As a company, TT should be able to take some criticism. They should pay keen attention to this thread because it is direct feedback from his potential customers! Baron has a right to protect his business but it is disappointing to see TT act in this manner. Hopefully, Andrew will return to ET and present a more thorough rebuttal than the one he posted. Frank, open discussion leads to healthy intellectual debate which only benefits all of the community.
It is not surprising. Thanks for your comments. I won't ask you to elaborate further because you might be banned if you do.
Gambit, ADL does not currently offer this but we do plan to offer it in the new platform. Right now it's looking like a 2016 deliverable. Let me know if you'd like to get an early look at it. We're always soliciting users to preview designs as early as possible to ensure that we get it right.
I am surprised you did not even mention that users can instead code against the standard API directly. Or am I assuming wrong? This is actually a perfect example. Can ADL provide the past 20 ticks or 20 1-minute bars (or bars of custom compression rate) so that the user can calculate some quick pair wise correlations on the fly by converting historical data to log returns or such?
Tell us why your organization repeatedly and steadfastly refuses to post and keep updated as a priority, on its Website, a list of brokers offering your new platform, TT, (so we can actually buy it!), while simultaneously taking the time and effort to post lengthy articles on this site on how you like to employ women and while also paying $s to advertise this "new platform" with no straightforward method of actually finding the options available to purchase it.
You can use the Time & Sales block to get tick data which you can feed into the Value Extractor and Value Bucket blocks to extract and store. And you can use the Analytics block to get live and historical bar data.
I fail to see how this works. Let's forget about the analytics block because I was referring to an in-memory data operation. Let's assume we can build up a block of historical data via incoming live data points. How can I build up a rolling window of all past data points over the past 70 minutes grouped into 10 7-minute buckets? Each new incoming live data point would become part of the rolling window and old ticks older than 70 minutes ago would drop out. I want to do this on the fly with each incoming tick rather than having to access some database each time and it would eventually form the basic building block for a pairwise correlation analytic implementation. Could you walk us through an example how to do this? Thanks
VolPunter, Have a look at the image and the link (the final post in particular) below. In this ADL forum post is an example of what Andrew Renalds was describing when he said, “You can use the Time & Sales block to get tick data which you can feed into the Value Extractor and Value Bucket blocks to extract and store.” In this example we utilize the Time&Sales block, which provides live tick data, to populate the storage of the Open High Low and Closing values in Extractor Blocks (or buckets). We do this on a one minute basis for the current and last 4 minutes. At the end of the minute the current values are moved to the “1 bar back,” the “1 bar back” is moved to 2 back and so on. The values that are 4 back are dropped as they are no longer of interest. In this example we look at OHLC but with ADL any value can be tracked or any calculation made. Also, we use a 1 minute timeframe in this example, but the time frame can be adjusted as you see fit. If you need any further information feel free to reach out. https://forums.tradingtechnologies.com/showthread.php?1196-Time-Series-Bars-in-ADL&highlight=bars
Thanks for the prompt reply but my question was not about storing data but updating a rolling window continuously with each incoming live tick. As described I want to make available the past 10 7-minute bars at all times and obviously only the last bar will change with each incoming tick until the bar is completed (via passage of time). Upon each bar completion the first bar drops out of the window and a new (last) bar is created. Also I would be interested in knowing how performant such solution is for someone who would have to do this for 100-200 symbol subscriptions simultaneously. Thanks