ABSTRACT
INTRODUCTION We made our forecast of Data Set A by searching the given sequence of 1,000 measurements for segments that nearly match the pattern of the last 21 and then guessing that similar antecedent segments are followed by similar consequent segments. This description fits almost any forecasting procedure, but, while most entries in this competition are interesting because of the computer pattern-matching techniques used, we did our pattern matching by eye-plotting segments of the data on vellum tracing paper, laying the plots on top of each other and shifting them to obtain the best match.
