[bumblebeard]

This study was conducted over the course of 15 years with 571 wine experts:

http://onlinelibrary.wiley.com/doi/10.1111/j.1755-0238.2008....

They found that individual judges were generally not very consistent in their rankings but that if you use small groups of judges then you can get better reproducibility.

The exact numbers are in the paper, but about 67% of the experts had statistically significant reliability for red wines and about 50% had statistically significant reliability for whites.

Also interesting is John Cleese's film about wine in which he conducted a blind taste test that found that most tasters couldn't discriminate between white and red wine. I couldn't find it online, but the whole thing is worth watching if you're interested in wine.