluggo
Roundtable Knight
Roundtable Knight
Posts: 2986
Joined: Fri Jun 11, 2004 2:50 pm

Post by sluggo » Wed Dec 31, 2008 10:13 am

Other systems, generally of the longer-term variety, use one price per day. Traditionally, people have used the Close as the one price per day, for a few different reasons: (1) It was perceived that the Close was the "highest liquidity" time of the session, so it represented the opinions of the greatest number of buyers and sellers; (2) It is the freshest, most recent price; (3) Everybody else is using the Close in their mechanical systems, so it must be a good idea.

For example, moving average systems take the moving average of, what? Answer: traditionally, they take the moving average of the Close.

But some traders may wish to reexamine this entrenched policy, in light of the market environment we face in early 2009. Maybe, the Close is not the most appropriate choice, in some cases.

For example, Mr. John Ehlers (bibliography) writes books about indicators that smooth price with not a lot of delay, so-called "low lag moving averages". He uses ((H+L)/2) rather than the Close, as his choice for the representative price of the day.

In the era of 24 hour trading, perhaps another choice might be the average of the Opens of the 24 one-hour-bars that make up a calendar day. (Or the 12 hours from 6 in the morning to 6 in the evening).

Some may decide to look at tick data or 5 minute bar data and try to identify times of highest volume (presumably, "highest liquidity") in the morning and the afternoon, and call these times the "Sort-of-Open" and the "Sort-of-Close". Then it will be necessary to decide how to handle Daylight Savings Time, and also to decide whether to permit different Sort-of-Open times for different markets. Is it a good idea for CME Swiss Franc to have a different Sort-of-Close time, than CME Australian Dollar?

LeapFrog
Roundtable Knight
Roundtable Knight
Posts: 695
Joined: Mon May 17, 2004 4:18 pm
Location: Boston, MA

Post by LeapFrog » Wed Dec 31, 2008 11:17 am

We talked about this issue over a year ago...I know I posted about it as well and looked into various solutions along the lines of what Sluggo suggested above. I like the idea of parsing intraday data around volume and creating synthetic daily bars, effectively throwing out thin volume overnight type data.

Unfortunately, there is no practical way of doing this that is within at least my financial means. What skimpy intraday data is available (on the 100s of commodites around the world) much of it is of questionable quality.

My way of dealing with this over the past 12 months has been to continue trading my long term systems which mainly use the Close prices and to use greater discretion in the application of my short term systems with the aim of beating those systems. So far this has worked out fine.

If anyone has knowledge of good intraday data that includes reliable volume data at an affordable price I would be all ears.

Then, there is always FOREX trading.

Post Reply