deep data history
deep data history
Does anybody know where I can get some very long term futures data, preferably backadjusted but not vital ? I use CSI and they are very good but only have long history for a few commodities.
-
- Roundtable Knight
- Posts: 2071
- Joined: Fri Apr 25, 2003 3:33 pm
- Location: London
- Contact:
http://www.globalfinancialdata.com/index.html
See if these people have what you need.
See if these people have what you need.
Wow!!!! & Double WOW!!!!AFJ Garner wrote:http://www.globalfinancialdata.com/index.html
See if these people have what you need.
Gold & Silver cash prices back to 1252!!!
(Testing a dual moving average crossover system with slow MA=200 years and 'fast' MA=20 years would be fun)
-
- Roundtable Knight
- Posts: 2071
- Joined: Fri Apr 25, 2003 3:33 pm
- Location: London
- Contact:
Yep, very interesting stuff. Bear in mind the further back you go the prices will be monthly or yearly. Bear in mind also that "cash" prices make no allowance for what we know know as contango and backwardation.rhc wrote: Wow!!!! & Double WOW!!!!
Gold & Silver cash prices back to 1252!!!
(Testing a dual moving average crossover system with slow MA=200 years and 'fast' MA=20 years would be fun)
Anthony & rhc, thank-you so much for posting links, this is just what I wanted. Like most, I'm in a long drawdown and have customers who ask me where I see a way out of this and if history is any guide to "what happens next". I realised that by only going back to the early 1980's with current data everything is dominated by an environment of generally falling interest rates but I really want to see how our model behaved during a bear market in bonds and also the interim period before they started the current long bull run. Most of this occurs around the late 1930's and early '40's but, as Anthony points out, the market structure at the time also has to be included if the results are to be accurate.
-
- Roundtable Knight
- Posts: 2071
- Joined: Fri Apr 25, 2003 3:33 pm
- Location: London
- Contact:
When Bonds Fall
To reinforce the concerns over data periods, the attached paper is of interest. CSI futures prices on the 30 year US Bond commence in 1977 (excuse my laziness but presumably from the commencement of T Bond futures). Thus available bond futures prices commenced only 3 years before the current 30+ year bond rally. Take a look at US corporate Aaa Bond futures for the period 1920 to date under the different interest rate environments.
What research there is to be done! To acquire data on the whole range of asset prices. To compare asset price correlation over the decades. To test mechanical trading strategies. To ascertain volatility and its quality for TF (noisy/not noisy). There are problems of course; the lack of daily data, the need to build in an approximation of futures market structure.
Not that it did JWH any good. I seem to remember from their website back in the middle of last decade that they made a similar boast to Winton: they had obtained grain prices back to the time Joseph saved Egypt from famine and apple prices back as far as the Garden of Eden. In the latter case that would approximate to 13.7 billion years of data if you are a fellow enlightenment enthusiast or a mere 6,000 years if you are a religious fundamentalist.
Excuse the levity. For those with a serious interest in the future of investment generally and TF in particular, the matter deserves at least a modicum of thought.
What research there is to be done! To acquire data on the whole range of asset prices. To compare asset price correlation over the decades. To test mechanical trading strategies. To ascertain volatility and its quality for TF (noisy/not noisy). There are problems of course; the lack of daily data, the need to build in an approximation of futures market structure.
Not that it did JWH any good. I seem to remember from their website back in the middle of last decade that they made a similar boast to Winton: they had obtained grain prices back to the time Joseph saved Egypt from famine and apple prices back as far as the Garden of Eden. In the latter case that would approximate to 13.7 billion years of data if you are a fellow enlightenment enthusiast or a mere 6,000 years if you are a religious fundamentalist.
Excuse the levity. For those with a serious interest in the future of investment generally and TF in particular, the matter deserves at least a modicum of thought.
- Attachments
-
- Welton-When_Bonds_Fall_(Visual_Insight_Series).pdf
- (1.28 MiB) Downloaded 667 times
This is just nuts. Need to get my hands on that somehow.AFJ Garner wrote:http://www.globalfinancialdata.com/index.html
See if these people have what you need.
Pinnacle have adjusted data back to 1969 in some cases (beans, wheat). Good company, good data: http://www.pinnacledata.com/clc.html#includes
-
- Roundtable Knight
- Posts: 126
- Joined: Mon Nov 07, 2005 3:39 am
- Location: Silicon Valley, CA
- Contact:
jas - so is CSI deep enough if you have the entire history they offer? I know for newer subscribers they charge a bit extra for this deeper history? For some US rates and series the St Louis Fed in the Fred database has data back to he 1950s.
As I'm sure you are aware of if one uses cash markets from Global Data for longer history then you will need to think hard about the impact of the roll or lack thereof in those prices.
As I'm sure you are aware of if one uses cash markets from Global Data for longer history then you will need to think hard about the impact of the roll or lack thereof in those prices.
svquant - CSI works perfectly for us on a daily basis, we cross-check it with 2 other data sources each morning and it barely misses a beat.
For research purposes though, we require much longer history. Originally we thought 30 years was sufficient, but we have recently decided to focus less on the portfolio level results and more on each asset class and even down to the individual contract to see more detail. This obviously involves many more periods of in/out of sample data.
We also run a genetic algorithm which benefits from as broad a database as possible, if only for proving it's pitfalls......this is just stress-testing if you like and the more variations of stress we can put the model through the better.
For research purposes though, we require much longer history. Originally we thought 30 years was sufficient, but we have recently decided to focus less on the portfolio level results and more on each asset class and even down to the individual contract to see more detail. This obviously involves many more periods of in/out of sample data.
We also run a genetic algorithm which benefits from as broad a database as possible, if only for proving it's pitfalls......this is just stress-testing if you like and the more variations of stress we can put the model through the better.