Market Calendar Edge: Expiry, Rebalance, Buybacks
What this does
Markets do have rhythm. It comes from deadlines — someone is obliged to buy or sell a certain amount on a certain date. Where that obligation exists, the rhythm is real. Where it does not, the pattern is usually an artifact of testing too many hypotheses.
Two jobs: find the rhythm that is real, and strip calendar noise out of flow analysis so a one-week spike is not mistaken for a trend.
The governing test
If you cannot state the causal mechanism in one sentence, do not trade the rhythm.
"On date X, participant Y is required to do Z." If you can say that, it is real. If you cannot, it is a coincidence found by searching enough data.
Sample output
📅 Supply-demand calendar — week of M/D
■ Institutional events
M/D Monthly options expiry → positioning unwinds after; volume spikes
M/D Investor-type flow data, 15:30 → information event, not a flow event
M/D Last business day of the month → the index is +6.2% MTD, so pension
rebalancing implies mechanical SELLING into month-end
■ Earnings
M/D cluster of 240 companies → attention fragments, index likely muted
■ Net read
Month-end supply meets a thin post-expiry tape. Expect drift rather than
direction until the first week of next month brings new contributions.
■ Exclusions for flow analysis
M/D volume increase is expiry-driven. Exclude from trend judgement.
What gets mapped
Within the month — options and futures expiry; month-end rebalancing (pressure runs opposite to the month's move); start-of-month contributions; scheduled data releases.
Within the year — fiscal year-end mechanics, ex-dividend dates, earnings seasons, tax-loss selling, index reconstitutions, thin summer liquidity.
Under-watched, and this is where the value is:
- Buyback blackout windows. Companies suspend repurchases for roughly four to five weeks before earnings. In the US, corporate buybacks are the largest single source of equity demand. When that buyer steps away, supply-demand loosens structurally. Declines run deeper during blackout and buying resumes shortly after.
- Dividend reinvestment timing. Dividends declared at period-end are paid months later and reinvested then — a delayed buying impulse disconnected from the ex-date everyone watches.
- Earnings density. How many companies report on a day matters, not just who. Dense days fragment attention; sparse days concentrate it.
- Month-end interacted with performance. A large monthly gain plus month-end rebalancing produces mechanical selling, because equity weights have drifted above target.
Testing anomalies
Five checks before relying on any "X effect": stateable mechanism, out-of-sample survival, survives transaction costs, how many hypotheses were tested, and consistency across markets.
Worked example — the lunar cycle. One widely cited study across 48 countries found returns lower around full moons. A later study tested 59 markets: full moon effects were significant in 6, new moon effects in 8. At a 5% threshold, testing 59 markets produces about 3 "significant" results even with no effect at all. And the results had no pattern — positive new moon effects appeared in markets sharing no geography, investor base, or structure. A real effect shows up consistently; scattered results are the signature of data mining.
Why it still feels real: a lunar month is about 29.53 days, a calendar month averages 30.44. They drift in and out of alignment with month-end rebalancing and monthly expiry, both of which are real. It is the calendar underneath, not the sky.
FAQ
How often should this run?
Weekly, at the start of the week. Also any time flow data shows a spike you are unsure about.
Does it work outside one market?
Yes. The institutional deadlines differ by market but the framework is the same. Name the market and it maps that calendar.
Does it recommend trades?
No. It describes structural supply and demand and its timing. Calendar effects are tendencies with wide variance, and they are stated with that uncertainty.
