Wine scores cloak wine with a false sense of objectivity and precision. If they are to have any rigor, they should be replicable under various situations. Joel Peterson, the Champagne-sipping founder of Ravenswood Winery, has a telling vignette in this regard:
I have made wines under two labels and I’ve had them scored in the same periodical as much as five points apart. Same wines, different package. Like all judging of wine, it suffers from the wine that came before, the wine that came after, the time of day. All those things can affect a wine by as much as five points.
Many Burgundy reviews have higher scores for each wine at a domaine as the prestige of the appellation increases. To add some rigor, it would be interesting to see if those results were replicable if the wines were tasted blind or intentionally mislabeled. But I guess the folks at Caltech already did something similar to that.