I thought the same thing. I expect it's just a mistake, but it could make sense in a way: it's possible over-estimates of price aren't counted, because any amount over and you lose. So if only successful guesses are averaged, only the negative side of the variance is going to be shown in the graphs.

If that's true and in reality variance in both directions is higher than it used to be, you might want to play it safe by reducing your guess, to limit the chances of going over.

... but that would probably still be a good way to lose the game, so most likely the author just got mixed up.