Advantages of daily rebalance to weekly sim

General
1

Hi, are there advantages of a daily rebalance of a live model to a backtested simulation.
What data does a backtested simulation only “gets” on a weekly basis versus a live model that rebalances daily.

TY

Andreas

2

Hey Andreas, some thoughts on daily vs weekly rebalance:

  • Generally I think it depends on the proportion of factors in the strategy that are updated daily (price, technicals) vs quarterly (financial statements, fundamentals, etc) - a strategy with more technical/price factors would be more sensitive to daily rebalance

  • Consider in a daily vs weekly - how much is noise vs signal?

  • Earnings season (when financial statements are updated) - for my strategies, this is when the highest turnover takes place. And can be very sensitive - rebalance buys/sells can change each day during an earnings week as stocks are re-rated/re-ranked. For small/microcaps, also need to watch how fresh the data is, i.e. stalestmt rule. Data isn’t always immediately updated after an earnings release, so a new “buy” can be delayed 1-2 days.

  • Still, I do believe there is potential opportunity rebalancing (or at least checking vs screen) more often than weekly to see what positions have changed rank. As strategies are “sell-driven” by default, if a new stock(s) has ranked very high (and assuming you have the max # of positions), it will not be highlighted as a buy unless another stock has dropped to a sell. There are some threads on how to make your system buy-driven in this case, but I haven’t found any significant difference in performance over the long-term (but still tweaking :slight_smile: )

*Or make a judgement call (i.e. override the model!)

I’ve tried to automate some of these, but I try and screen daily or a few times a week during earnings season.

My $0.02.

Cheers,
Ryan

1 Like
3

Thank you, yes also tweaking!!!

What I am also interested in: Sims do not get some of the data, but daily rebalances get all of the data (for example earnings revisions?).

That difference is also my question: What are the differences between what a sim does (not capture) and what does a daily rebalance capture aditionly?

4

Andreas,

I am hesitant to try to give a list of assets that have daily data. But you ask a great question!!! For earning estimates I ran a screen (not a sim) with daily rebalance. I got what I thought was an answer based on the fact that the assets in the screen only changed once per week.

Anyway, confirm this yourself. But maybe this a way to get an answer for any particular factor. Especially, as there can be errors from the members (including me) in the forum.

FWIW.

P123 staff (including Yuval) would be better equipped to give you an exhaustive list with no errors.

Jim

5

Daily rebalances of a live strategy take into account all the information that you would get in a screen including all the data that’s been updated overnight and the latest ranks. Simulations, on the other hand, are ranked only weekly, and even a simulation that is rebalanced daily is restricted to the ranks that were available on the weekend. Rebalancing in a simulation will still take place daily if specified, according to buy rules, sell rules, and universe rules, but ranks will not change during the week.

1 Like
6

Thank you!!!

Add on questions:
Are screen Backtest capturing the daily rank or only the weekly?

Daily updates: is the following data updated daily (which then could be captured by rebalancing by hand every day):

Earnings estimates (weekly update or daily)

Earnings up and downgrades (weekly update or daily)

Insider Data (weekly update or daily)

Quarterly reports (earnings, sales etc. what they report in earnings season) (weekly update or daily)

Also, please, please push the improvement on the insider data!!!

Thank you!!!

Best Regards
Andreas

7

Screen backtests are also based on weekly ranks.

All data (but not ranks) are updated daily except for EPS revisions (CurQUpRevLastWk, NextFYDnRev4WkAgo, and others like those), which are updated weekly.

2 Likes
8

Andreas,

I think Yuval may mean all data is updated daily for live ports (except for EPS revisions).

Let me stay away from sims and mostly let Yuval address that. But ports can be updated daily with good up-to-date data.

Also you can run a weekly port on Thursdays and expect it to do about the same as a port rebalanced every Monday. There does seem be a little a day-of the-week effect with models in my experience.

So, I think what I said is probably correct: a port rebalanced every Thursday will do about as well as one rebalanced on Mondays.

The classic day-of-the week effect would suggest Fridays do best and I think that is probably true. With Tuesday rebalances doing the worst (again exactly as one would expect with the published day-of-the-week studies).

One can rebalance a live port daily but with the exception of a few factors in the buy and sell rules you cannot test this in a sim.

I have run some paper traded ports with daily rebalances to see how they did–as daily rebalance of ranks cannot be tested in a sim (but can be tested in a live port). Maybe I gave up too early but it did not seem to be worth the trouble. Obviously, as this was only what I was looking at the time it would be a pretty limited sample.

But LIVE screens and Ports work well on any day of the week unless you are using EPS revisions (as Yuval has confirmed above). Sims and screen backtests, not so much. These work with daily rebalancing only when using buy and sell rules and the number of factors that have daily data for a sim is a small number at that. Technical factors mainly, I believe but I have not studied this.

I think this supports what Yuval has said but I am happy to be corrected on any errors.

Jim

9

So to summarize: daily manual rebalance might capture data that screens and sims do not.
But we can not really test it (other by manually rebalance for a longer time).

  1. Any Chance we can change this (optional) (e.g. Sims capturing what a manual rebalance of a live strat
    would be able to capture?)

  2. Are people rebalancing daily and what the performance experiance of it doing it?

The reason why I am asking:

With the capabilities we have we are able to capture alpha and beta based on our weekly rankings.

But I am pretty sure there is alpha left:

If a stock has a great value momentum sentiment ranking already.
Now another data set (mainly earnings, analyst buy or sell ratings, insider data) comes out and is positive: This are important inflection points we could get as early as possible.

This would let p123 go more in the direction of “trading”, right now p123 is Trade/Investing (something in the middle). But if it is optionable people could choose…

Just my 2 cents.

Best Regards

Andreas

10

Just adding experience of my so far daily rebalance:

  • Big Earnings miss and the stock reacts negatively: Selling right away at the open or even premarket gives you a head start (10 of 10 last earnings misses I had).

  • Rankjump systems (esp. based on analyst upgrades) : very limited experience, rebalanced those the last 8 Weeks so far it was better to rebalance every day

  • Hugh Vol Spikes: (I have sell signals on huge vol spikes) depends on the market, raging bull market small caps with huge volume do well, but in all other markets high volume and small caps are doing terrible → so selling them on huge vol spikes on a daily rebalance might be beneficial…

So, thinking it through: theoretically a daily rebalance should give better performance… will do and give some feedback on it…