Semiannual estimates

Until tonight, we weren’t loading semiannual estimates. We have now fixed this. About 140 semiannual stocks will now have semiannual estimates.

Please note that for stocks with semiannual estimates, every other interim estimate and actual is blank: HistQ2, HistQ4, and NextQ factors as well as odd offsets for the ConsEst functions will all give NA for semiannual data. Also note that there’s a 6 month gap instead of a 3 month gap between HistQ1 and CurQ, but it’s a compromise that needed to be made for usability.

We’ve created a new factor: ConsMonths. This will be 3 if the estimates are quarterly, 6 if they’re semiannual, and NA if there are no estimates.

For these semiannual stocks, does this “Q” (as in CurQ) actually cover six months?

Yes, that’s right.

Yuval, I do not know what happened, but the change might have hit my backtesting in a great way, see pictures below.



Pleeeeeaaaaaaase investigate, Thank you!!!

deleted

All my modells are heavily hit by this (or another?) Change!

See old similation and new simulation.

1 or 2 missed rebalances the next 1-2 weeks would not make a big difference so I am save for now.
But it would be really nice if we could get that fixed, otherwise I am dead in the water because my model rely heavily on earnings estsimate.
Thank you!!!



This seems unrelated to the change we made. The change affects only quarterly estimates, and you’re not using quarterly estimates in your systems.

In general, live strategies will differ markedly from simulations because live strategies used Compustat data until a year ago and simulations use only FactSet data. This is especially true if the strategies have very few holdings. Moreover, FactSet revises some of its data from time to time, so simulations run today will be different from simulations run six months ago.

Were your live strategies converted from simulations recently or were they converted from simulations a long time ago? If they were converted very recently, then there may be some cause for concern and we could dig a little deeper. But if they were converted a while ago, this behavior is entirely expected.

Both strategies are based on backtest that I have finished developed around the 18th of may this year, then they have been converted to Live Strategies.

Also I have done simulated backtests from the last week (!) that replicated both strategies almost 1:1 (because I am tracking simulation regulary versus the live ports to see if I run in issues). There where some minor differences, but realy realy small (1-2% range).

Thank you!!!

Have a look at https://www.portfolio123.com/port_summary.jsp?portid=1655621

it hase been simulated on 08/07/21 and it almost matches the strategy above (first picture): https://www.portfolio123.com/port_summary.jsp?portid=1647390

Sorry for having hijacked this thread, but I think I have a good hint!

I should not be able to change the ranking system (because it is used by a designer model). But it seems to have been changed in the process of copying it, meddling with it.

But it got changed, so the value part of the ranking system got set to 0.

now I need to find the original settings in the ranking system and try the backtests again.

I let you know!!! (gee, if this is it I am really sorry, beginners mistake after using P123 for 10 Years, arrrrrggggg…)

o.k. false alarm!!!
that is the cause, I am really sorry!!! Friday evening overworked and I see white mouses around me, lol!

Now I have to find the original weightings of the ranking system and I am good to go!

Please close the support ticket I send you via email and thank you 100000000 times for your help!
Once again thank you!!!

deleted