Going back a few pages, I want to bring up a discussion we had a couple months ago.
http://www.mtgthesource.com/forums/s...=1#post1034448
As a quick refresher, the basic idea is that we could rate cards as good or bad in our deck based on what they interact with, or what they don't interact with.
A similar discussion is being had on MTGS right now for some Modern ratings, and I took the opportunity to use the discussion alongside some free time I had to try a test run of this system.
Essentially, the way it works is that each card is rated against each other card through a series of rules, the card in the column is rated by the card in the row, and at their intersection point in the grid we rate 1, 0, or -1 (still not sure on the -1). A 1 indicates the card matches up favorably according to the rules and a 0 means it doesn't.
For this test case I took the current top 40 most played cards in Modern and excluded the lands. That left me with 21 cards, which is 441 pieces of data. It took an average of 8 seconds per card to manually rate these (which is partially why I would like to build an automated system to do it).
My notes on this weren't the most legible but here's the copy/paste of the rules I used
Ratings rate first cast of each spell, at earliest opportunity.
Creatures:
Successfully fights (trades in combat, bounces in combat, wins in combat... does not die for nothing)
Edge cases - Fight exception If CA is gained (Wall of Omens example), variable p/t (ooze, goyf, etc) given arbitrary values, I'm using 4/5 goyf, 4/4 ooze
Removal:
Kills at target CMC+1 or less. (Terminate passes vs Birds of Paradise, Hero's Downfall does not)
Corner cases – Delve fails vs CMC of 1. Delve creatures assumed to fully delve, conditional enablers (metalcraft, revolt, etc) always assumed on. Must trade for full card (Electrolyze passes Lingering Souls, beats one side and draws another, Kolaghan's Command does not, fatal push loses to BBE), treat counterspells as removal
Discard:
Opponents choice discard never beats a card
Player choice discard must within choice cmc+1
Trading:
Use traditional CA counts, draw = +1, discard = -1, pass if parity or better (Cryptic Command hits parity against BBE, so it passes). Edge case, 2 for 1 passes at mana+3 vs mana +1. Additional cards gained such as 3 for 1 pass at mana +2 for each additional card.
Planeswalker:
May evaluate all modes usable on ETB, if any usable mode passes vs a given card, the card passes
No interaction:
If no interaction between cards (Vial vs Lighting Bolt), result is a -1.
So what people are probably wondering is... why mention all this Modern stuff on a thread for a Legacy deck? And my answer to that is that I want to build something similar for Nic Fit, or perhaps the format as a whole. But, in order to get it right, it's important to make a proper rating system. Such a rating system needs to not be subjective, and follow clearly defined rules so that everyone gets the same results when rating cards.
Here's the initial results I got
Raw scores
https://imgur.com/a/ItrujFZ
No -1's
https://imgur.com/a/xBiPzah
From this initial set of results, I've found that my ruleset is currently incomplete. The biggest question is what the proper rating is for two cards that totally ignore each other like Aether Vial vs Lightning Bolt. It seems incorrect to me to label that as an unfavorable interaction, so I was using -1 essentially as a null value. Introducing three values to the system: good, bad, and none. But I can also see an argument to rate with two values: good and other than good.
The other issue is that cards need to follow different rule sets based on the objective of a card. So a card would need to be rated both offensively and defensively under separate rules. What I'm coming up with so far is that cards need to be categorized as threat (wins the game), answer (stops opponent from winning), or enabler (makes another card more effective). Can anyone think of something that wouldn't fall into one of those three categories?
