Going by "effect with" (i.e. number of standard deviations better than expected I perform) my top 5 are:
1) Bag of Gold: 20.23
2) Treasury: 2.52
3) Bureaucrat: 2.17
4) Monument: 2.11
5) Torturer: 2.12
Its not that I buy these cards a lot but rather that I feel comfortable with when to use them. Which for Bag of Gold, my number one, is almost never. My %+ is 3.4% for this card, compared to 17.8% in general. As a rule of thumb, unless Bag of Gold creates a guaranteed Province for me this turn (say, a card drawer and a Remodel in hand), I'll take a Duchy instead.
Likewise with Treasury, I think I see people buying it too often. To me its only worthwhile if greening is likely to occur late in a game, and there are engine components of high value and high cost. I also like it if there are means to gain rather than buy the green, such as Border Village, Ironworks, Horn of Plenty, etc. Otherwise, I tend to not rate it much.
Bureaucrat and Monument I think I know when to use, which is basically weak kingdoms where money matters, or where there's a specific benefit to them (like a game with lots of green predicted for bureaucrat (silk road, gardens rush, workable tunnels) or an opportunity to play multiple monuments per turn in a province game (say chapel, village, monument)). I probably veer more towards money than a lot of players, so I play Bureaucrat as an opener quite a lot, as a lot of kingdoms don't seem fast enough to compete against its mix of treasure generation and light attack.
Torturer I tend to play only when I am playing for a Torturer lockdown, and I almost never "splash" into this card. This card isn't really worth its cost until you play two of them in a turn.
Incidentally I think "effect with" is probably the best column to go by when judging your effectiveness with a card, as "winrate given availability" tends to be down to the strength of the cards mostly.