How this site measures pattern hit rates

Every figure on this site comes from a written rule applied to price histories the site keeps itself: ten Indian large-cap stocks (Reliance Industries, HDFC Bank, ICICI Bank, Infosys, TCS, State Bank of India, ITC, Larsen & Toubro, Axis Bank, Bharti Airtel) and the Nifty IT index, ten years of daily bars each, read once from Yahoo Finance on 6 November 2026. This page explains what is counted, what counts as a hit, and what the count leaves out.

What counts as an occurrence and a hit

An occurrence is a stretch of history where a written rule finds the pattern; each model has its own rule, quoted on its page, and a figure counts only when a close completes it. A hit is then judged by one definition for all models: from the completing close C, and the ATR of that session, a close 2 ATR from C in the textbook direction within the next 20 sessions, before any close 1 ATR the other way. A close 1 ATR against first is a miss; neither within 20 sessions is undecided and counts as not a hit. Only closes decide.

The three numbers were fixed in the measuring tool before any count was run and have not been changed. A figure with fewer than 20 sessions of history after it is left out and counted separately.

How the history is kept

The site stores ten years of daily bars, 7 November 2016 to 5 November 2026, for each of the 11 instruments, read once on 6 November 2026. The history is not extended as new closes arrive and no count is redrawn from a newer copy: the figures on the pages are the figures of that stored copy, and the date is printed under every chart. Nothing ticks live.

One cleaning step comes before any search. A bar whose high equals its low carries no trading range; the source prints exchange holidays and empty days that way. Such bars are left out before the rules run: two bars on each of nine of the stocks and one on ITC, 19 in all, and none on Nifty IT. They cannot pass for an inside bar or distort the ATR.

A rare pattern stays rare in the table and a share that rests on few occurrences stays rough. Each table prints the count next to the share, and a yardstick (the same test from every session) next to both.

What the count leaves out

The count ignores spreads, commission and financing costs. It also ignores the size of the winning moves compared with the losing ones. A pattern that hits often but loses more on its misses can still lose money overall.

The count is a measurement of past occurrences in the stored history. It is not a probability for the next occurrence, and it says nothing about what will happen after the pattern appears again.

Reading the tables without misreading them

Each model page carries a table above the text. The columns show the instrument, the number of occurrences the rule found, how many were counted, the hits, misses and undecided, the hit share and the average result in ATR, with a final row for all instruments and a line for the yardstick, the share the same test passes from any session in the same direction.

The share is a fraction of the counted occurrences, not a forecast. A small number of occurrences makes a rough share; one result changes it by 100 divided by the count. The average result is the distance from the completing close to the deciding close in ATR, positive in the textbook direction, and it is before costs.