|
PureBytes Links
Trading Reference Links
|
Robyn Greene wrote:
> I've been interested in bad data ever since - about 4 years ago - I and 3 of my
> online friends compared MCO numbers - and we each had different numbers (and
> different data vendors). Turns out not a one of us had the correct number. And
> the most technical of us (the engineer) said - how can I ever risk money on the
> basis of systems which are developed or traded using faulty data.
>
> At this point - I only get about 5-6 important data streams now (i.e., the ones
> that are in my systems) - and I check them out on a random but routine basis.
> Frankly - I don't know how people who download 2000 stock prices a day could do
> that. Robyn
I download about 2000 stocks eod each night and wind up spending a good
deal of time maintaining this database - much more time than I want to
spend. I use dmi and chartscanner to pick up bad price data. If the
spread is very wide it's usually a bad price bar. Bad volume is another
problem. I know if vol is 2x average, and if it's zero. If vol is 2x
normal on an unremarkable price bar i check it, and if it's zero i check
it. Sometimes you don't see these things till you actually look at the
chart, perhaps a week or more after the fact. For historical data on a
single symbol on a specific date in the past, I usually use
http://www.iqc.com/quote/history.asp
The scarcity of error-free data is the reason I use eod data. If I had
to spend my life editing bad ticks and battling server anomalies I'd go
insane.
Monte
|