I am thinking of turning the daily seasonal alpha into production...

[nikolas-joyce]

Dmitry has done this interesting seasonal algo that uses daily futures bar data to extract seasonal patterns and builds average commodity performance.

https://10.0.1.2:8888/notebooks/strategies/Strategy_Seasonality_DailySeasTracking_StrategyModule_new.ipynb

Because our data does not go back this far and due to the fact that our new data infrastructure is not complete Dmitry has used Quandl to access the history and build the data set.

My question is whether this is an issue in your mind (should we only use data that we have contained in our own ecosystem). Or is this a non issue and we could put it into beta production where i could watch this type of strategy in our realtime while we develop our data infrastructure more. thx

[alexveden]

The only issue that I can see is a eod price synchronization, so we need to use intraday data or previous eod bar for decision making

Отправлено из Mail.Ru для Android четверг, 09 февраля 2017г., 22:44 +04:00 от nikolas-joyce notifications@github.com :

Dmitry has done this interesting seasonal algo that uses daily futures bar data to extract seasonal patterns and builds average commodity performance. https://10.0.1.2:8888/notebooks/strategies/Strategy_Seasonality_DailySeasTracking_StrategyModule_new.ipynb Because our data does not go back this far and due to the fact that our new data infrastructure is not complete Dmitry has used Quandl to access the history and build the data set. My question is whether this is an issue in your mind (should we only use data that we have contained in our own ecosystem). Or is this a non issue and we could put it into beta production where i could watch this type of strategy in our realtime while we develop our data infrastructure more. thx — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub , or mute the thread .

[nikolas-joyce]

Dmitry is the eod price sync an issue? You had done a sanity check on this method already?

[dmveden]

You had done a sanity check on this method already?

Yes, it passed the test.

is the eod price sync an issue?

I did not understand what price sync means in this context.. Price data from quandl and exo series does not have to be same length or synchronized by date, because the seasonal average series, which this algo tracks, have a not standard index

[nikolas-joyce]

Can you expand on the not standard index in an example svp. I am dumb.

thx

Nikolas Joyce najoyce@tml1.commailto:najoyce@tml1.com mobile : 604 616 0489 office : 604 984 7244

---

From: dmveden notifications@github.com Sent: February 9, 2017 12:00 PM To: trendmanagement/tmqrexo_alexveden Cc: nikolas-joyce; Author Subject: Re: [trendmanagement/tmqrexo_alexveden] I am thinking of turning the daily seasonal alpha into production... (#107)

You had done a sanity check on this method already?

Yes, it passed the test.

is the eod price sync an issue? I did not understand what price sync means in this context.. Price data from quandl and exo series does not have to be same length or synchronized by date, because the seasonal average series, which this algo tracks, have a not standard index

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://github.com/trendmanagement/tmqrexo_alexveden/issues/107#issuecomment-278755467, or mute the threadhttps://github.com/notifications/unsubscribe-auth/ARobLP1nlfAibzKJCJVyRBm-CI8S-htdks5ra3B5gaJpZM4L8gyW.

[dmveden]

dataframe like this is used for decision making. Algorithm goes through index array of an exo, and if day and month of current index is equal to day and month of dataframe above it sets a trigger value from 'long_signal' or 'short_signal'.

For example. We want a bullish alpha. Current index within a loop is equals to 1960-01-11. So for this index - day = 11, month = 1. The corresponding value of 'long_signal' column in dataframe above on index day=11, month=1 is True. So trigger for this alpha on index 1960-01-11 is set to True, and alpha will go long

On next index (1960-01-12) trigger will be False

[alexveden]

I did not understand what price sync means in this context..

If you don't use last day in calculations this should work fine.

The nature of this issue connected with time of the close of the day, for example Quandl quote could have Close price marked at 16pm, but our quotes are calculated at 12pm at the decision time. This could be a problem in real-time execution, but if you use previous day in calculation this should be fine.

[nikolas-joyce]

So there is no urgency to remove the Quandl reference from the code? It would be suitable for production in the current state Alex?

thx

Nikolas Joyce najoyce@tml1.commailto:najoyce@tml1.com mobile : 604 616 0489 office : 604 984 7244

---

From: alexveden notifications@github.com Sent: February 9, 2017 12:26 PM To: trendmanagement/tmqrexo_alexveden Cc: nikolas-joyce; Author Subject: Re: [trendmanagement/tmqrexo_alexveden] I am thinking of turning the daily seasonal alpha into production... (#107)

I did not understand what price sync means in this context.. If you don't use last day in calculations this should work fine.

The nature of this issue connected with time of the close of the day, for example Quandl quote could have Close price marked at 16pm, but our quotes are calculated at 12pm at the decision time. This could be a problem in real-time execution, but if you use previous day in calculation this should be fine.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://github.com/trendmanagement/tmqrexo_alexveden/issues/107#issuecomment-278762893, or mute the threadhttps://github.com/notifications/unsubscribe-auth/ARobLJKbLhhaTvs_lCRbkqXyTs4WWxTvks5ra3aOgaJpZM4L8gyW.

[alexveden]

Seems that @dmveden using quandl data until 2009, so this won't be the issues I've expected. solved.