Best and Worst Month Analysis: What Extreme Months Reveal About Risk

Best month and worst month are exactly what they sound like: the largest gain and the largest loss a strategy or portfolio produced in any single month of its history. They sit near the bottom of most performance reports and get skipped, which is a mistake, because they are often the two most informative numbers on the page.

The reason is that most performance statistics are built to describe the middle of the distribution. Averages, compound growth rates and volatility figures all summarise typical behaviour, and in doing so they smooth away the extremes. Best and worst month do the opposite. They report the extremes directly, unsmoothed, and the extremes are what actually determine whether an investor stays with an approach.

What they measure

The calculation is as simple as it looks. Take the return for every calendar month in the history. The best month is the highest of those values, and the worst month is the lowest. Some reports also show the best and worst quarter or year, and the more useful ones show the full distribution of monthly returns rather than only its endpoints.

Two related figures usually sit alongside them and are worth understanding together:

• Percentage of positive months. The share of months that ended above zero. This is the monthly analogue of a win rate, and it carries the same caution: frequency of positive months says nothing about their size.
• Average positive month and average negative month. The mean of the up months and the mean of the down months, which describes the asymmetry between good and bad periods.

Together those give you a rough shape of the monthly return distribution: where its centre sits, how often it is positive, and how far it reaches in each direction.

How to read them

Start with the worst month, not the best. The worst month is a lower bound on unpleasantness that you know actually occurred, as opposed to a modelled estimate. Volatility and value at risk both estimate how bad things could get from assumptions about the distribution. The worst month is not an estimate. It happened. That makes it a useful reality check on any modelled risk figure: if a model implies that a month of that size should be extremely rare, and yet it appears in a fairly short history, the model’s assumptions deserve inspection.

Then compare the worst month with the maximum drawdown. These measure different things and the gap between them is informative. If the maximum drawdown is roughly the size of the worst month, the strategy’s bad episodes are short and sharp. If the drawdown is far larger, the damage was accumulated over a run of moderately negative months rather than one dramatic one, which is a very different experience: slower, longer, and in many ways harder to sit through.

Then treat the best month with the same suspicion you would give any outlier. A very large best month raises an immediate question: how much of the total result is attributable to that single window? The test is direct. Remove the best month and look at what remains. If the picture changes materially, the record is an account of one episode plus some noise, not a description of a repeatable process. This mirrors the deletion test applied to individual trades in average trade profit.

Then check when the extremes occurred. Extreme months are rarely random in timing. Large negative months in an equity strategy tend to cluster in broad market falls, and large positive months often follow immediately afterwards. Knowing whether a strategy’s worst month coincided with a market-wide fall, or happened in isolation, tells you whether you are looking at market exposure or something specific to the rules.

Then check whether the extremes are still possible. A history that contains no severe month may simply not have covered a severe period. The absence of a bad month is evidence of a short or benign sample at least as often as it is evidence of resilience. Always ask what the sample window includes and, crucially, what it excludes.

What you see	The reading
Worst month close to maximum drawdown	Bad episodes are short and sharp
Maximum drawdown far larger than worst month	Damage accumulated over a run of poor months
Best month very large relative to the rest	Result concentrated in one window, test by deleting it
No severe month in the history	Possibly resilience, more often a short or benign sample

What they do not tell you

They do not tell you the shape of the distribution between the extremes. Two strategies can share an identical best and worst month and behave completely differently in every other month. The endpoints are two observations out of many, and reading only them discards nearly all the data. The full monthly distribution, ideally as a histogram, is the honest version.

They do not tell you about sequence. A worst month followed by a swift recovery is a different experience from a worst month followed by two flat years. Best and worst month are single points with no memory of what came next. Recovery time and the underwater curve answer that question, and this is the province of drawdown recovery analysis.

They are artefacts of the calendar. A fall that begins in the third week of one month and ends in the second week of the next may never show up as a large monthly number at all, even though the peak-to-trough loss was severe. Monthly bucketing is a convenience, not a natural unit of risk, and a strategy can have a benign worst month and a nasty worst six-week stretch.

They are bounded by the sample. The worst month in a record is the worst month so far. It is not a limit, a floor, or a promise. Reading it as a maximum possible loss is one of the more dangerous misreadings in performance analysis, and the reason stress testing exists as a separate discipline.

They say nothing about why. An extreme month has a cause: a concentrated position, a sector event, a leverage decision, a liquidity squeeze, or simply the market. The number alone contains none of that, and the cause is usually more informative than the magnitude when you are judging whether it can happen again.

They are affected by how the return was measured. Gross versus net of costs, and whether the figure is a time-weighted return or a money-weighted one, both change monthly numbers. As always, the basis has to be stated before the number means anything, which is the same point that separates gross from net profit and loss.

The practical value of these two numbers is psychological as much as statistical. Most people can describe the return they want. Far fewer have thought concretely about the month they would have had to sit through to get it. Best and worst month put that month on the page in one line, and reading it honestly, before rather than after, is what turns a performance report into something decision-useful.

Related reading

• Portfolio and Backtest Metrics, Explained: the hub guide to portfolio and strategy metrics and the limits of each.
• What Is Maximum Drawdown: the largest peak-to-trough fall, and why it matters more than volatility to real investors.
• Volatility and Standard Deviation, Explained: what standard deviation of returns captures, and what it smooths away.
• Drawdown Recovery Analysis: time to recover, underwater curves, and the arithmetic of losses.
• Stress Testing a Portfolio: how teams look beyond the worst month the history happens to contain.

This article is educational. Altys Labs is not a registered research analyst or investment adviser, and nothing here is investment advice or a recommendation to buy, sell, or hold any security.