Kim Man Lui and Lun HuMany technical price patterns in financial time series used as trading strategies are learned by traders’ astute observations on the movement of price vectors. The authors propose data mining techniques to explore relationships between price movements and price vectors automatically. One such relationship, candlestick patterns, was discovered by human observation.The authors studied three years of data from 42 Hong Kong Hang Seng Index composite stocks.By two-level clustering to deal with shape and scale of the daily price vector of a time series of a stock price, a time series is converted into a symbolic price sequence. The previous price trend and the next price trend on each trading day can be converted into two symbolic trend sequences. The authors then looked for correlations between the symbolic price sequence together with the previous trend sequence and the next price trend sequence. The result shows three patterns have statistical significance (p < 0.05) on only two out of 42 stocks and indicates that the discovered patterns for one stock may not be the same kind for others, meaning that there is no generic pattern for the two assets. The authors conclude that price patterns, if any, as reported in technical analysis literatures, should not be equally applicable to any time series of stock prices.