Screener: Is it supposed to be historically accurate?

I wish to inquire what using the “As Of” date is supposed to return in the screener. When using a previous date, is the screener supposed to return the stocks that would have passed on that date? Are there any known reasons for possible discrepancies? My impression is that the weekly ranking updates seem to influence the stocks chosen by the screen so that the stocks returned later using the As Of feature for a selected date are not necessarily the stocks that would have been selected if the screen was run on the date itself. Is there documentation for this? Thank you.

As far as I know, yes it is. P123 always say that their data is supposed to be “point in time”.

The only issues I’m aware of are that
(1) the screener will use the data from the previous weekend update
(2) there have been a few bug fixes in the past which might have changed results

Why don’t you take a screenshot of your results and then post any discrepancy? I’m sure they’d look into it.

The P123 historic fundamental data was saved on the weekend, and the stock prices (Open, Hi, Low, & Close) and daily volume were saved daily over night. So an “as of date” screen that has fundamental rules (or uses a ranking system that has fundamental factors) will use the last weekend’s data, and any screen that uses price or volume rules will use the previous day’s data.

P123 downloads fundamental data daily as it is available from Compustat (but only saves it for historic data over the weekend). Therefor, a screen that is run during the current week will use yesterday’s new fundamental data (if there is any), + prices, & volume data.

PS: be aware that Compustat didn’t download the open price prior to mid 2004. Because of that P123 calculates an open price bases on previous close and the day’s H & Low. You can see Marco’s explanation HERE .

Thank you for your responses. I think I’ll take some screen shots and see if in a week they are consistent with the results I expect. I have this strange feeling that the results I got for a certain day last week aren’t the same as the one I’m getting now using the As Of feature for that date but I’m second guessing myself thinking maybe I was just doing something differently or misunderstanding something and wanted to ask first if their are relevant issues the community is aware of. I’ll follow up if I see any discrepancies.

Denny, I’m a little unclear on what you are saying. Are you saying that the results for the screen that I ran for example last February 11 should match the results running the same screen today with As Of set to February 11 (or maybe Feb. 12 taking time of day differences into account), or should I expect them to be slightly different? Are As Of screens point-in-time? Ranking data for a screen run February 11 would use February 7 fundamentals. If I ran the screen today using February 11 as an As Of date it would still use February 7 ranking data right?

Sterling,
The way I understand it and what I think Denny was saying is if you ran the screen on the 11th it had the most up to date data available: updated that morning.

Running the “as of” for the 11th today will be different than when you ran it the 11th. Run today, the rank data will be from the 7th-or sometime that weekend. As Denny said, some of the price data in the buy rules will be more up to date. Screens and ports have the most up to date data available (some data is only updated weekly); however, for backtests and sims the ranks are weekly only.

So when you say: “I have this strange feeling that the results I got for a certain day last week aren’t the same as the one I’m getting now using the As Of feature for that date,” you are probably correct.

So what you are saying Jrinne is that the rank data may be point in time (both using rank data from the 7th in my example) but the price data would have since been updated and that may cause differences? So there is no way to faithfully reproduce with confidence the screen results from more than a week ago?

Sterling,

Only the weekend data (Monday morning) is available for backtesting purposes and for “as of date”. Therefore you should be able to reproduce all Monday morning data.

Chaim

Sterling,

If you ran a screen on a Thursday last week it would use fundamental and price/volume data downloaded from the previous Saturday through Wednesday. However, if you run the screen on any day this week with an as of date of last Thursday, you will be using fundamental data from the week ago weekend.

Any fundamental data downloaded during last week’s week days will have been saved with an as of date of this last weekend. The fundamental data downloaded on Monday, Tuesday, or Wednesday nights that was available for a screen run last Thursday is no longer available for a rerun of a screen with an as of date of last Thursday. The price/volume data downloaded on Monday, Tuesday, and Wednesday nights is saved with those as of dates.

Thanks for the clarification. Does this mean that simulations work the same way?