Three rules. The SIP that survives the next crowding unwind.
Recency loads tomorrow's SIP into yesterday's winner. Crowding writes the exit. Three checks before the click. None of them ask what last year was.
01
Name the factor. If the screen has a label, the crowd has one too.
A smallcase that ranks by trailing twelve month return is a momentum tilt. A screen sorted by P or B is value. A screen filtered on ROCE above 18 is quality. Once a factor has a name, more capital runs it and the spread narrows. The 26 percent post publication decay in McLean and Pontiff 2016 was paid by the next cohort that did not know the trade had a name.
02
Read the longest window the strategy has, not the loudest one.
A 60 percent year is not the factor's mean. It is one tail of a distribution that also contains a minus 28, a minus 21 and a minus 19 across three regimes. If the rolling worst 12M window is missing from the brochure, the rolling worst is doing the heavy lifting. Ask for it. The mean tells you the prize. The tail tells you the toll.
03
SIP into the strategy you can hold through its next crash.
A factor returns to its average. Bigger deviations snap back harder. The SIP that gets cancelled three months into a 20 percent drawdown earns the drawdown without the recovery. The right test is not "does this beat last year". It is "can I keep adding the same Rs every month while the chart goes red for nine straight months". If the answer is no, the screen was right and the strategy was still wrong for you.
Closer
The brain reads last year's return. The market is already reading next year's reversion. Last twelve months is a record. The forecast is the regime you are about to step into.