The Mpemba Effect

A graph of freezing rates for two water samples demonstrating
the Mpemba effect. The gray line is water at 42.9°C and the black
line is water at 18.6°C. Note the hotter water spends considerably
less time in the 0–4°C range, likely due to convection.
Source: picotech.com.

As odd as it seems, sometimes warmer water freezes faster than cooler water. This observation is known as the Mpemba effect, names after Erasto Mpemba from Tanzania. Even though the effect had been known for years to layman such as plumbers and ice rink workers, the scientific community had not taken the claim seriously. But Mpemba, a high school student at the time, provoked a visiting physics professor to test his claim and to his surprise it held true. They published the results jointly in 1969. The Mpemba effect is only observed under certain conditions and there are many factors which can effect how quickly water cools. 
  Conduction is often the biggest factor. Take two similar containers, one containing cold water and the other containing an equal amount of hot water and place them both in the freezer. The hot container will melt any ice on the freezer shelf that it comes in contact with and refreeze as the container cools, creating a good thermal contact between the freezer shelf and the hot water container which will remove heat from the warmer water more quickly. On the other hand, the cooler water container will just sit on the layer of poorly-conducting freezer frost, so it will take longer to cool down. This is a pretty good explanation but even if you control for it by insulating the contact points to the freezer you can still observe the effect. 
  The explanation most-often cited is convection. As the warmer water rapidly cools at the surface, convection currents will develop and create an uneven distribution of temperature with hot water nearer the surface and therefore more evaporation from hot water than from cold. More evaporation means less mass to freeze so it would cool faster. And while this is probably the biggest contributing factor it’s not the only mechanism that drives the Mpemba effect, as not enough water would be lost to evaporation to account for the difference.
  Supercooling is another possible factor that contributes to the phenomenon. Initially cooler water is thought to dip farther below 0°C before freezing, but the mechanism for this is not well understood. Another factor could be the amount of dissolved gases which would be less in water that has been heated. But the most probable conclusion is that there just isn’t any single unique explanation as to why hot water sometimes cools more quickly than cold.