The Truth About 100% Win-Rate Seasonal Patterns

Where “100% win rate” comes from

A seasonal pattern is a recurring calendar window — for example, “long this stock from early October to mid-October” — that has tended to move in one direction at the same time each year. If you scan a decade of history and keep only the windows that won every single year, you get a list of patterns with a flawless record. That is exactly how the patterns in our weekly screener are surfaced: index members, a 10-year lookback, at least 10 occurrences, and a 100% win rate.

A perfect record is a powerful headline. It is also, on its own, almost meaningless — because the record was used to select the pattern. The interesting question is not “did it win every time in the window we picked it from?” (it did, by definition) but “what happens next, and what happened before we were looking?”

So we ran the test. We pulled all 37,430 patterns that currently carry a 100% record across 634 index constituents, then reconstructed every yearly occurrence of each one back to 1981— 1,368,538 individual trades in total. For each trade we recorded the day-by-day path and the final result, and we split the data into two halves: the recent 10-year window the patterns were selected from (where, by construction, every trade is a winner) and all the earlier years — genuine out-of-sample data the selection never saw.

Question 1: If the first day goes against the pattern, is it doomed?

This is the question most traders actually ask. You enter a seasonal long, the stock drops the first day, and you wonder whether to bail. On the surface, the data looks damning. Out of sample, when the first full day held moves with the pattern, the trade finishes green 63.1% of the time; when day one moves againstit, that drops to 48.6% — below a coin flip.

But that gap is a trap. The first day's move is already part ofthe window's total return — a red day one mechanically drags the sum down before anything else happens. The shorter the hold, the bigger that single day looms. And sure enough, the gap explodes for short windows and shrinks for long ones:

That is the signature of an artifact, not a signal. To settle it, we ran the decisive test: condition on the first day's direction, but then measure only the rest of the trade— from the day after entry to the exit, with day one excluded from the outcome. If an early adverse move genuinely predicted trouble, the remainder should suffer too.

The signal vanishes. Once day one is out of the outcome, an adverse first day leaves you with an identical— in fact a hair higher — chance of the rest of the trade working. The same held for the entry day's own move: whether the stock was up or down on the day you entered made no difference to the eventual result (56.2% vs 57.1%).

Question 2: Does the 100% record actually hold up?

Here is the result that matters most. In the 10-year window the patterns were selected from, the win rate is 100% — it has to be. But step back to the years before that window, which the selection never touched, and the same 37,430 patterns won just 56.6% of the time (about 58% for the long-only subset that makes up the bulk of the list).

That is not a flaw in seasonal trading — it is what selection does to any metric. A 56–58% base rate with a positive average return is a real, tradable tendency. But the gap between 100% and 57% is the single most important number in this study: it is the difference between a marketing headline and an honest expectation. Win rate alone, especially a perfect one, tells you how a pattern was chosen, not how it will behave.

This is precisely why we don't stop at win rate. A pattern with a flawless recent record can still be deep in a cold stretch — which is the entire idea behind equity curve regime filtering: reading where a pattern sits in its own hot/cold cycle rather than trusting the headline number.

Question 3: Can trend or market regime filter the losers out?

The obvious next move is to add a confirmation filter — only take longs when the stock is above its 200-day average, or when the broad market is in an uptrend. We tested all of them on the out-of-sample data. They did nothing useful, and several were mildly backwards:

Every “favorable” column is flat or slightly worse. These long seasonal patterns tend to behave like dip-buys: they did marginally better when the stock and the market were soft, not strong. The intuitive “only trade with the trend” rule is not just useless here — it points the wrong way.

Question 4: So what does hold up?

Two conditions — and only two — moved the needle out of sample.

1. Magnitude, not win rate

All 37,430 patterns share the same 100% win rate, but they differ enormously in how much they historically moved. Rank them by average historical return and the out-of-sample win rate climbs cleanly: the weakest third won 55.7%, the middle third 59.3%, and the strongest third 61.1%. Size of edge generalizes; a flawless-but-tiny pattern does not.

2. The calendar month of entry

This was by far the strongest factor — a 23-point spread. Long patterns that entered in October and December held up best out of sample; those entering in August and September fell apart, despite their perfect recent records:

Restricting long entries to October alone lifted the win rate from about 58% (all months) to roughly 66%, with the average return rising in step. This is the well-documented strength of the fourth quarter and the late-year rally bleeding through — and the August–September weakness is exactly where the over-fit patterns concentrate.

The short side, for the record, was weak everywhere: short patterns won only ~43% out of sample, and — counter-intuitively — shorting the seasonally weak August and September months was the worst approach, not the best. July was the only month where shorts cleared a coin flip, and only barely. The durable edge in this data is long, and it is concentrated in the calendar.

What this means for trading seasonal patterns

None of this says seasonal patterns don't work — a 56–66% base rate with a positive expected return, on a systematic, rules-based window, is a genuine edge. What it says is that the metric on the label is the wrong one to lean on.A perfect win rate is the easiest number to manufacture and the least informative. The things that actually traveled out of sample were the size of the historical move, the time of year, and — not tested here, but the natural next filter — where the pattern sits in its own win/loss cycle.

Practically: don't panic over an early adverse day, don't bother with trend confirmation, rank patterns by magnitude rather than by a perfect record, weight toward the strong end of the calendar, and check the regime before you assume the streak continues. That is the difference between trading a backtest and trading an edge.

How we ran the study

We selected every seasonal pattern carrying a 100% win rate over a rolling 10-year lookback, with at least 10 occurrences, on the index members in our universe — 37,430 patterns across 634 symbols. We then reconstructed all 1,368,538 yearly occurrences back to 1981 using daily closing prices, scoring each trade close-to-close in the pattern's direction. “In-sample” means the recent window the patterns were chosen from; “out-of-sample” means every earlier year, which the selection never saw. One caveat: occurrences of the same pattern across years overlap and cluster, so effective sample sizes are smaller than the raw counts — we treat 1–3 point differences as noise and the 5–23 point differences above as signal.

Frequently asked questions