JOURNAL · 9 JAN 2026
Why a 20-period average arrives after the turn you already see
The window still contains the old closes. We count the sessions between a bend in price and the flattening of the average, and we call that count the lag.
A simple moving average of twenty closes is the sum of those closes divided by twenty. When the path bends, the window still holds the older numbers. The average cannot forget them until they slide out of the twenty. That delay is ordinary arithmetic, and it is the whole subject of evening 5.
Malai’s January sheet marked a high in a past series of daily closes, then marked the later session where the 20-period slope finally went flat under the triangle test. The count between those two sessions was the lag you could point at. It was not a forecast and not a flaw in the pencil. A longer window would have taken still more sessions. A shorter one would have turned sooner and lied more often about small bends.
People who meet moving averages on a forum usually meet them as arrows. The arrow hides the window. In the chart room the window is written as a bracket under the line: these twenty closes, no others. If you cannot draw the bracket, you are not ready to talk about the turn. Several students that month discovered their line was twenty sessions on the label and seventeen in the arithmetic, because they had skipped a holiday with no close. The bracket caught it.
Exponential lines lag differently. They never drop a close off a hard edge; they fade it. The line still turns late, and the lateness is softer. Evening 6 redraws the January sheet with an exponential average so the two delays sit on top of each other. The exponential hugs the bend sooner. It also fidgets on the quiet weeks. The tradeoff is visible without anyone declaring a winner.
Take one archived month and count the lag yourself before you change indicators. If the count surprises you, the surprise belongs in the margin. The average did not betray the chart. It reported the window you gave it, a little late, on purpose.