How do you compute momentum, and why skip the last month?
Momentum needs no accounts and no ratios — only prices, which exist for every listed share back decades. The recipe fits in one line and has one detail that every beginner omits and every practitioner insists on.
Twelve minus one
The standard measure is the return from twelve months ago to one month ago: 12-1 momentum. Take the adjusted close on the date a year back and the adjusted close a month back, and divide. The most recent month is skipped on purpose.
The reason is a separate anomaly with the opposite sign. Over horizons of a month or less, shares tend to reverse — last month's sharpest risers lag, last month's fallers bounce — Jegadeesh found it in 1990, three years before the momentum paper. Include the last month and the two effects fight; skip it and the momentum signal is cleaner. The convention is so standard that "momentum" in a factor paper means 12-1 unless it says otherwise.
Measured, 3 September 2026
Ten names, adjusted closes on 3 September 2025, 3 August 2026 and 3 September 2026:
| Ticker | 12-1 return | Last month | 12-0 return |
|---|---|---|---|
XOM.US |
+41.7% | +5.3% | +49.1% |
KO.US |
+29.5% | +2.2% | +32.4% |
IWM.US |
+28.1% | −0.3% | +27.6% |
AAPL.US |
+27.6% | +8.3% | +38.1% |
QQQ.US |
+23.4% | +2.5% | +26.5% |
NVDA.US |
+21.3% | +10.6% | +34.1% |
JPM.US |
+20.0% | +2.7% | +23.2% |
SPY.US |
+19.0% | +2.0% | +21.4% |
PFE.US |
+8.0% | +15.1% | +24.4% |
MSFT.US |
−2.9% | +4.8% | +1.8% |
Two rows show why the skip matters. Pfizer's 12-1 momentum was the weakest positive figure in the list, and its last month the strongest: a 12-0 sort lifts it from near the bottom of the table to the middle, still below Apple; a 12-1 sort leaves it near the bottom. Microsoft's twelve months were negative until the last four weeks; on the standard measure it is a momentum loser, on the naive one it is flat. Which measure you use is which stock you hold.
What the sort then does
Rank the universe on 12-1, buy the top tenth, hold a month, re-rank. In the table the winners are an oil major and a soft-drink company, and the loser is the largest software company — momentum has no opinion about businesses, only about what has been repricing. A year earlier the ranking looked different, and a year later it will again, which is the turnover the next lesson charges for.
In the data
Each row of the table is three prices. Here are Pfizer's:
Adjusted closes, because a dividend or a split in the window is not momentum: Pfizer paid dividends during that year, and on the prices as traded each payment would read as a small fall. A real sort needs the same three prices for every name in the universe, a thousand names and three thousand prices, which is work for a program rather than a page; the quant-coding domain's dashboard course builds it.
Try it now
- Reproduce one row. From Pfizer's three adjusted closes above, compute 12-1 and 12-0 and check them against the table. Then do the same for a share of your own choosing that reported results in the last four weeks: open it in the Terminal (the link starts on Pfizer; change the symbol), read its closes a year ago, a month ago and today off the chart, and write which measure your share prefers.
Open PFE.US in the EODHD Terminal
- The two anchors on a chart, so the skipped month has a shape:
Measure from a year ago to a month ago, then from a month ago to today. The first is the signal; the second is what the signal ignores. 3. Out loud, without notes: what would a 12-1 sort have done in the spring of 2009, holding last year's winners and short last year's losers, when the losers were the banks. The drawdowns lesson is the answer with numbers.