Hello everybody,
I'm new to futures, LTTF and this forum. Currently I'm in the process of finding my system and setting up my operation. I'm pretty sure by now that LTTF and a mechanical system suit me well. Therefore good data quality and a long history are important.
I came to the conclusion to order data from CSI and I have two questions to the community and the more experienced people:
a) Do you have any comments regarding CSI as data provider? Would you approve of this decision?
b) How far back shall I order historical data?
CSI offers some markets - especially the agriculturals - all the way back to 1949. However, in the net I mostly find comments about testing the last 20 to 25 years. Are data from 1949 to 1980 useful to have and to include in tests? If no, why not? If yes, why do you think most people only use 25 years?
Regards,
Robert
Data history: how many years back is useful?
-
TrendMonkey
- Roundtable Knight
- Posts: 154
- Joined: Fri Apr 22, 2005 9:14 pm
- Location: Vancouver, Canada
I signed up with CSI a couple week ago, so far so good. They were very responsive, the software works great, and is is blazing fast.
My main interest/concern with data is the effect of different futures back adjusting algorithms. CSI has five or six. My first little experiment with different algorithms was disconcerting; going back only to November 2004, there is a huge spike visible on one chart and not the other.
I haven't tried to understand this yet, though, so perhaps its actually OK.
My main interest/concern with data is the effect of different futures back adjusting algorithms. CSI has five or six. My first little experiment with different algorithms was disconcerting; going back only to November 2004, there is a huge spike visible on one chart and not the other.
I haven't tried to understand this yet, though, so perhaps its actually OK.
-
Jason
- Roundtable Fellow
- Posts: 57
- Joined: Sat Apr 26, 2003 10:31 pm
- Location: Oakville, Canada
- Contact:
Using CSI is a good decision.
A number of years ago when I first ordered CSI, I took a day off from my job at the time figuring that I would be able to just set it up and start testing. I ended up taking the rest of the week off. I could have somehow got my hands on a portfile.adm file and moved on. However, understanding what you are doing when creating these contracts was and is critical IMHO. The scenic route is well worth your time.
To this day, a larger than expected amount of my research time is spent either within CSI or a data table of some sort.
Have fun!
Jason
A number of years ago when I first ordered CSI, I took a day off from my job at the time figuring that I would be able to just set it up and start testing. I ended up taking the rest of the week off. I could have somehow got my hands on a portfile.adm file and moved on. However, understanding what you are doing when creating these contracts was and is critical IMHO. The scenic route is well worth your time.
To this day, a larger than expected amount of my research time is spent either within CSI or a data table of some sort.
Have fun!
Jason
-
Roger Rines
- Roundtable Knight
- Posts: 2038
- Joined: Wed Oct 06, 2004 10:52 am
- Location: San Marcos, CA
We have been using CSI data for a long time and think they are the best daily download service we have ever used in almost 20-years of trading.
If you are going to get into long term trading you'll find their Unfair Advantage Software to be very handy at keeping your back-adjusted long term contracts updated.
Good luck with your venture!
If you are going to get into long term trading you'll find their Unfair Advantage Software to be very handy at keeping your back-adjusted long term contracts updated.
Good luck with your venture!
-
Roger Rines
- Roundtable Knight
- Posts: 2038
- Joined: Wed Oct 06, 2004 10:52 am
- Location: San Marcos, CA
Hello Robert,
For testing a system you can't have enough data. So the longer the contracts you have, the more likely you'll get a sense of how the system will trade those markets over that period of time.
To generate signals, this depends upon the type of system you are using and how much data it wants to see for making decisions. For example, a system may not take a new high or a new low that it hasn't seen before. To feed this approach, give the system all the data you can find so you'll have as relistic an exposure to the data as is possible. Not all system need a lot of data, but with today's computers being so fast, we don't think it is worth the effort to maintain a long and a short series of contract files. Everything we trade uses the same set of files no matter what system we are applying. All this means, look at the rules of the system and consider if the data range you are providing will do the best job possible.
Another point about data is to look at the length of time of the trades. If you try to trade from individual contracts that are tied to an expiration month, your system might loose the signal when you move from one expiration to the next. This is one of the main reasons we only trade from back-adjusted continous contracts. CSI's UA will create these contracts for you, but be sure to understand how they are being spliced so you can gage whether your trading approach will follow the splicing action. If it doesn't you could find yourself rolled into the next month and have to adjust the signal prices using the difference between what your trading and what the splicing action is doing when it updates the data.
We get around this splicing problem by updating our contracts by hand. This way we can always have our contracts reflect our actions in the market. It also prevents any bad data prices from getting to the signals.
To understand how you might trade, make some decision on when you'll roll your positions forward. Some traders use a fixed calendar setting that varies by market. Others watch the money on the floor and move when the floor changes its front month. In some case you might find that you'll need to use both approaches.
To state the obvious, a lot of trading is understanding what to do when. There isn't much out there that talks about the tiny details that are important when you pull the trigger, so consider developing that experience by paper trading for a while so you'll get some insights into these areas.
For testing a system you can't have enough data. So the longer the contracts you have, the more likely you'll get a sense of how the system will trade those markets over that period of time.
To generate signals, this depends upon the type of system you are using and how much data it wants to see for making decisions. For example, a system may not take a new high or a new low that it hasn't seen before. To feed this approach, give the system all the data you can find so you'll have as relistic an exposure to the data as is possible. Not all system need a lot of data, but with today's computers being so fast, we don't think it is worth the effort to maintain a long and a short series of contract files. Everything we trade uses the same set of files no matter what system we are applying. All this means, look at the rules of the system and consider if the data range you are providing will do the best job possible.
Another point about data is to look at the length of time of the trades. If you try to trade from individual contracts that are tied to an expiration month, your system might loose the signal when you move from one expiration to the next. This is one of the main reasons we only trade from back-adjusted continous contracts. CSI's UA will create these contracts for you, but be sure to understand how they are being spliced so you can gage whether your trading approach will follow the splicing action. If it doesn't you could find yourself rolled into the next month and have to adjust the signal prices using the difference between what your trading and what the splicing action is doing when it updates the data.
We get around this splicing problem by updating our contracts by hand. This way we can always have our contracts reflect our actions in the market. It also prevents any bad data prices from getting to the signals.
To understand how you might trade, make some decision on when you'll roll your positions forward. Some traders use a fixed calendar setting that varies by market. Others watch the money on the floor and move when the floor changes its front month. In some case you might find that you'll need to use both approaches.
To state the obvious, a lot of trading is understanding what to do when. There isn't much out there that talks about the tiny details that are important when you pull the trigger, so consider developing that experience by paper trading for a while so you'll get some insights into these areas.