Currently, I make 16 card packs, and have everyone toss the 16th card in an 'undrafted' box, and I note that on my spreadsheet. Each time a card goes undrafted it gets a 'point'. Getting pointed doesn't make it auto-removed, but it does give me a chance to see what things are being ignored.
thread bump. i searched the forum for threads titled "data" and this was the most recent one i could find.
cube building seems to be a blend of science and art. we can make lists and calculations on mana curves and creature counts, and we will always have pet cards that have great history and value in our cube even if no one else seems to run them. what i'm wondering is if anyone has found a great method to determine what card to cut when you finally acquire the sick foil mythic from the latest set, or that $400 p3k card for your cube. i'm interested in hearing any and all ideas that people use (or being directed to a compilation list if one exists).
first, some comments:
1) i'm aware cubetutor.com exists and i highly value it as a resource for data collection. while it's possible to go there and look through your cube's lowest draft picks, i think that data is going to be somewhat skewed and shouldn't be the only resource for making cuts. for most of us who don't have featured cubes and aren't wtwlf, our cubetutor cubes are mostly drafted by ourselves and friends who have nothing better to do at work. while the new castle ai is going to help somewhat, these draft numbers still might be skewed towards our personal preferences instead of what are objectively the weakest/least cohesive elements of our lists.
2) i'm aware that there probably isn't one exact best method of coming up with what cards to cut, and that there are certain methods that work better for certain cubes/managers. i'm interested in hearing ideas and counterpoints, not so much in finding one solution for perfect cube management.
3) there are other factors to take into account besides what is objectively the "worst" card in a list and cutting it. maybe someday goblin guide is the worst red one drop in your cube, but cutting it takes too much away from red's aggro suite. i get that there needs to be some finesse to the science, but i'd still like to figure out if there's a way to at least narrow down to 5 or so cards that maybe gives a range of costs and abilities within a color to help make the tough cuts.
on to the ideas:
one method i've heard of is to throw all of the 15th picks into a pile and keep a tally of all last picks. when it's time to make a cut, look at your cards that are last picked the most frequently. this idea makes some sense, but i think there are some problems with it. first, the list would need to reset every time you add a new card to the cube, because it would start at 0 last picks and automatically have an edge over some of the older cards. second, there are days when nobody wants to draft certain archetypes, and on those days cards that might be great for that archetype might fall down the pick list even though they are superb cards for that archetype. third, there are sometimes packs that have 5 red cards in them, and only a couple people at the table are playing red. this could cause certain cards to be last picks again with no consideration of the actual merit of the card. those are just some of my thoughts, and i don't personally use this method.
the method i use is as follows. after playing for the night, i have everyone keep together their maindeck and sideboard, and write on a piece of paper their name, record for the night, and name of their deck (r/w aggro, ub reanimator, whatever). then i go through the deck and keep a running tally of each card's wins and losses (which gives each card a win % over time) and whether the card was maindeck or sideboard (cards that were drafted but not on color/theme for the deck i leave off, because it's possible they were hatedrafted, picked late or picked early and abandoned and i don't want that to affect their score). then i multiply the maindeck % by a multiplier (60 at the moment) and the win % by a multiplier (40 at the moment) and add the two scores together for a final card score. so a card like sol ring may have a score of 80 ((100% md x60) + (50% winrate x40)=80).
i'll start out by saying i don't get to live draft my cube as often as i'd like, and it's a 720 card cube, so i don't have anywhere near enough data to come up with any conclusions on if this method delivers any productive data. it's been years since i've taken a college level statistics class, so if there is no way to make any results relevant, please feel free to let me know so i can stop wasting my time and move on to something else. i'm here to learn from other people's ideas.
below are some of the things i'd hope to accomplish with the data i've collected:
find out what an "average" score is for a card in my cube. i'm thinking 65 would be a good average. ((.75 md% x60)+(.50 win% x40))=65. This would mean the card is maindecked in about 75% of decks that play the card's color, and wins about half the time. using the same card score of 65, a card that is only played in ~48% of the decks it's played in would have to win 90% of its matches to come out with the same score.
this card score would not need to reset every time new cards come and go from the cube. older cards would have more data, but a newer card could shoot to the top or bottom of the score table based on a few drafts. i'm also able to see which cards have the least amount of total games played, to avoid cutting newer cards that haven't had a chance or focus on cards that have been in the cube forever but never seem to go into any decks.
maybe weed out some cards that are "too good" for my environment. i run an unpowered 720 with no planeswalkers but include the card sol ring. many people agree that sol ring is too good for unpowered lists, and it's certainly a first pick a vast majority of the time around here. but i'm more comfortable running sol ring if the decks it's played in are only winning 50-60% of matches rather than 85-95% of matches. this method would allow me to see over time if any cards really are a power-level concern (aside from anecdotal evidence from drafters that card x is broken).
i also feel that adding win% to the equation might help shine some light on cards that are strong but underdrafted. for example, in my playgroup, no one ever seems to pick psychatog. i don't personally like it either, but many cube builders on the forums argue vehemently for its power level. it's possible that it goes undrafted and unplayed in my cube for a while, but when a player outside the normal group, or someone looking to experiment with cards they don't normally pick wins with it a few times, maybe the win% keeps its score high enough for a while that others get a chance to see it in action and figure out how to win with it.
in combination with the card score, i use the other information recorded (player name, deck name, win/loss record) to keep another sheet with player profiles and deck profiles. this lets me see what decks and colors are winning/losing so i can try to keep an eye on archetypes that may need more or less support. and it helps us brag about who has the best overall record.
anyway, these are some of the things i think about when managing my list, and i'm interested both in hearing what others do for data collection and hearing why my own method is terrible/brilliant/has potential with tweaking of the formula/data collection/etc.
Performance is 10x more important than draft % or win %. Draft percentages can be effected by what's open, what's the right deck to draft in that particular draft, or simply player preferences. Win % has more to do with the deck than the individual cards, and it doesn't give you any context. A card could have a low win %, but it could've been the best performing card in a series of losing decks, and get unjustly punished for it. Without an ungodly amount of data (like years worth of draft and win percentage data) the correlation won't show itself to be of any use. And even then, it's still largely effected by the biases of the playgroup and the skill of the players wielding the cards.
The only information that really matters is how the card performs. When people did decide to include it in the final 40 over other options, how did it perform in action. Feedback from your players is the only way to make informed decisions about what to add and cut from your cube. Since the draft % and win % is going to be largely based on the preferences of the playgroup anyways, you might as well base your cuts on the only concrete information you have; which is if the card was good or not for your players when they used it.
Edit: And as an aside, despite the high draft numbers on my CubeTutor list, I would never, ever base a cut on the pick/pass ratio in that data. I think it's garbage information.
the method i use is as follows. after playing for the night, i have everyone keep together their maindeck and sideboard, and write on a piece of paper their name, record for the night, and name of their deck (r/w aggro, ub reanimator, whatever). then i go through the deck and keep a running tally of each card's wins and losses (which gives each card a win % over time) and whether the card was maindeck or sideboard (cards that were drafted but not on color/theme for the deck i leave off, because it's possible they were hatedrafted, picked late or picked early and abandoned and i don't want that to affect their score). then i multiply the maindeck % by a multiplier (60 at the moment) and the win % by a multiplier (40 at the moment) and add the two scores together for a final card score. so a card like sol ring may have a score of 80 ((100% md x60) + (50% winrate x40)=80).
Currently, I make 16 card packs, and have everyone toss the 16th card in an 'undrafted' box, and I note that on my spreadsheet. Each time a card goes undrafted it gets a 'point'. Getting pointed doesn't make it auto-removed, but it does give me a chance to see what things are being ignored.
cube building seems to be a blend of science and art. we can make lists and calculations on mana curves and creature counts, and we will always have pet cards that have great history and value in our cube even if no one else seems to run them. what i'm wondering is if anyone has found a great method to determine what card to cut when you finally acquire the sick foil mythic from the latest set, or that $400 p3k card for your cube. i'm interested in hearing any and all ideas that people use (or being directed to a compilation list if one exists).
first, some comments:
1) i'm aware cubetutor.com exists and i highly value it as a resource for data collection. while it's possible to go there and look through your cube's lowest draft picks, i think that data is going to be somewhat skewed and shouldn't be the only resource for making cuts. for most of us who don't have featured cubes and aren't wtwlf, our cubetutor cubes are mostly drafted by ourselves and friends who have nothing better to do at work. while the new castle ai is going to help somewhat, these draft numbers still might be skewed towards our personal preferences instead of what are objectively the weakest/least cohesive elements of our lists.
2) i'm aware that there probably isn't one exact best method of coming up with what cards to cut, and that there are certain methods that work better for certain cubes/managers. i'm interested in hearing ideas and counterpoints, not so much in finding one solution for perfect cube management.
3) there are other factors to take into account besides what is objectively the "worst" card in a list and cutting it. maybe someday goblin guide is the worst red one drop in your cube, but cutting it takes too much away from red's aggro suite. i get that there needs to be some finesse to the science, but i'd still like to figure out if there's a way to at least narrow down to 5 or so cards that maybe gives a range of costs and abilities within a color to help make the tough cuts.
on to the ideas:
one method i've heard of is to throw all of the 15th picks into a pile and keep a tally of all last picks. when it's time to make a cut, look at your cards that are last picked the most frequently. this idea makes some sense, but i think there are some problems with it. first, the list would need to reset every time you add a new card to the cube, because it would start at 0 last picks and automatically have an edge over some of the older cards. second, there are days when nobody wants to draft certain archetypes, and on those days cards that might be great for that archetype might fall down the pick list even though they are superb cards for that archetype. third, there are sometimes packs that have 5 red cards in them, and only a couple people at the table are playing red. this could cause certain cards to be last picks again with no consideration of the actual merit of the card. those are just some of my thoughts, and i don't personally use this method.
the method i use is as follows. after playing for the night, i have everyone keep together their maindeck and sideboard, and write on a piece of paper their name, record for the night, and name of their deck (r/w aggro, ub reanimator, whatever). then i go through the deck and keep a running tally of each card's wins and losses (which gives each card a win % over time) and whether the card was maindeck or sideboard (cards that were drafted but not on color/theme for the deck i leave off, because it's possible they were hatedrafted, picked late or picked early and abandoned and i don't want that to affect their score). then i multiply the maindeck % by a multiplier (60 at the moment) and the win % by a multiplier (40 at the moment) and add the two scores together for a final card score. so a card like sol ring may have a score of 80 ((100% md x60) + (50% winrate x40)=80).
i'll start out by saying i don't get to live draft my cube as often as i'd like, and it's a 720 card cube, so i don't have anywhere near enough data to come up with any conclusions on if this method delivers any productive data. it's been years since i've taken a college level statistics class, so if there is no way to make any results relevant, please feel free to let me know so i can stop wasting my time and move on to something else. i'm here to learn from other people's ideas.
below are some of the things i'd hope to accomplish with the data i've collected:
find out what an "average" score is for a card in my cube. i'm thinking 65 would be a good average. ((.75 md% x60)+(.50 win% x40))=65. This would mean the card is maindecked in about 75% of decks that play the card's color, and wins about half the time. using the same card score of 65, a card that is only played in ~48% of the decks it's played in would have to win 90% of its matches to come out with the same score.
this card score would not need to reset every time new cards come and go from the cube. older cards would have more data, but a newer card could shoot to the top or bottom of the score table based on a few drafts. i'm also able to see which cards have the least amount of total games played, to avoid cutting newer cards that haven't had a chance or focus on cards that have been in the cube forever but never seem to go into any decks.
maybe weed out some cards that are "too good" for my environment. i run an unpowered 720 with no planeswalkers but include the card sol ring. many people agree that sol ring is too good for unpowered lists, and it's certainly a first pick a vast majority of the time around here. but i'm more comfortable running sol ring if the decks it's played in are only winning 50-60% of matches rather than 85-95% of matches. this method would allow me to see over time if any cards really are a power-level concern (aside from anecdotal evidence from drafters that card x is broken).
i also feel that adding win% to the equation might help shine some light on cards that are strong but underdrafted. for example, in my playgroup, no one ever seems to pick psychatog. i don't personally like it either, but many cube builders on the forums argue vehemently for its power level. it's possible that it goes undrafted and unplayed in my cube for a while, but when a player outside the normal group, or someone looking to experiment with cards they don't normally pick wins with it a few times, maybe the win% keeps its score high enough for a while that others get a chance to see it in action and figure out how to win with it.
in combination with the card score, i use the other information recorded (player name, deck name, win/loss record) to keep another sheet with player profiles and deck profiles. this lets me see what decks and colors are winning/losing so i can try to keep an eye on archetypes that may need more or less support. and it helps us brag about who has the best overall record.
anyway, these are some of the things i think about when managing my list, and i'm interested both in hearing what others do for data collection and hearing why my own method is terrible/brilliant/has potential with tweaking of the formula/data collection/etc.
happy cubing
The only information that really matters is how the card performs. When people did decide to include it in the final 40 over other options, how did it perform in action. Feedback from your players is the only way to make informed decisions about what to add and cut from your cube. Since the draft % and win % is going to be largely based on the preferences of the playgroup anyways, you might as well base your cuts on the only concrete information you have; which is if the card was good or not for your players when they used it.
Edit: And as an aside, despite the high draft numbers on my CubeTutor list, I would never, ever base a cut on the pick/pass ratio in that data. I think it's garbage information.
My 630 Card Powered Cube
My Article - "Cube Design Philosophy"
My Article - "Mana Short: A study in limited resource management."
My 50th Set (P)review - Discusses my top 20 Cube cards from OTJ!
I wonder where this idea came from.
I used to write cube articles on StarCityGames, now for GatheringMagic and podcast about cube (w/Antknee42.)