Treat a volume-price surge as a contingency table

Start with a table

Popular Volume-price analysis treats a surge as a chart story. The evaluation task is to write that story as a contingency table: normalize the inputs, test whether volume category and price-change magnitude interact at all, and only afterward ask whether direction or profit follows.

A cross-tabulation of normalized volume bins against absolute price-change bins can be built from a random sample of 50 stocks to test whether larger volume categories coincide with larger price moves.

How the sample table was built

The working sample was drawn from every New York and American exchange listing between 1993 and 1994 as 50 randomly chosen stocks with about two months of daily data each.

Each observation is an ordered pairing of normalized volume and absolute price change. The bins, not a freehand chart, are what Hypothesis testing can judge.

What the cells showed

In that sample, daily price changes below 5% remain the most frequent outcome in every volume bin, and a log frequency scale is used because one cell contains 1,318 observations.

Price-change magnitudes above 10% become noticeable only after normalized volume reaches the highest categories, with the >300 bin showing the concentrated large-move cells.

Not every stock showed the surge effect. Higher-priced shares were less susceptible to large percentage moves.

What a Chi-square test can answer

A Chi-square test of the volume-by-price-change table answers whether the rows and columns interact at all, which is a weaker claim than a rank test that volume and surge size move together.

Resampling by generating thousands of chance samples estimates how often a chosen table pattern would appear under randomness. A tiny share supports treating the pattern as nonrandom.

Direction and profit stay later questions

Price surges greater than 5% were confirmed to relate to volume, but the sign of those surges was not shown to be determined by volume rather than chance, for individual stocks, each exchange, and the full sample.

Dollar gains during surges in the sample exceeded dollar losses by $1.4 million, but that increment sat inside a $66 million total change in stock values, so the profit implication of volume remained unclear.