Best place to purchase historical stock data for backtesting??

Quote from mgabriel01:

Only 5 deep on Depth of Market
Could be deeper, but that hasn't impacted anything so far
More...


Based on this link I found on Google:

http://www.ninjatrader.com/support/forum/showthread.php?t=47387

and this link provided by Kyle @ Ninjatrader customer service:

http://www.ninjatrader.com/support/helpGuides/nt7/index.html?importing.htm


[Open Quote
Tick Format
Each tick must be on its own line and fields must be separated by semicolon (;).

The format is:
yyyyMMdd HHmmss;price;volume

Sample data:
20061107 000431;1383.00;1
20061107 000456;1383.25;25
20061107 000456;1383.25;36
20061107 000537;1383.25;14
End Quote]



I don't see any sub-second time stamp ability from their API. I can promise you that there is zero chance that five levels deep of market data on ES amounts to 20MB per day unless it is cut down substantially.

For example, Bloomberg data for SPY pulled via their API ranges from 56MB to 86MB per day (over the last 5 years) and that's stripped down - quotes and prints only. No cancels, no broker codes, no partial/full designation, no crosses, no order type, no crosses, etc.

If you are at 20mb per day and you only have market depth five levels deep you are probably not getting every exchange and are certainly not getting the full trade details with every message.
 
Quote from bwolinsky:

Agreed, minute and daily data is useless.
More...

I disagree with this. I use minute based data for intraday backtesting and daily data for swing based testing. As long as you know and can adjust your code to avoid the weakness of the data precision, you should be ok. For the average trader, tick data actually has many disadvantages.

Quote from mgabriel01:

intraday tick
unfiltered
More...

One caveat with unfiltered tick data is a bad tick on one feed may not be present on another feed. Thus, comparing data from multiple, unfiltered vendors will yield different test results. Although this is true will all brokers and feeds, many fail to recognize this fact.

TickData cleans their data. They even have a white paper on the subject "High Frequency Data Filtering - Tick Data".
 
Quote from AK Forty Seven:

I have and a lot of their data is inaccurate.I would not buy from them again
More...

How so?

I've been using pi's data for backtesting and haven't had any problems with it. I've compared it against IB's historical data, and they seem to match up.

Also, to those of you suspicious of how cheap it is: how much do you think it costs them to burn you a DVD and send it to you? Some places charge a shitload of money because they know that a lot of people correlate cost with quality, and want to help those people get rid of that extra cash. Other places just want to attract a lot of customers. In the end, I doubt it costs any of them that much to subscribe to a data source, save it to a disk, and then burn it to a DVD for subscribers.
 
Quote from sneakoner:

I want to focus on Nasdaq stocks with a volume over 300,000 and price over $10. A quick scan results in 500+ stocks. And its still $8400 for 500 equities :(
More...
===============
Mr.Sne;
Another way to reach your Nasdaqqq goal;
divide the market into bull markets/bear markets/200 day moving average. Then Test. You may want to buy some books recommended [top green line this website/page], or spend$ $8,400+ - but that just the first step in a lengthy process.

And get your hands dirty with the data[hourly, 4/5 hours,40 or 50 days...;
sure its exspensive in time, but it stays in you mind so much better.There are no shortcuts, but differences in bull & bear is much.
Some who are visual like charts[picture=2,000 words];
my old CPA doesnt like charts, he likes numbers.............:cool:
 
Another reason i prefer the time exspense of printed/paper charts;
several different data companys ''forgot'' to mention Citigroup split 10 times reverse, its really a $3.25 penny stock.LOL:D Big BIG difference.......................................

MSFT split once or more...............................................;
i noted that by hand on my charts-amazing the mistakes that ''hide'' or are simply concealed in lots of data.:cool:
 
Quote from dyson:

Is this site a joke? Who would pay for this?
More...

Yeah, it was way too expensive. $34.95 per symbol? I don't need that when all I have to do is click edit data.
 
Quote from Mike805:

TradeStation also has good data.
More...

+++

$99/mo if you open an account. First three months are free.

And the backtesting engine is the quickest / easiest I've tried.

And it will paper trade your algos when you're ready to forward test.

And it can walk forward optimize on a portfolio....

:D
 
Quote from WinstonTJ:

Based on this link I found on Google:

http://www.ninjatrader.com/support/forum/showthread.php?t=47387

and this link provided by Kyle @ Ninjatrader customer service:

http://www.ninjatrader.com/support/helpGuides/nt7/index.html?importing.htm


[Open Quote
Tick Format
Each tick must be on its own line and fields must be separated by semicolon (;).






The format is:
yyyyMMdd HHmmss;price;volume

Sample data:
20061107 000431;1383.00;1
20061107 000456;1383.25;25
20061107 000456;1383.25;36
20061107 000537;1383.25;14
End Quote]



I don't see any sub-second time stamp ability from their API. I can promise you that there is zero chance that five levels deep of market data on ES amounts to 20MB per day unless it is cut down substantially.

For example, Bloomberg data for SPY pulled via their API ranges from 56MB to 86MB per day (over the last 5 years) and that's stripped down - quotes and prints only. No cancels, no broker codes, no partial/full designation, no crosses, no order type, no crosses, etc.

If you are at 20mb per day and you only have market depth five levels deep you are probably not getting every exchange and are certainly not getting the full trade details with every message.
More...



Interesting!
I'll bring it up with them and see what they say
Thanks!
 
You must log in or register to reply here.
Back
Top