Moon Phases Trading Strategy: Research, Rules, and Backtesting

A moon phases trading strategy uses astronomical dates to schedule market exposure. Published research has found historical associations between lunar phases and stock returns, but that finding does not establish that buying each new moon and selling each full moon produces a tradable edge. The useful starting point is to separate the research question from the trading rule, then test the rule on data that did not determine its design.

What the Original Research Found

Kathy Yuan, Lu Zheng, and Qiaoqiao Zhu's 2006 paper, Are investors moonstruck? Lunar phases and stock returns, studied stock indices in 48 countries. It reported lower returns around full moons than around new moons, with an annualized difference of 3%–5% for its global portfolios. That is a historical return comparison, not a published net return for an individual investor following the rule below. The paper also explicitly acknowledges that a detected relationship could be spurious.

The authors found that differences in volatility and trading volume did not explain the return pattern. Their result therefore does not support the claim that full moons necessarily produce higher volatility or require wider trading stops. The published paper is available through the Australian National University repository.

A later study, Are investors moonstruck? Further international evidence on lunar phases and stock returns, reported a new-moon effect but no overall full-moon effect in its analysis. These results make the exact market, sample, and window definition material; they do not justify treating all lunar rules as equivalent.

Investor mood is a proposed explanation, not a mechanism established by this article. No independent Pineify replication or recent performance series is presented here.

A Phase Window Is Different from a Holding Period

A return window centered on a new moon includes observations before and after that event. A position opened at the new moon and closed at the full moon covers a different interval. Transferring a result from one definition to the other changes the hypothesis.

Question	Required definition	What it can establish
Do returns differ around lunar phases?	Fixed event windows, return series, and overlap treatment	An association within the selected sample
Does a new-moon-to-full-moon position earn money?	Entry, exit, tradable asset, costs, and cash treatment	Simulated performance of that specific rule
Does a lunar filter improve another strategy?	Identical baseline with and without the filter	Incremental effect on the same observations

A mean lunar month is roughly 29.53 days, but actual event timestamps should drive a test. Repeatedly adding a rounded interval can move the assigned event away from the astronomical phase.

Define One Rule Before Looking at Returns

The following is an experimental specification, not a recommended allocation:

1. Choose one tradable instrument and data provider. Record its exchange, session, time zone, dividend treatment, and data coverage.
2. Obtain new- and full-moon timestamps from the U.S. Naval Observatory's primary moon phases service. Preserve its time convention when converting events to exchange sessions.
3. Enter long at the first scheduled session open strictly after each new-moon event. If the event occurs during a session, use the next session's open under this rule.
4. Exit at the first scheduled session open strictly after the following full moon. Remain in cash until the next entry.
5. State how cash earns interest, how distributions enter returns, and how fees and spread are charged. Use the same conventions for comparison portfolios.
6. Freeze the rule and test interval before running it. Changing the event window or entry delay creates another candidate that must be counted in the research record.

A stock index itself is not directly tradable. If using an ETF, begin with that ETF's available history. If using futures, specify the contract and roll treatment. Do not silently splice an index series into an ETF backtest.

Calculate Results That Another Reader Can Reproduce

For each completed long trade, calculate the gross return as:

Gross trade return = exit price / entry price − 1

For a cash-equity example, subtract the cash cost of entry and exit from the position's dollar profit before dividing by entry notional. This makes it explicit whether a quoted cost applies per side or per round trip. Use a daily account equity series, including time spent in cash, for portfolio statistics.

CAGR = (ending equity / starting equity)^(1 / elapsed years) − 1

Drawdown at time t = equity at t / highest equity reached through t − 1

Report the largest drawdown magnitude alongside CAGR, market exposure, completed round trips, and the full date range. Twelve months of similar calendar exposure can still contain very different numbers of trading sessions. Do not present a count of individual orders as a count of completed trades.

Keep these comparisons separate:

• Buy and hold: Shows what happened to continuously invested capital over the same dates.
• Cash: Makes the return assumption for uninvested periods visible.
• Calendar placebos: Shift the event schedule by predetermined offsets while preserving the holding rule. This tests whether the lunar label is doing more work than a recurring exposure schedule. Repeatedly searching offsets also creates multiple-testing risk.

Check Whether the Result Survives New Data

Reserve a chronological holdout before choosing parameters. If the proposed rule was motivated by an older paper, a later interval can test whether its association persisted, provided it was not already used to select the rule.

Save every version, including failures. Report how the effect changes across subperiods and markets without selecting only favorable examples. International indices share market shocks, so observations across countries are not automatically independent; uncertainty estimates should account for the dependence in the chosen dataset.

A few lunar cycles can reveal mistakes in timestamps and order handling. They cannot establish a stable return advantage. Likewise, an attractive in-sample p-value cannot substitute for an executable rule, costs, and a holdout result.

Frequently Asked Questions

Does buying a new moon and selling a full moon work?

This article does not provide a verified net-performance result for that rule. The cited papers test specified historical relationships; their findings cannot be used as a current forecast for a different implementation.

Should stops be wider around a full moon?

The cited 2006 study does not establish that requirement. A volatility-based stop needs a measured volatility rule and its own comparison, rather than a lunar label alone.

Which timeframe is best?

Choose the observation interval needed to implement the timestamp and fill rule. Daily data may be sufficient for a next-session-open test; an intraday event rule requires intraday data. Neither choice establishes superior performance.

How much capital should be assigned to lunar trading?

There is no evidence here supporting a universal allocation. First determine whether a predefined rule survives costs and untouched data. Portfolio sizing would then require information about loss tolerance, liquidity, and other exposures that this article does not have.