Rebalancing Day of the Week - Thoughts?

[brianj999]

My gut feeling is that Monday might not be the best day to rebalance a portfolio. I suspect it's when many others are also rebalancing after running their models over the weekend, potentially leading to bidding up/down the same stocks. I'd love to hear other perspectives on this.

If my memory serves correctly, P123 uses end-of-week data for backtesting and simulations regardless of the chosen rebalancing day. If this is still true, what would be an efficient way to test the impact of the rebalancing day?

Happy New Year, everyone!

[Jrinne]

Actively traded for a while but paper-traded at the close now IN ORDER TO ANSWER THIS VERY QUESTION, So traded for over a year with P123 editing the price (therefore little human error).

As far as Monday being a particularly bad day, I am just not seeing it with my personal trading:

Jim

[pfrommert]

just rebalance every day

[Whycliffes]

Thank you for the overview; it is very helpful. It is unfortunate that we do not have the ability to investigate the best days for trading in the simulations.

Do you have any explanation or hypothesis regarding why Wednesday performs significantly worse?

I am not aware of how much your strategies overlap with those used by other P123 users, but there has been a theory that many of us trade micro and small-cap strategies, and with numerous similar stock criteria and a limited universe, we impact the prices on Mondays.

Regarding the strategy you trade on Mondays, what type of strategy is it, and will it face challenges with low volume if many P123 members are trading the same stock?
How many trades or what level of turnover do you experience with your strategies? I am curious to know the statistical basis behind your performance table.

[Jrinne]

@Whycliffes

Do you have any explanation or hypothesis regarding why Wednesday performs significantly worse?

I don’t have a clear explanation for the Wednesday outlier.

@Whycliffes

And will it face challenges with low volume if many P123 members are trading the same stock?

It’s a slightly more liquid version of the Easy To Trade Universe, if that helps answer your question.

@Whycliffes

How many trades or what level of turnover do you experience with your strategies? I am curious to know the statistical basis behind your performance table.

Here’s the turnover data for all five days. Each portfolio holds 15 stocks, and I started the first portfolio on 12/1/23. Also note the number of trades for each port is about 150 trades in the statistics:

@Whycliffes

There has been a theory that many of us trade micro and small-cap strategies, and with numerous similar stock criteria and a limited universe, we impact the prices on Mondays.

I think that’s a great theory, and it would be fascinating if P123 members collectively influenced small-cap markets globally. However, my small study doesn’t definitively support this hypothesis, so I looked for additional data.

I couldn’t get Claude to program code to download and analyze volumes from the Russell 2000 sorted by day, so I had ChatGPT search the topic. This study seemed relevant: Monday Effect: Trading Patterns of Individual Investors.

“This study involved using actual individual investor data from the New York Stock Exchange from May 2006 to April 2016… study results demonstrated day of the week was not a factor in terms of average daily returns, and individual investor daily trade percentages on Monday did not significantly differ from those on Tuesday, Wednesday, and Thursday. However, the percentage of individual investor trades on Fridays was significantly lower than individual investor trades on other weekdays.”

At this time, I don’t have data to support the idea that P123 members are significantly impacting markets, but it’s an intriguing hypothesis worth exploring further.

My study and the referenced paper don’t address nano-caps, so there may still be unexplored effects in those smaller universes.

Jim

[Jrinne]

@Whycliffes

Curious to know the statistical basis behind your performance table.

Great question. Let me share my thoughts—or, hopefully, they rise to the level of being called a "statistical analysis of some sort."

Caveat:

I doubt that my study has much statistical power. This means there could definitely be a difference between the days, but I simply don’t have enough data to detect it. This limitation applies no matter how you look at it.

I might run this through JASP as a statistical exercise later, but Claude 3’s approach using a Repeated Measures ANOVAseems appropriate. (For reference, this is similar to a paired t-test but works with more than two groups—in this case, five days.)

Here’s Claude 3’s analysis for context:

—

Claude 3’s Repeated Measures ANOVA Results

1. Overall ANOVA Results:

• F-statistic: 0.411 (df = 4, 224)
• Effect size (Partial η²): 0.0073 (indicating only 0.73% of the variance is explained by the day of rebalancing)
• Interpretation: The very low F-statistic suggests no significant differences between the days.

2. Pairwise Comparisons:
   The largest differences were observed between:

• Wednesday vs Thursday: -0.2033% (t = -1.7831)
• Wednesday vs Friday: -0.1798% (t = -1.0843)
• Monday vs Wednesday: 0.1507% (t = 0.8358)However, none of these differences were statistically significant at conventional levels (p < 0.05).

—

Practical Implications

1. Timing of Rebalancing: The day of the week (Monday through Friday) does not appear to significantly impact the excess returns of this strategy.
2. Magnitude of Differences: The largest observed difference was only about 0.20%, which is small.
3. Effect Size: The effect size (Partial η² = 0.0073) suggests that other factors are likely far more important than the day of rebalancing in determining returns.

This result is actually reassuring. It suggests:

• You have flexibility in choosing your rebalancing day without a significant impact on performance.
• Your strategy appears robust to the timing of rebalancing.
• Practical considerations—like trading costs, liquidity, or operational convenience—can take precedence when deciding on a rebalancing day.

[Note: Remember the low statistical power of this study when interpreting these results.]

—

@Whycliffes

Do you have any explanation or hypothesis regarding why Wednesday performs significantly worse?

While I don’t have a clear explanation, it’s important to emphasize that the difference observed for Wednesday was not statistically significant and could very well be due to chance.

So, I think the realistic conclusion is that you cannot reject the null hypothesis with this weakly powered study. Not rejecting the null hypothesis in this context does not exclude the possibility of a day-of-the-week effect—especially for nano-caps, which were not included in my analysis.

Jim

[Ju]

Hello Jim,

Was it done with running the model on Mondays only, or were you re-running it for each of the days (I understand that it would change any ranking based on price but not any other)?

Same strategies any day? With which % of volume traded at worst? With VWAP or any other order? Basically trying to assess if orders spread on the week are affecting other days performance due to market impact.

Best regards.

Julien

[Jrinne]

Monday was rebalanced on Monday each week and only Monday. Once per week rebalance, like many Designer Models.

Tuesday was rebalanced on Tuesday each week and only Tuesday.

Wednesday, Thursday, and Friday followed the same pattern.

Prices are automatically adjusted by P123 to reflect the closing price (reducing the chance of data-entry errors). I do have to take about 15 seconds each day to click “rebalance” to collect this data.

—

Symmetry and My Assumptions

I have a bit of a physics background, and much of modern physics relies on symmetry assumptions. For me, this study boils down to two alternate views:

1. Symmetry: The days are essentially the same, with no meaningful differences in performance.
2. Broken Symmetry: There’s something unique about certain days—perhaps a bunch of people trading Designer Models on Mondays or activity from platforms like Zacks influencing specific days.

To me, it’s a bit like the breaking of matter-antimatter symmetry in the early universe, which led to our very existence.

Okay, maybe that’s a stretch—but it illustrates the idea that while symmetry might be the natural assumption, symmetry can and does break, sometimes without any messy human interactions.

—

My Inclination

I tend to lean toward assuming symmetry unless there’s evidence to suggest otherwise. However, I find the idea of P123 members (and maybe a few people at Zacks) collectively moving the market on certain days intriguing.

That said, my study doesn’t provide enough data to definitively support or reject either hypothesis.

—

Bayesian Perspective

In a Bayesian sense, this new data doesn’t shift my prior belief much. There’s simply not enough data to make a strong case one way or the other. It’s the only data I have, but I’m very open to seeing additional studies or perspectives.

—

I hope this answers your question and provides some context for my approach and thoughts.

Jim

[korr123]

I had the same suspicion about Mondays but my live experience is that day of week didn't matter.

I now focus on layering signals. i.e. diversify across time as much as possible to reduce noise.