Prior research shows positive abnormal returns follow the announcement of a new share repurchase program, but do firms earn abnormal returns on their actual repurchases? An obstacle to answering this question has been poor data on the timing of corporate share repurchases. I build a near complete dataset of monthly share repurchases using software I wrote to extract the data from 10- Q and 10-K filings. Constructing portfolios based upon the share repurchase signal, I find small to medium firms earn abnormal returns while large firms do not. Furthermore, I find evidence consistent with a market reaction to the disclosure of share repurchases. Firms experience positive abnormal returns around their 10-Q and 10-K filing date relative to non-repurchasers with public programs, and using machine learning techniques to forecast share repurchases, I find the share repurchase surprise, actual minus forecast repurchases, is associated with positive abnormal returns around a firm’s earning announcement and 10-Q, 10-K filing dates.,My main contributions are thus threefold: I provide systematic, machine extracted data on repurchases, show that small to mid-size firms have positive abnormal returns while large firms do not, and use machine learning techniques to forecast repurchases and show abnormal returns around earning announcement days and 10-Q and 10-K filing dates are positively associated with the unexpected component of repurchases, the share repurchase surprise.