The issue: I'm fairly new to all this, but in a short period I have already observed many discussions in the MP lobby regarding under/overvaluation of cards. People complain that some cards marked as bronze are way better than that bronze marking would suggest, for instance. It would be nice for people trying to comprehend the format if the colored title bars on the cards actually reflected the cards' quality. The suggestion: Separate the title bar from the rest of the card image, and draw the title's background dynamically. So when you display, say, "Parry," you grab the title, then you grab the card image minus the title, then you cobble them together, and you paint a background under the card title based on the card's 'actual quality.' A card's 'actual quality' would depend on how frequently the card is used by winning teams. I imagine that Blue Manchu has access to this data and that it wouldn't be hard for them to siphon it into a database of table in order to produce this functionality. For reference, here is an example of someone ranking Magic: The Gathering cards based on criteria that I think could work well for this suggestion. I think if Blue Manchu had a database automatically ranking cards like this constantly, and then had cards display a quality based on that ranking, it would be pretttty sweet! Let me know if there was anything unclear in my post, first time here
I'm not sure this will work too well. Partly because cards interact with other, and partly because cards are associated with items. If there is a great item everyone uses with two good cards and a drawback, the drawback will get boosted! If you get the data from PvP there is the metagame to consider too. You shouldn't be worrying too much about the colour of a card anyway. What matters is what it can do for you.
The fact that the metagame factors in seems like a bonus to me rather than a drawback. Card quality does increase or decrease based on the meta, and reflecting that accurately would be awesome. Drawback cards probably should be boosted—all traits, even negative traits, are almost strictly good from a deckbuilding perspective. Armor & blocks could be set to only register when the die roll is a success, and weak action cards on an item (like weak strike) are unlikely to be played much even if winning teams draw them. Something to consider. I think this system would mitigate the expectation issues I mentioned in my previous post while serving to just be more true. This seems a bit silly to me. Obviously when you're experimenting with builds or trying to find a way to topple the meta, you should look at "bad" cards, but the quality indicators are there for a reason: they are information. If the information is not to be used, it shouldn't be there, plain and simple. Obviously I think the current static indicators do have some value, and are intended to be used to some extent. But I think they could be improved.
That sounds interesting... but I see some minor issues (aside from the work involved). You know what would be immediately ranked as the best card in the game? Walk. So we'd need to impose our opinions on a card's importance at least a bit, and not just leave it up to which card gets used in the most winning decks. I'm sure that in your linked example, there's probably a similar system for ruling out essentials like land, but in the end, it creates a trend for certain cards to seem significant when they're nothing more than staples, or something you're even forced to use. I'd imagine that the data created would end up just being one more thing you have to make a lot of allowances for... which is exactly what we all do right now anyway with regards to the dev's opinion of which card should be of a certain color quality. I'd look at this new data and think, "Well, this card isn't used by a lot of winning decks, but it's probably just because it's rare. This card is used by almost everyone, but that's probably just because it's tacked on to a lot of things". Emphasis on probably. I'm still exerting my opinion as much as I ever was. :/ Your point about the traits is a good one, too. I regularly stack combustable on my mage to get at the good stuff that comes with it and reduce my deck size to increase the chances of drawing the stuff I want reliably. This would be a great clue to build into a card's quality; that people use it ins spite of the drawbacks. I wonder if this idea wouldn't work slightly better for pieces of gear, and not individual cards?
I think you could do a double pass and check both the % of games won in which a given card was played, and the % of games in which a card was played that the team playing the card won. Let's take an example card, Walk. Let's say for the sake of argument that every team runs Walk, and it gets played every game. This means that the winning team played Walk. But it also means the losing team played Walk. So while 100% of winning teams play Walk, only 50% of Walk-playing teams win. Because Walk is probably played close to evenly between winning and losing teams, I don't think it would rank very high. Still for argument's sake (arbitrary examples here), let's say Whirlwind Enemies is such a powerful card that if you play it, you're guaranteed to win. It's also rare, so maybe only 60% of winning teams play Whirlwind Enemies. But 100% of Whirlwind Enemies-playing teams win their games. Now we decide on our weight algorithms in such a way as to rank Whirlwind Enemies above Walk, and the numbers DO mean something. So I'm not sure that the "staple" problem is a problem at all, even a minor one. There could in theory be three, four, five, or even more different criteria for ranking, each with their own weight. Each one complicates the process slightly, obviously, but it's not hard to imagine a fairly intelligent ranking system. I agree that this would be sweet to see on items, but as I understand it, item rarity is a mechanical thing and not a purely informational; I don't know what consequences changing item rarity based on actual item quality would be, and I didn't intend to go there.
Okay. I think I'm getting a better picture. You'd probably want old data to slowly be removed over time as well. As new cards get introduced (as I'm sure they will be), the effect they have would slowly change the weight of any preexisting ones, and you want this to update dynamically, right? It would be interesting over time to see cartain cards gain/lose ratings, change their statistics, update themselves, and then possibly effect their scarcity over time. Sort of like games that let you mouse over an object in your inventory, and it will tell you in it's stats what it's currently selling for in the auction house. Sounds like a lot of work.
This is not going to work. Popularity (or winrate) of cards does not determine its power in this game. Players choose items to build deck, unlike choosing card to build deck in MTG. Some cards will be very common in pretty much any item. (Bludgeon line). Whether that item is bad or good, its pretty much got some standard attack on it. Item take the job of distributing cards, which makes any metrics measured on cards skewed toward actual availability of that cards on item. For example, cards exclusive on a slot will not showing up as common as another card on multiple slot, even if it is used by many people.
Data would just be less accurate. Not completely useless. Lot would certainly depend on actual algorithms used.
It would be simpler to just look at (number of times winning player used card) - (number of times losing player used card).
That would indeed be very simple, and possibly effective. I think we still want to express these things in percentages such that rarer cards that "always win" rank up there with more common cards that often win, even though they're played less often. Anyway, there are several totally viable approaches to this, which your post exemplifies.
