@JThorne: Okay let's talk about this for a second because I have a little bit of experience with it.
Straight rating systems have their problems. Why? Because everybody has a different scale. On a scale of 1-10 someone might never rate a card a 10 because he thinks it's not a must buy on every board and therefore not worth a 10. For the same reasoning he decides to never give a 1 because there are situations where you might want to buy it. Another person has no problems giving out 10s and 1s. You might think it doesn't matter but it does. Someone goes through the list and starts let's say with Cartographer (just an example) and reasons "It's a really useful card on many boards, way above average, let's give it a 7." Later he compares the rating of 7 to stronger cards and he gives out a lot of 8s and 9s. Because he started so high he has a lot of high ratings and Cartographer ranks in his list below average even though he thinks it's an above average card. So he rated maybe a card actually worth around 4.5 a 7 which is a huge difference. Most people tend to rate higher anyway. IMDB has on some page written what the average movie rating is. The last time I checked I think it was at 6.7, so you see a movie with a 6.5 and might think it's a decent movie, but in fact it's slightly below average. I think BGG has the same problem although I can't prove that. And no way the voter is going to correct all of his ratings after he found out that he tends to rate higher. But ranking is way better. The numbers I give out with each ranking is like a rating, but only in percent instead of a number. We can say that the community thinks of a card with 50% as an average card, a card with 70% is a good card and so on. There is no difference in 70% and 7/10.
I can see that ranking is more cumbersome than rating. But there is a solution. Every rating can be converted into a ranking. Let's say you rated 101 cards and you gave 15 cards a 10 and 32 cards a 9 and so on. The 15 cards get then 101-(1+15)/2 = 93%, the 32 cards get 101-(16+47)/2=69.5% which is fine as you rated almost half the cards a 9 or 10, so the ranked value has to be way lower. As you can see rating on a value of 1 to 10 isn't great when you evaluating 101 cards, but it's up to you if you rate 1 to 10 or 1 to 100. I already plan to give another option to rate the cards. Not everyone wants to do the ranking, so that's fine, but if someone rates it, we at least get more votes and more accurate data. Also it gives the option to rate the cards from like 1 to 10 and then use the vote method to finer rank the cards within the same rating. That means less comparisms should be made to get a full ranked list.
But that still gets not rid of the grouping problem. Even if you rate all the cards from 1 to 10, and you rate a Cartographer and then a Bonfire you have trouble finding a right rating while you ask yourself which of those is better automatically. So I stand by that. Grouping the cards by cost is still the best method as they are more comparable. If you rate your cards 1-10 and don't care about grouping, fine, you can do that (now I basically promise to program the new rating feature until next time, I hope I can manage that!).
Edit: Well I wrote that while eating dinner, so some time has passed and 14 new replies, but glancing over them, some of the stuff I wrote was already covered. I hope that by adding that feature we get a better result because more people vote.