Sneak Peak

Synergy data mining application developed by AdaptiveTradingSystems.com
jamess
Site Admin
Posts: 55
Joined: Wed Jan 04, 2017 9:52 pm

Sneak Peak

Post by jamess » Wed Jan 04, 2017 10:56 pm

Previews of new functionality yet to be released appear under this topic.

jamess
Site Admin
Posts: 55
Joined: Wed Jan 04, 2017 9:52 pm

Re: Sneak Peak

Post by jamess » Sun Feb 19, 2017 4:47 am

The Pinnacle futures data includes volume and open interest columns. However, the futures volume and open interest are always one day behind. When doing exploratory modeling runs using multiple Pinnacle futures data series, deleting all of the volume and open interest columns was time consuming and error prone. In Synergy 2.6.8 you can select one instance of the volume and open interest columns, right click and select Remove Columns of Type to remove all volume and open interest columns from the loaded data. Makes life easier.

Kind Regards,

James
Attachments
loaded_data_remove_columns_of_type.PNG
loaded_data_remove_columns_of_type.PNG (10.44 KiB) Viewed 5494 times

CarlosR
Posts: 51
Joined: Sun Jan 08, 2017 10:52 pm

Re: Sneak Peak

Post by CarlosR » Sun Feb 19, 2017 2:26 pm

Nice!

Now if we could just get Pinnacle to fix their data. (or at least the volume part -- I think the OI part is up to the exchange)

jamess
Site Admin
Posts: 55
Joined: Wed Jan 04, 2017 9:52 pm

Re: Sneak Peak

Post by jamess » Mon Feb 20, 2017 12:37 am

Hi Carlos,

You are right the open interest being one day behind is due to the CME. I was browsing the free futures data on Qandl (https://www.quandl.com/) yesterday and noticed that the volume data appeared to be up to date and the open interest data was always 1 day behind. One of the free futures data providers, I think it was the Chris database, includes continuously linked contracts for the nearest, 1 month out, 2 months out, etc. contracts. That could be very useful. Quandl has an API so I'm thinking I could write a C# application to manage downloads from Quandl. Haven't looked into that yet.

In Synergy 2.6.9 you can remove loaded data series from the Data Chart. That way when you identify a 'bad' data series you do not need to close the chart, remove the series from the Loaded Data list and then open the chart again.

Kind Regards,

James

CarlosR
Posts: 51
Joined: Sun Jan 08, 2017 10:52 pm

Re: Sneak Peak

Post by CarlosR » Mon Feb 20, 2017 4:28 am

James,

I just sent a second email to Ed Pavia at Pinnacle asking him to *please* get the volume data on the correct date line. Maybe if you and some of the Synergy users were to do the same, it might help.

His email is pinnacle@netacc.net

Short of that, I'm all for trying out other sources of data. I have a feeling that we may be leaving some good predictive power on the table, by not having access to the volume data.

CarlosR
Posts: 51
Joined: Sun Jan 08, 2017 10:52 pm

Re: Sneak Peak

Post by CarlosR » Tue Feb 21, 2017 3:28 pm

James,

I just uploaded 2.6.9, and the addition of the description of the formulas in the User Guide is fantastic. It must have been a bunch of work, I'm sure, but it really helps one understand what is going on inside the models.

Thanks much for doing this.

jamess
Site Admin
Posts: 55
Joined: Wed Jan 04, 2017 9:52 pm

Re: Sneak Peak

Post by jamess » Tue Feb 21, 2017 11:35 pm

You are welcome Carlos. Thanks for the feedback.

Kind Regards,

James

jamess
Site Admin
Posts: 55
Joined: Wed Jan 04, 2017 9:52 pm

Re: Sneak Peak

Post by jamess » Mon May 08, 2017 6:23 am

Version 3.0.1 of Synergy will be available very soon and includes the Data Snoop for quickly identifying which data series are useful for modeling a given traded series.
zt_idx.PNG
zt_idx.PNG (37.01 KiB) Viewed 5061 times
The above image lists the data evaluation results for the ZT futures contract trading on the close with the trading delay set to 1. The $LSTEER series has the highest rating and a large number of samples. I had previously identified the cash price for live steer as the best data series for modeling live cattle futures. This provides some confirmation that the Data Snoop is working as intended.

Kind Regards,

James

CarlosR
Posts: 51
Joined: Sun Jan 08, 2017 10:52 pm

Re: Sneak Peak

Post by CarlosR » Wed May 10, 2017 12:53 am

Hi James, this version looks very interesting indeed! I'm having fun playing with the Data Snoop function, which looks like it could be very useful.

One thing I did not find in the User's Guide is a description of the "denoise" process? Can you explain what that does, and why you might or might not want to use it? (if it's in the Guide and I missed it, my apologies)

Thanks much.

jamess
Site Admin
Posts: 55
Joined: Wed Jan 04, 2017 9:52 pm

Re: Sneak Peak

Post by jamess » Wed May 10, 2017 4:13 am

Hi Carlos,

I haven't added a description in the user guide for the denoising yet. Will do it now. It's very simple,

y(t) = 0.25 * x(t-1) + 0.5 * x(t) + 0.25 * x(t+1)

In words, 0.25 times the value at the previous row plus 0.5 times the value at the current row plus 0.25 times the value at the row after the current row. The denoised traded series is only used for measuring the fitness during the particle swarm optimization process, so there is no future leak. This filter dampens, but doesn't entirely remove, the very high frequency fluctuations in price. I suspect that it would be most useful when combined with a linear model or similar.

The Data Snoop is a game changer in terms of productivity. I could not be happier with the results I've seen so far.

Kind Regards,

James

Post Reply