Reverting back to the Legacy engine for the weekend and next week.

Dear All,

Looks like there’s a few too many issues with the new engine.

We’ll revert back to the legacy engine as the default for the website , and for this weekend rebalancing. You will still be able to access the new engine by selecting ‘Current’ in the menu.

Additionally , we’re getting ready to announce our new API and accompanying Python application. Should help tremendously with testing the different engines.

Sorry about this and thank you for your patience

Just a short word to thank you all Marco and the P123 team for the hard work you are delivering with engines and data sources.
I / we appreciate this is a particularly difficult exercise.

Do not get discouraged by the various bugs popping-up. This was to be expected and you are doing a stellar jobs at solving them one by one.
Keep at it!

Thank you

Jerome

Marco,

FactSet is already working fine for me. I know any details will be worked out.

I just want to make sure you understand that you have the best platform available to someone like me and that you have the potential to take such a commanding lead that no one will ever remember a name like Quantopian in the future.

I think you will find that most people who have any experience in machine learning have most of their experience in cross-sectional data. This is something that P123 is already good at with its present models and methods. I think there is a whole world out there ready to use their old college projects with new (P123) factors. This includes people with finance degrees who have econometric models ready to go.

Quantopian does not seem to understand machine learning for Cross-Sectional data (i.e., the factors and functions P123 uses). Quantopian is a fine platform for automated day trading (time series data). P123 was never for day traders and does not need to be.

I do not understand marketing nor how (or why) ideas catch on in the market place. But with a little luck in this area this should be just the beginning for P123.

I know you have some better ideas that I have not thought of. But if I were you, I would provide a neural-net with default settings at even the trail membership level (computer resources permitting). Or something like a Ridge Regression if computer resources are limiting. This is just a marketing idea (FWIW). I do not need this myself as I have my own code that I will use at some point (financial resources permitting).

Bottom line: I do not understand how anyone could not notice. Your market is not limited to present members who may be here for the screening and financial analysis (or other great P123 features).

Best,

Jim

Compustat legacy appears to me to be differing somewhat from a few weeks ago. Check out DbtTotQ for ticker UGI. I’ve seem a number of 0s for DbtTotQ where a few weeks ago compustat had large debt numbers and factset currently has those large numbers.
Best,
Scrooge

Avner,
Thanks for pointing out this issue. The change that caused this was introduced back in June 2019. Because it only affected stocks for the duration in which the new period was indicated in Compustat as filed but none of the data was yet added to Compustat, it has gone unnoticed since then. We’ll be reloading all servers today to correct this behavior.

We are still working on this. Sorry for the delay. We’ll try to have it up later tonight. Expect the Current server to return by 7PM.

Can someone refresh my memory on the difference between Legacy and Current?

Here, again, is the list of differences between the Legacy engine and the current version of Compustat using prelims.

Shares now returns NA when “Shares Outstanding” is unavailable instead of returning a value from SharesBasic or SharesFD.

Skip preliminary fallback if numerator is present and denominator is invalid (0 or negative) in preliminary: DbtS2NIQ, DbtS2NITTM, GMgn%, NPMgn%, PMgn5YCGr%, TxRate%.

AnnounceDaysPYQ, AnnounceDaysQ, WeeksIntoQ, WeeksToQ, WeeksToY use the date of the preliminary release rather than the “effective” date.

Div%ChgA, Div3YCGr%, and Div5YCGr% are -100 instead of NA or 0 when going from positive to zero.

Preliminary fallbacks have been revised for these factors: EV2EBITDA, ValROETTM, Pr2BookPQ.

Hi all.

Just a question about the new engine and the new data-base and all that stuff.

The “Factor & Function Reference”, will change or remain the same??

I mean, the information about when the factors start working, and so…

Thanks.