Fixed before any outcome, on the 95% range of the difference: wholly above zero and reaching 0.31 gives yes if the loss side is near zero and the difference reaches 0.31 with the share's recent volume and change in price swings allowed for and where both thirds' gains overlap, else partly. Above zero but under 0.31, or under 0.31 and including zero, gives not at the size tested. Anything else, or a failed floor or precision check, gives not settled.
Data and method
Each session, shares priced five or more at the previous session's close are split into thirds by recent trading value; within each third, the day's volume against its own recent normal is fitted on the gain and loss of the previous twenty trading days, allowing for the day's move, recent price swings, price and trading value. The thirds' slopes are differenced and averaged, read in stretches of about three months, because neighbouring days share most of the same twenty-day gain.
The answer
Partly. Over 1,259 sessions from 2018 to 2022, the least traded third's volume rose by 1.30 log points per unit of return and the most traded third's by 0.61, on average across each third's own gains: a difference of 0.69 (95% range 0.51 to 0.87), wholly above the 0.31 set in advance, where 0.31 means about six percent more volume after a twenty percent gain. With the share's own volume over those twenty days, and how much its price swings changed, allowed for, the difference was 0.47 (0.34 to 0.61), reaching the bar. It reached the bar inside groups of shares of similar price, 0.69 (0.49 to 0.89), and over the range of gains both thirds share. On the loss side the difference, 0.21 (from 0.08 below zero to 0.49 above), was neither shown near zero nor shown to differ.
Main estimate
0.690 ratio (95% range 0.514 to 0.867), from 1259 observations
What it changes
Under the rule fixed in advance, the steeper average rise in the least traded third is recorded as present at the size with its loss side open, and no rule for reading trading volume in later studies changes. The next study choice weighs one repair of this test against the programme's next test: whether the rise differs for shares above or below the average price they traded at over the past year.
How it was made
Run took 68 minutes; independent code review: approve; the recorded calculation reruns to the same result.
Fingerprints
plan 63bed399cf5a50917c8b60fa9ed4c3e6a38c4741fdbf9a08335bfb698ce3696c result aa57cbc7bbe89317c3f23087b9f9ba158707dcd36ff75a5b9e60b692cd948942
Exploratory work on development data; not a performance claim.