Methodology: how the PredictivePick momentum signal works

What the model predicts

Every trading day, for each stock it covers, the model estimates one probability: the chance that the stock's closing price 34 calendar days later will be above its 150-day exponential moving average (EMA) at that time.

The 150-day EMA is a long-term trend line that gives more weight to recent prices. A close above it means the price is running ahead of its own longer-term trend. A high probability therefore says the model expects the stock to be trading above that trend about five weeks from now. It is not a forecast of a particular price, and it does not predict by how much the price will move.

The model and its inputs

The signal comes from a classification model: for each stock and day it outputs a class (above or below the trend line at the horizon) and the probability behind it. The model includes a neural network used to find relationships between the price movements of different stocks, so one stock's estimate can reflect what related stocks are doing.

Its inputs include:

• Daily change in the closing price
• Trading volume
• Relative Strength Index (RSI)
• Reported company fundamentals
• Volatility
• Sharpe ratio and risk/return ratios
• A small number of further derived features

The model is trained on the previous five years of market data.

From daily probability to signal

A single day's probability can jump around, so the published signal uses the probability averaged over the last 5 trading days:

• High Signal: average probability of 80% or more
• Moderate Signal: from 60% up to 80%
• Low Signal: below 60%

These thresholds are read from the research platform's data every time it refreshes, so the numbers on this site always match the ones the app uses.

Coverage, data and updates

• Coverage: India: 96 stocks from the Nifty 100 (as of 8 Oct 2026). See all covered stocks.
• Data: publicly available end-of-day market data (adjusted closing prices and volume) and reported fundamentals.
• Updates: the model runs after each trading day's data arrives; the stock pages on this site refresh within a few hours of that.
• Recorded on the day: each day's probability is produced that day by the model running at the time and then stored. Past values are not recalculated later with a newer model.

What the signal is not

• It is not a price target and not a recommendation to buy, sell or hold.
• It is a probability: a High Signal can be followed by a fall, and a Low Signal by a rise.
• It does not account for news or events after the data date shown on each page.
• Markets change, and how the model has behaved before does not guarantee how it will behave next.

Where to see it

Each covered stock has a page on this site with its current signal, a 12-month price and probability chart, and the research notes written about it, starting from Stocks. The PredictivePick app shows the same signals with daily rankings. Research is led by Ruchika Patel, founder and quantitative researcher.