Starts to freeze at 38°F? The equilibrium melting point of pure water is 32°F (459.6°K). That means that, as you cool it, it _begins_ to freeze at 0°C (273.15°R)*.

For crystals to form, there must be a lack of local motion. The closer to freezing temp you're at, the more "quiescent" (immobile) the material must be. If it's moving, it ain't gonna freeze easily. That's why waterfalls freeze more slowly than pools. Friction is an insignificant factor, compared to the effect of local motion.

The points that Gio makes about the effects of impurities and atmospheric pressure are valid.

Regardless of all the above considerations, I don't think I've ever paddled in water that was significantly below 32°F. The min air temp that I've braved was somewhere in the low teens F (-10 kg), but the water was probably right at 32.

(Credentials, or "why you should believe what I say": I make my living studying, primarily, crystallization and melting phenomena. It's not water that I [usually] study, but freezin is freezin.)

* I've heard that one can make snow at temperatures above 32°F. If so, that technology employs the heat of vaporization of some water particles to cool some other particles into freezing. However, that's not practical. For snowmaking, in fact, one important hurdle to get around is the heat of fusion: as it freezes, the water actually gives off heat, which makes it difficult for the rest of the water to freeze. http://www.arecosnow.com/Snowmaking.html says that it's pretty much uneconomical to make snow above 27°F (-3 ml).