Data - Where from and how much?
- Thread starter nizar
- Start date
tech/a
No Ordinary Duck
- Joined
- 14 October 2004
- Posts
- 20,403
- Reactions
- 6,323
Try Premium data Richard Dale.
Will be meaningless if your thinking you need go back that far for systems testing.
90% of stocks listed in 1960 wont be there today.
T/T was designed on 8 yrs of data.
Will be meaningless if your thinking you need go back that far for systems testing.
90% of stocks listed in 1960 wont be there today.
T/T was designed on 8 yrs of data.
Yeh and I recall you saying you saw that as one of its weaknesses.
But you're right, the further you go back potentially the more inaccurate the data is and the more the market dynamics have changed.
tech/a
No Ordinary Duck
- Joined
- 14 October 2004
- Posts
- 20,403
- Reactions
- 6,323
Nah Not me!
Its been argued by many that this was a weakness I remember Duc having a good bash at statistical significance.
Just get a heap of different Bourses and test them both in and out of sample (If you wish) I have 8 bourses and did/do use them during test design,part of the Just Data package.
The best results came from the Hong Kong Bourse.
Richard Dale
Norgate Data
- Joined
- 22 February 2005
- Posts
- 269
- Reactions
- 197
Firstly, the quality of data out there available before 1992 is questionable - even the ASX can't sell you anything before 1990.
Pretty much all of the data vendors history that goes back this far is riddled with duplicate days etc. I think this may be indicative of the quality of data available at that time.
As has previously been raised, having LISTED stocks going back that far is fairly useless if you don't have the delisted data going back that far too. You're just introducing significant survivorship bias.
Lastly, the ASX changed character significantly in 1992 due to compulsory employer superannuation.
Cheers,
Richard.
Pretty much all of the data vendors history that goes back this far is riddled with duplicate days etc. I think this may be indicative of the quality of data available at that time.
As has previously been raised, having LISTED stocks going back that far is fairly useless if you don't have the delisted data going back that far too. You're just introducing significant survivorship bias.
Lastly, the ASX changed character significantly in 1992 due to compulsory employer superannuation.
Cheers,
Richard.
