How often has it actually happened?
A base rate is how often something has happened before. It is measured across thousands of past instances. Each one is compared against what the market did on the same days. The result is reported with the uncertainty attached.
Three rules, and they are the whole argument
Registered before it was computed
Every pattern was written into a public dated log before any data was touched. All 300 variants were written there too. The headline uses the middle setting chosen in advance, not the one that looked best.
Every number has a placebo
A medical trial without a placebo arm proves nothing. If 70% of patients recover on your drug, that is meaningless. It stays meaningless until you know that 68% recovered on sugar pills. A 60% rate in a rising market is a sugar pill. Every number here is published next to its placebo.
The ones that showed nothing are here too
1 of the patterns below could not be told apart from the market. They sit in the same list, at the same size, as the rest.
Gaps
| Pattern | Gap vs control | 95% interval | Effective n | Sample |
|---|---|---|---|---|
| Gap up above 20dma The stock opened at least 2% above yesterday's close, on a day when it was already trading above its 20-day average price. | −3.7pp | −4.4 to −2.9pp | 34,316 | Strong sample |
| Gap down below 20dma The stock opened at least 2% below yesterday's close, on a day when it was already trading below its 20-day average price. | −0.5pp | −1.5 to +0.5pp | 20,564 | Moderate sample |
Price structure
| Pattern | Gap vs control | 95% interval | Effective n | Sample |
|---|---|---|---|---|
| Close crosses above 20dma The stock closed back above its 20-day average price after having spent at least five of the previous ten sessions below it. | −1.7pp | −2.3 to −1.2pp | 60,340 | Strong sample |
| New 60 day high The stock closed at its highest level in sixty sessions, on volume at least equal to its own twenty-day average. | −2.1pp | −2.9 to −1.4pp | 68,015 | Strong sample |
| Three down days in an uptrend The stock closed lower three sessions running while still trading above its fifty-day average price. | +3.3pp | +2.7 to +3.9pp | 79,956 | Strong sample |
Volume
| Pattern | Gap vs control | 95% interval | Effective n | Sample |
|---|---|---|---|---|
| Volume spike without price move Volume was at least three times the stock's own twenty-day average while the closing price barely moved — under 1% either way. | −1.3pp | −2.4 to −0.3pp | 9,111 | Strong sample |
Volatility
| Pattern | Gap vs control | 95% interval | Effective n | Sample |
|---|---|---|---|---|
| Narrowest range in seven sessions The day's high-to-low range was the narrowest of the last seven sessions, and under 2% of the closing price. | +0.9pp | +0.5 to +1.3pp | 1,32,421 | Strong sample |
Seasonality
| Pattern | Gap vs control | 95% interval | Effective n | Sample |
|---|---|---|---|---|
| Early session of the month The third trading session of a calendar month, for a stock trading above its hundred-day average price. | +1.8pp | +0.6 to +3.0pp | 9,638 | Moderate sample |
Market breadth
| Pattern | Gap vs control | 95% interval | Effective n | Sample |
|---|---|---|---|---|
| Market wide down day A session on which the median eligible security fell at least 2%, for a stock trading above its hundred-day average price. | +2.1pp | +1.2 to +3.0pp | 8,246 | Strong sample |
Registered, and not answered
These were written down with the others, but not measured yet. The data they need is not available here. They stay on the list.
The sessions following a quarterly results announcement, split by whether the stock rose or fell on the announcement day itself.
NOT COMPUTED. It needs a point-in-time announcement date for every result between 2011 and 2026, stamped by filing date rather than period end. The daily bhavcopy carries no such field and no free public archive reconstructs it back to 2011. Registered rather than dropped, and it stays here unanswered until that dataset exists.
Registered 2026-09-05. Log entry post-earnings-announcement-drift.
A single session's percentage change in a stock's futures open interest, split by direction.
NOT COMPUTED. It needs the derivatives bhavcopy rather than the cash one, and a futures-eligible point-in-time universe to go with it. Registered now, at the same time as everything else, so that whatever it eventually shows was declared before it was seen.
Registered 2026-09-05. Log entry raw-open-interest-change.
The sessions around a security being added to a broad index at a scheduled reconstitution.
NOT COMPUTED. It needs as-of index membership back to 2011. Today's constituent list cannot answer it — using it would be the reconstitution survivorship channel the method section warns about, measured instead of controlled for.
Registered 2026-09-05. Log entry index-reconstitution-entry.
Before you read a single number
How often you have to be right
A frequency has no scale until you know what your own bracket and charges demand. Work it out first. Every table here carries the same column.
Open →What the words mean
Base rate, matched control, effective sample size, break-even frequency — each with what it does not tell you.
Open →The registration log
Every pattern, with the settings chosen in advance and the date it was written down.
Open →Questions about this hub
What is a base rate?
A base rate is how often something has happened before. It is measured across thousands of past instances. Each one is compared against what the market did on the same days. The result is reported with the uncertainty attached.
Why is every number shown next to a control?
Because the Indian market rose in roughly two of every three rolling twenty-day windows over this period. Any long-side pattern therefore shows a frequency near 60% and means nothing on its own. For every event, comparable securities are drawn from the same date and measured exactly the same way. The publishable number is the gap between the two.
Why do some pages say the pattern showed nothing?
Because every pattern is registered in a public log before it is computed. Every registered pattern then gets a page, whatever the result. A site that publishes only what worked has told you nothing about how often it looked.
Does this name any stock?
No. Every figure is an average across hundreds of stocks over many years. The price data used is at least 30 days old, as SEBI's circular of 8 May 2026 requires.
Can I use these numbers elsewhere?
Yes. Every page has a downloadable file under the CC BY 4.0 licence. It holds every figure the page shows, with the sample, the quality checks and the robustness checks. Please credit the source and link to the method page.
How every figure here is produced
The method page covers the point-in-time universe, which keeps delisted securities. It covers the matched control drawn from the same dates. It explains the block-bootstrap interval. It lists the quality checks a sample must pass before any rate is shown. It also names the four things this data cannot do.
Read the methodology