I show that the growth of high-frequency trading, due to its heavy reliance on computer algorithms, can be associated with a reduction of human errors and financial anomalies in the market. Trades in which a non-high-frequency trader is the liquidity demander exhibit abnormally high buy (sell) pressure when prices are immediately below (above) a round number due to psychological effects, while the pattern is completely reversed when a high-frequency trader is the liquidity demander. As a result, the overall sample does not exhibit such imbalances. Furthermore, high-frequency traders earn higher returns when trading around round number prices.