|
PureBytes Links
Trading Reference Links
|
John,
Thanks for the detailed reply. I guess my data provider is doing a pretty
good job for me. It is very rare to get bad data and their automated tool
to adjust for splits and dividends saves me a lot of time. My analysis
program doesn't like negative prices and the auto adjustments done by my
data provider will create negative prices if I include a lot of history. So
before I run my analysis each night, my program runs through the stocks
flagging any with negative prices historically. I shorten the history on
those files and then let the analysis run. It seems that this occurs 1 - 3
times per week. On the weekend I check for commodity contracts that need to
be rolled to the next month. So my manual time spent each week to keep the
data how I like is less than 1 hour. Everything else is automated and very
reliable. I am currently doing this process with the stocks that make up
the S&P 500 and also about 70 commodity contracts.
Thanks again for your input.
Tim
-----Original Message-----
From: owner-metastock@xxxxxxxxxxxxx
[mailto:owner-metastock@xxxxxxxxxxxxx]On Behalf Of jhmtn
Sent: Tuesday, January 23, 2001 3:50 PM
To: _Metastock List Server
Subject: Re: stock splits
Tim,
Right now my Metastock data comes in end of day. (My real time provider
uses a composite page that I can't get into Meatsotck format.) My provider
has a two hour window between West Coast market close and when we are
suppossed to perform our downloads that they use to scrub the data. If
there was a split, they adjust it during the scrub. But sometimes, the
scrub doesn't get the split adjustment correction until the next day if
there was an abnormally high amount of "dirty" data that day to correct.
Also, I would have to delete and re-add the download to be able to
redownload it from the vendor. As there is a cap on the number of quotes
per month before I shift to a higher monthly charge, I tend to not want to
delete and reload 10+ years of daily data for every stock split.
All this considered, given the small amount of time total to maintain the
stock price history DB, it is just easiest and frugal for me to handle the
splits.
What is worse than the splits, is the oddball bad data that inevitably makes
its way into the quotes. Some are obvious, such as a price quote that is 2
or more times yesterday's high or low without a volume spike. The vendor's
"scrub" usually finds and corrects these, but some get through. The Equis
Downloader has a function to check an issue for that spike up or down in
price, but that is for one issue only and using it is not feasable for more
than a dozen or so stocks. Again, I wrote a quick Exploration to locate
potential ones of these. I finally realized that the split finder would
find these anomalies, and just run it.
There is one other category of errors that all data vendors seem to fall
prey to at one time or another. That is they sometimes just miss a quote.
That leaves a whole in the data. The problem with this type of error is
that it looks the same as an issue that is no longer trading under that
ticker due to name/ticker change, being bought out (or bankrupt, or went
private, or became delisted, or ...... etc.). Even with the S&P 500, which
I track in sector folders, there are adjustments and movement in and out of
the index to track. So I had to write another Exploration to find issues
which downloaded no data today and manually validate them by check Web quote
history sources.
Worse still is when the vendor's data scrub just doesn't catch bad data due
to multiple reasons. Maybe I downloaded to early, or maybe the data was
just very dirty that day. This happens once or twice a year. When I find
several hundred or more anomalies in my scans, I use a day data stripper
that the vendor provided (it was supppossed to be a Holiday date stripper
originally). I run my normal trade finder scans, ignoring the ones with bad
data. After I've made my selection(s) for the day, I then run the day
stripper and remove that day's activity for all stocks in the DB and attempt
to download again. If I still find the same number of errors, I print the
list to a file and send it attached to an eMail to the vendor then call them
the next day to find out what their proposed solution is. I've never
experienced more than 1 day without those anomalies being fixed. (The
vendor tells me it is related to gliches with web traffic and/or errors on
the part of their data providers.)
After I run my trade hunting Explorations, I manually examine each potential
trade find. Every now and then, price action will just look unreasonable
and upon verification of a price or two at the Yahoo (or CNN or MSFN etc.)
web site will show the anomally. In the worst case, the error had gone on
for several months for one issue and represented a glitch in the vendor's
processes. I called them the next day and they found and corrected to price
data (which I did delete and reload) and their processes (which took about
2-3 days to correct for all stocks).
Maintaining our own price history DB's takes time. There is just no way
around that. I have been caught with seriously contemplating a trade before
I realized the data was incorrect (and to my knowledge, I've not yet place
atrade based on bad data -- "knock on wood!"). Generally this happened when
I only ran my error scans on the weekends or even skipped a week or two.
I've learned to take the extra 10 minutes or so a day and validate my data
download as best I can prior to running my trade hunter Explorations. In
his book "The Ultimate Trading Guide" (which I highly recommend), John Hill
(founder of the publication "Futures Truth"), devotes a section (pgs.
166-171) to data. Hill references the September 1999 article by Sheldon
Knight in "Futures" magazine in which Knight analizes the differences in
data vendors. Knight found that some vendors report a bid or ask as a high
or low and other vendors only use actual trades. So if true highs and lows
are important for your technique, you need to validate with you data vendor
which method is being used !! Hill further reports a test of various data
vendors (Bridge, CSI, Dial, Omega, Reuters and 5 others) to determine how
"clean" the data was for Soybeans and the S&P500. Bottom line for the
results was that ALL 10 vendors tested showed bad data to one degree or the
other! The book was published in 2000, so the results are fairly recent.
The only defense is to do what you can to verify your data download daily!
Hope this helps! ........... John
----- Original Message -----
From: Tim Stevenson
To: metastock@xxxxxxxxxxxxx
Sent: Tuesday, January 23, 2001 6:20 AM
Subject: RE: stock splits
Thanks for the reply John. What does your data provider do to affect the
split? Mine will adjust the data automatically. This will produce negative
prices historically in some cases. I'm not sure if this adjusting history
to reflect splits is OK or not. I just want to be sure that the prices
reflect the true price swing turning points. The software I have written
identifies price swings and their levels.
Thanks again,
Tim
-----Original Message-----
From: owner-metastock@xxxxxxxxxxxxx [mailto:owner-metastock@xxxxxxxxxxxxx]On
Behalf Of jhmtn
Sent: Sunday, January 21, 2001 1:12 AM
To: metastock@xxxxxxxxxxxxx
Subject: Re: stock splits
Tim,
I have 60+ Metastock folders holding some 2,100 stock issues. In addition,
some of those are duplicated in multiple folders by price, exchange, or
issue type. It includes the S&P 500.
The way I handle the split issue is simple. I do my daily price download
then I run a small eploration I wrote that finds the common 2:1 splits, then
another one that finds the 3:2 splits. I check the lists produced against
the Yahoo site, then use my data provider software to afect the split.
Generally, it only adds about 3 - 6 minutes to my nightly preparation time.
Then, on the weekend, I again go to a web site such as Yahoo and scan their
split list. From that I pick up the oddball splits that I missed such as
the 5:1 and the 1:8. The weekend effort takes on average about 10 -15
minutes. A small price to pay.
In addition to this, I do a more manual review of a potential trade before I
take it. This will show any other odball conditions that surface. Plus,
I've downloaded BigEasyInvestor and run its update about twice a week. It
reports splits due for the week and splits that occurred today at market
open. Altogether, the effort to maintain the DB for 2,100 issues is about
1/2 hour per week total, with some weeks taking up to a hour for that week.
I've heard that some data providers will update the issue's back history if
there is a split, but the one I use doesn't unless I delete and readd the
issue, which I don't.
Hope this helps ......... John
----- Original Message -----
From: Tim Stevenson
To: metastock@xxxxxxxxxxxxx
Sent: Friday, January 19, 2001 6:12 AM
Subject: stock splits
I have been trading mostly commodities and have developed a system that I'd
like to run against stocks as well.
My question involves stock splits and stock dividends and capital gains. My
data vendor is CSI and they include an option in their Historical Stock
Adjustments settings that allow for adjusting for stock splits and also for
stock dividends and capital gains.
What have you found to work well with dealing with these issues? I haven't
dug too deeply into this yet, but I'm thinking I might have to use the data
from the most recent split and go forward. This requires constant
monitoring of when splits are occurring. This isn't a big deal on a few
stocks but I am running the system on the 500 stocks that make up the S&P
500.
Thanks,
Tim
|