Hi Keith

I missed this thread so I apologize for the late reply.  

I did a memtest86 and HD scan etc when I originally built the machine a little 
over a year or so ago.  While I haven't checked the memory since, I regularly 
run various HD intensive stuff (defraggler, looking for dupe files across all 
drives, large backups/copies etc) and I do a pretty regular diskchk so that 
should be ok.

I think the problem (which has since reappeared since I've rebuilt everything) 
is more a problem in code "somewhere" (in AB or otherwise) rather than a random 
bit in memory getting flipped (which I think would result in the OS crashing).

While I haven't done any extensive tests, speaking anecdotally, I think the 
problem is more likely to come up if/when the quotation data cache size is at 
or exceeds the max value.  Unfortunately, it takes a while to run an explore 
using backfills across so many symbols, so I don't/can't necessarily just 
hunker down to try and troubleshoot the problem.  I hope to figure it out one 
day though.

Thanks very much for your input.


--- In [email protected], Keith McCombs <kmcco...@...> wrote:
>
> Kurasake --
> Could it be that you might have a hardware problem?
> To test you RAM, run MemTest86+ at least overnight.  Free from:
> http://www.memtest.org/
> 
> and your hard drive(s), run SpinRite. $89US from:
> http://www.grc.com/cs/prepurch.htm
> This is an excellent product, though a bit expensive and geeky.  Manual 
> is available at:
> http://www.grc.com/files/sr5_manual.pdf
> That is the latest manual, even though the latest version of SpinRite is 
> 6.0.
> And for everything you don't need to know about SpinRite:
> http://www.grc.com/files/technote.pdf
> 
> You might read the following for more info on other HDD testing and 
> repairing ideas:
> http://www.techsupportforum.com/hardware-support/hard-drive-support/461234-questions-about-spin-rite.html
> 
> -- Keith
> 
> On 4/30/2010 14:23, kurasake wrote:
> >
> > Hi Patrick.
> >
> > Thanks for taking the time to write so much.
> >
> > I've cross checked a few symbols in AB vs QT and I'm definitely only 
> > seeing the spikes in AB. For example for WDC I'll see the spike in the 
> > daily chart in AB, I'll switch to the 1-minute chart in AB and see the 
> > spike(s), note the date/time and look at the symbol in QT and scan 
> > through the entire range of date/time in QT for WDC and everything is 
> > normal. Another way of looking at it for me is that while the bad data 
> > spikes don't happen all the time or on every symbol in AB, I've never 
> > seen bad data spikes in QT (minute view, historical or raw data 
> > window) so I'm fairly certain the issue has to be in in AB or 
> > something related only to AB.
> >
> > I would also tend to believe that if the bad data were consistent 
> > across QT and AB that it would most likely be a problem in the IQFeed 
> > data stream or the IQ agent in which case I would think it would be 
> > affecting a large number of their customers.
> >
> > Curiously, no matter what symbol, the spikes always seem to pop up 3-5 
> > days prior and after I perform an explore in AA. So I could have 
> > looked at WDC earlier, and everything could've been fine. Then I'll 
> > run an Explore and then look at WDC again and the spikes would appear. 
> > I also don't think I've ever seen the spikes occur for the current day 
> > or for any day >5 days ago which is strange (though I'm not 100% 
> > positive about this part). This further increases my suspicion that it 
> > is something in AB.
> >
> > The spikes also only seem to appear (or at least I notice it more) 
> > when I've already looked at a large number of symbols, so I'm thinking 
> > it might be an issue with the DB/memory cache but it's still hard for 
> > me to rationalize how/why it happens. I've tried playing around with 
> > the database and memory cache settings before but I guess I'll try 
> > other combinations.
> >
> > As for the large number of symbols, I wish I could settle on 200 or so 
> > but my "system" (for lack of a better word) relies on a combination of 
> > technicals and candles and I don't always get the right signals 
> > triggered if the number of symbols I'm scanning across is too small. I 
> > don't always have that many symbols though and there are days when I 
> > don't need to perform an Explore and the spikes aren't so numerous as 
> > to be overwhelming (though some days are worse then others).
> >
> > I'll figure this out eventually.
> >
> > Cheers!
> >
> > --- In [email protected] <mailto:amibroker%40yahoogroups.com>, 
> > "NW Trader" <pk47hargus@> wrote:
> > >
> > > Hi Kurasake,
> > >
> > > I know exactly what you are experiencing. However, the bad data is 
> > in QT, but then once you have imported it to AB, it will persist. It 
> > will most frequently occur when you do a backfill of some or all 
> > stocks in QT, but depending on your data source and the spike settings 
> > you have defined in QT, it can happen anytime -- especially on 
> > volatile opening, closing, or FOMC days. And I see spikes virtually 
> > every day on symbols like SPY just after the close. Be aware that the 
> > spikes do not always appear on a QT chart, especially the instances of 
> > a 0 value for a close on an intraday bar (like a one minute bar). But 
> > they are there and you can find them by opening a QT data window 
> > (<CTRL>D) and scrolling to the time of the bad data.
> > >
> > > Here is my solution. First I've created an exploration which I run 
> > (daily setting in AA) on all symbols, at least once a day. The output 
> > of the exploration is the current Close, the HHV of high for the past 
> > 21 days, the LLV of low for the past 21 days, and a 5 period ATR 
> > divided by the close. Why 21? Well I have a version of QT which allows 
> > 20 days of historic data before it starts to drop off, so this will 
> > cover all QT data available on a given day. Why the ATR/C ? This is a 
> > quick volatility tool which can help spot when a stock may have a 
> > split, an outlier data point, or some other radical move.
> > >
> > > So you run the scan and identify the stocks that have data that 
> > needs correcting. Now open the Quote Editor in AB and sort the columns 
> > (e.g., if you have a LLV(L,21) of 0, sort the lows and find the bars 
> > where 0 exists. You could delete or edit those bars, and exit the 
> > Quote Editor, however this won't fix things so long as the bad data is 
> > in QT. So once you've identified the time of the bad data in AB, you 
> > have to find and delete the bad data in QT. This can be accomplished 
> > either with the QT data window (go to the bad data and delete), or 
> > zoom the QT chart in on the time period with the bad quote (even tho 
> > it may not be visible on a QT chart) and press <SHIFT> <CTRL> L to 
> > remove a low point (may have to do this more than once if multiple 
> > minutes are bad) or <SHIFT> <CTRL> H to remove a high quote.
> > >
> > > Splits can also be adjusted in QT, as well as in AB. See the help 
> > files for each program, tho if you fix data in QT it should be ok in 
> > AB. Again, always fix the QT data first, then AB.
> > >
> > > Finally a gratuitous observation & suggestion. IMHO, trying to track 
> > 1500 stocks intra day is neither worth it (given all the data 
> > maintenance you will be doing) nor going to leave you much time to 
> > trade. Consider narrowing the stocks you are following intraday to 
> > those you realistically would trade based on price, volume, and your 
> > other trading metrics. I now track approximately 200 symbols intraday, 
> > importing them to AB, but I trade only about - 2 to 3 dozen regularly 
> > (based on my Price/Volume and other metrics). There are an additional 
> > 4 dozen or so that are possible trade candidates that are not 
> > regularly traded but may move into the regular camp as others drop 
> > out. The remaining symbols are indexes, sector ETFs, and longer term 
> > holds and swings that I want to generally keep an eye on. For me this 
> > works well as I can thoroughly know the stocks I trade regularly; I 
> > study their intraday support and resistance levels, daily patterns, 
> > know when they report earn ings or have other news releases, and such. 
> > I didn't always have 3 dozen traders -- years ago I tried to follow 
> > hundreds, but realized that the concept of span of control was causing 
> > me to miss many trades and make some bad decisions on trades I took. 
> > So I focused on about a dozen stocks and over the years I've expanded 
> > the number of my traders based on experience, knowledge of those 
> > stocks, and developing a trading rhythm for how and when I trade.
> > >
> > > Of course if you want to track 1500 stocks for end of day study 
> > purposes, consider either a separate data base set to daily data, or a 
> > different data provider.
> > >
> > > This works for me, but your mileage may vary.
> > >
> > > Peace and Justice --- Patrick
> > > ----- Original Message -----
> > > From: kurasake
> > > To: [email protected] <mailto:amibroker%40yahoogroups.com>
> > > Sent: Thursday, April 29, 2010 12:19 PM
> > > Subject: [amibroker] Easy data refresh?
> > >
> > >
> > >
> > > I've been having a problem with bad data appearing in AB (abnormally 
> > high volume or wrong OHLC) in AB when I perform an "Explore" from the 
> > "Automatic Analysis" on a large number of symbols (>1500) when I have 
> > "wait for backfill" checked even for something as simple as filter=C>O.
> > >
> > > Fortunately the bad data points are very easy to identify and seem 
> > to occur on the same days for the affected symbols and since I haven't 
> > been able to resolve it, I've sort of learned to live with it (Note, I 
> > don't get any bad data in QuoteTracker running on the same machine 
> > using the same RT data feed so I'm pretty sure it's in AB).
> > >
> > > Anyway, when I see the bad data, I right click on the "Plug-in 
> > Status" area of the status bar and do a "Force Backfill" and all the 
> > data is correctly filled in. My question is, is there an easier way, 
> > either from the menu (so I can setup a macro) or better yet, via some 
> > code/formula that will
> > > 1. Keep all the symbols but delete all the EOD and RT data
> > > 2. Automatically "Force Backfill" from my RT feed?
> > >
> > > If someone has an idea on what's causing the problem in the first 
> > place, that would be helpful too but I've posted this regarding this 
> > problem before and haven't seen a solution so I'm guessing I'm a 
> > unique case.
> > >
> > > I've also tried flushing the DB cache and playing around with 
> > different database parameters, in-memory cache size, setting up a new 
> > database etc.
> > >
> > > Thanks.
> > >
> > > Roy
> > >
> > > System:
> > > Windows XP Pro/Sp3
> > > Intel Core i7, 4GB Ram, many gigs of HD free
> > > AB v5.3 (tried uninstall/reinstalling)
> > > RT Data from IQFeed using the latest client
> > >
> > >
> > >
> > >
> > > ------------------------------------
> > >
> > > **** IMPORTANT PLEASE READ ****
> > > This group is for the discussion between users only.
> > > This is *NOT* technical support channel.
> > >
> > > TO GET TECHNICAL SUPPORT send an e-mail directly to
> > > SUPPORT {at} amibroker.com
> > >
> > > TO SUBMIT SUGGESTIONS please use FEEDBACK CENTER at
> > > http://www.amibroker.com/feedback/ <http://www.amibroker.com/feedback/>
> > > (submissions sent via other channels won't be considered)
> > >
> > > For NEW RELEASE ANNOUNCEMENTS and other news always check DEVLOG:
> > > http://www.amibroker.com/devlog/ <http://www.amibroker.com/devlog/>
> > >
> > > Yahoo! Groups Links
> > >
> >
> >
>


Reply via email to