To be fair, I don't think it's been actually stated. The goal of measuring interaction with a clearly defined model is an objective measurement for each deck. That can then be used to roughly compare decks, evaluate the overall mix of a metagame across time, etc. At this point we're stuck with "I know it when I see it" evaluations of interaction and that quickly bogs down in arguments over the trees rather than taking in the forest.
That's why my main focus is a simulation based approach that's driven by AI, capable of playing decks against each other very quickly and generating data sets on the matchups. My end goal is to take that approach and merge it with auto generated card evaluation to solve metagames... but that's currently a long ways off.
So are you saying that the design of the Bogles deck isn't to pre-emptively interact with a potential opponent's decision tree, to reduce as many possible branches as possible?
My labels of interactive and non interactive in the previous post were referring to the pilots decisions. Lantern and Bogles are similar in that they seek to reduce the opponents ability to interact, but on the pilot side they're very different. Lantern cares about what the opponent can potentially do, Bogles doesn't. This also has nothing to do with decision trees, Affinity is a deck that's low on interactivity but very high on decision trees.
Edit: There are also many cards that do interact with Bogles, one cannot make a non interactive deck, they can only play cards that dodge the popular cards in the format. Hollow One is an example of this, their whole deck is resistant to the popular cards in the meta. If Barter in Blood were a popular card, Bogles would lose to it. Engineered Explosives can disrupt them. Settle the Wreckage gets them. And so on... there's more to it than simply trying to play cards your opponents can't answer, because given a stable metagame those answers will appear.
If you're looking at an approach that is driven by AI, then you may be interested in why minimizing the opponent's ability to interact while maximizing our own is rather relevant.
The approach Shannon suggested is today called Minimax (named after John Vonn Neuman’s minimax theorem, proven by him in 1928) and would be hugely influential for the future game-playing AIs. It is perhaps the most obvious approach one can take to making a game-playing AI. The idea is to assume both players will consider all future moves of the whole game, and so play optimally. In other words, you should always choose a move such that, even if the opponent chooses the absolute best response to that move and to every future move of yours, you will still get the highest score possible at the end of the game.
It was specifically used, and is currently used, for designing AI's to properly evaluate gamestates and decisions. I think what many people might be missing is that gamestates aren't solely determined after the opening hands are kept in Magic. They are primarily influenced, above and beyond all else, by the construction of the decks.
The approach Shannon suggested is today called Minimax (named after John Vonn Neuman’s minimax theorem, proven by him in 1928) and would be hugely influential for the future game-playing AIs. It is perhaps the most obvious approach one can take to making a game-playing AI. The idea is to assume both players will consider all future moves of the whole game, and so play optimally. In other words, you should always choose a move such that, even if the opponent chooses the absolute best response to that move and to every future move of yours, you will still get the highest score possible at the end of the game.
I've read about it, I don't think it's the right approach for Magic because variance is too high. I also think it's unrealistic to base expectations on making the best possible play because on average any given player will not make the optimal play.
Between these two factors you wind up with a situation where the AI begins to make non obvious plays that the player can't learn from and that the player won't make. You also wind up with a situation where games are played very slowly. If I remember the paper I read on this a couple years ago correctly, it was taking over an hour per game. This doesn't work in Magic because with such a small data set you can't explore all of the possible outcomes.
The approach I've taken instead, which has worked with several decks at this point is to seek to make a good move based on known information rather than simply the most optimal. While it can take upwards of a minute to determine the optimal play it only takes a small fraction of a second to determine a good one. As a result, you can make up for quality with quantity. This quantity allows for the rapid testing of multiple deck configurations and matchups. The heuristic I've been using is player clock speed with the primary goal being to keep your clock speed below your opponents, and secondary goals being to lengthen your opponents clock or speed up your own. This approach is highly effective for aggro, midrange, and control, but doesn't work for combo.
I've read about it, I don't think it's the right approach for Magic because variance is too high.
The approach I've taken instead, which has worked with several decks at this point is to seek to make a good move based on known information rather than simply the most optimal.
These particular sentences are interesting to me, because that's where we can come to appreciate the application of expectiminimax.
I understand that it's a lot of work, but despite the excessive amount of work to apply it, it seems to be the best method. Just because something is difficult doesn't mean it's not correct.
I understand that it's a lot of work, but despite the excessive amount of work to apply it, it seems to be the best method. Just because something is difficult doesn't mean it's not correct.
Processing resources are a thing. The way I like to sum it up is that Magic has something of an uncertainty principle in it. If you seek to get too detailed in your move analysis you wind up with fewer moves and fall victim to variance, making your data unreliable. If you leave out too many details in favor of more speed, you fail to capture necessary information making your data unreliable.
The best approach excepting access to a supercomputer is to strike a balance between the two. One big factor here besides variance is that Magic is always evolving, new cards enter the Modern card pool every 3 months, new cards are added to decks shifting meta balance every week. In order to have any meaningful results that can be applied to the current metagame, processing time is a very large factor.
I understand that it's a lot of work, but despite the excessive amount of work to apply it, it seems to be the best method. Just because something is difficult doesn't mean it's not correct.
Processing resources are a thing. The way I like to sum it up is that Magic has something of an uncertainty principle in it. If you seek to get too detailed in your move analysis you wind up with fewer moves and fall victim to variance, making your data unreliable. If you leave out too many details in favor of more speed, you fail to capture necessary information making your data unreliable.
The best approach excepting access to a supercomputer is to strike a balance between the two. One big factor here besides variance is that Magic is always evolving, new cards enter the Modern card pool every 3 months, new cards are added to decks shifting meta balance every week. In order to have any meaningful results that can be applied to the current metagame, processing time is a very large factor.
Did you not look at what expectiminimax is? Or read the article I linked, in which they explain how to reduce the necessary processing power required while applying minimax?
Did you not look at what expectiminimax is? Or read the article I linked, in which they explain how to reduce the necessary processing power required while applying minimax?
The approach they use relies on perfect information. Magic does not have perfect information in most situations. Not only do you not know what is in your opponents hand, but you cannot even calculate those percentages because you don't know what is in the opponents deck. As an example, lets say you're on some Eldrazi deck and your opponent is on Grixis Control. You have the option of playing a Mind Stone or a Matter Reshaper. How do you make a distinction between the two when you don't know your opponents hand, can reasonably assume they play Kolaghan's Command in their deck, but have no idea if they play Magma Spray? So how do you weigh the two cards to make a play?
...expectiminimax does not rely on perfect information relies on "perfect but incomplete information". That's why I linked it, and that's why I'm confused as to why you are saying it doesn't apply.
EDIT: To expound on this, in your example, you use expectiminimax to calculate the chances of your opponent having any specific card in their hand. No, you don't know every single card in the opponent's deck, nor do you have to. Are you at least familiar with the metagame and statistically probable decklists? If not, then we have much more basic problems to solve to improve, long before we get to game theory.
You then use that calculation to determine what is the best play to minimize the opponent's decision tree branches, while maximizing our own.
EDIT: To expound on this, in your example, you use expectiminimax to calculate the chances of your opponent having any specific card in their hand. No, you don't know every single card in the opponent's deck, nor do you have to. Are you at least familiar with the metagame and statistically probable decklists? If not, then we have much more basic problems to solve to improve, long before we get to game theory.
I'm trying to answer a much broader question. You're starting from the position that a metagame is predetermined and that there's only a handful of cards each deck plays. My end goal, is that I want to take any arbitrary draft format, set up 8 bots to draft and build optimal decks, and solve each of those formats. Something like Modern is only a stepping stone.
I want to take into account that the opponent could be playing anything that's generating positive results for them, not just statistically likely cards in any given archetype.
Ah, that sounds much easier (smaller cardpool, so less processing power to analyze all possible cards). Remember that article I linked? It sounds like you're working towards the Monte Carlo Tree Search, described in it. Did you read all three parts of the article? I know it's a long read, but I figure that if you really wanted to get the best product, you would try to acquire as much information as possible to make sure you're heading down the right path.
EDIT: @tronix I think that's a misconstrued version of what I'm saying. I'm not saying any quality that makes a card or deck "good" is a form of interaction. I think it would be more precise to say that what reduces the amount of interaction that a potential opponent might have is what makes it "good".
do you have an example of one but not the other?
Private Mod Note
():
Rollback Post to RevisionRollBack
Modern: UWGSnow-Bant Control BURGrixis Death's Shadow GWBCoCo Elves WCDeath and Taxes (sold)
Amulet being lumped into the non-interactive pile feels like a disingenuous way to make a point, and makes it difficult to take these conversations seriously after a certain while.
It's certainly, not the most interactive, as some of it's nuts draws are combo like, but you are talking about a deck that has multiple counter spells, sweepers, and dismembers in the 75, as well as EE and Ballista as control pieces, the ability to flash in Bojuka Bogs or Ghost Quarters... Bogles it is not.
Can any deck that's not U/R meet the demand for interactivity that some people are clamoring for?
EDIT: @tronix I think that's a misconstrued version of what I'm saying. I'm not saying any quality that makes a card or deck "good" is a form of interaction. I think it would be more precise to say that what reduces the amount of interaction that a potential opponent might have is what makes it "good".
do you have an example of one but not the other?
Ya. You might notice the order in which I talk about the three main concepts on which a competitive deck is built upon. I always list "minimize the amount/significance of the opponent's interaction" before "maximize it's own amount/significance of interaction with the gamestate". The mini comes before the max.
An example would be something that draws lots of cards. Sure, drawing cards is very useful. It's strictly increasing the number of options available, maximizing possible decisions. However, minimizing the opponent's options is consistently more impactful when it comes to controlling the game.
An extreme example of this, to point out how much of a difference this makes, is if we have two decks. One can draw tons of cards.. Tons of options. The other deck, however, doesn't draw extra cards, but instead makes it so the opponent cannot make any relevant choices with the cards drawn. Thus, even if the opponent draws any single card in their deck, if there is no coherent branch from having that card in the hand that leads to a winning node on the decision tree, then it just doesn't matter.
The quality of a card is primarily defined by how much it prunes branches off of the opponent's decision tree. Second to that is maximizing it's own branches (and the "health" of those branches), and third to that is whether it helps do the first two things consistently.
EDIT: I do think that it's probably important to point out - The imaginary decision trees that exist during the game are part of the gamestate, whether we can physically perceive them or not.
EDIT 2: And by "during the game", I'm including the deck design/choice phase of the game.
well the subjective, biases, and semantics is par for the course in state threads. going down this path for discussions just shines a brighter light on that.
Private Mod Note
():
Rollback Post to RevisionRollBack
Modern: UWGSnow-Bant Control BURGrixis Death's Shadow GWBCoCo Elves WCDeath and Taxes (sold)
EDIT: @tronix I think that's a misconstrued version of what I'm saying. I'm not saying any quality that makes a card or deck "good" is a form of interaction. I think it would be more precise to say that what reduces the amount of interaction that a potential opponent might have is what makes it "good".
do you have an example of one but not the other?
Ya. You might notice the order in which I talk about the three main concepts on which a competitive deck is built upon. I always list "minimize the amount/significance of the opponent's interaction" before "maximize it's own amount/significance of interaction with the gamestate". The mini comes before the max.
An example would be something that draws lots of cards. Sure, drawing cards is very useful. It's strictly increasing the number of options available, maximizing possible decisions. However, minimizing the opponent's options is consistently more impactful when it comes to controlling the game.
couldnt it be said that having more cards improves your ability to minimize your opponnents decisions?
the same goes for other strategic enablers like card selection or ramp spells. they are a part of deck construction used to strengthen your strategy, and are thus instrumental in your ability to impact the decision trees.
why is the level of impact considered when both effects have the same objective. cutting off one branch should be interaction the same way cutting off ten is.
Private Mod Note
():
Rollback Post to RevisionRollBack
Modern: UWGSnow-Bant Control BURGrixis Death's Shadow GWBCoCo Elves WCDeath and Taxes (sold)
EDIT 2: And by "during the game", I'm including the deck design/choice phase of the game.
I definitely agree on this part. My little software project has significantly changed how I approach deck construction. Deck slots are a huge resource in attempting to answer your opponents cards. Sometimes it involves something you actively want to play, other times it involves denying a cards usefulness.
When it comes to your first edit, which I didn't quote, I like to think of the branching of a decision tree as difficulty. At the start there is very little difficulty, in a typical deck your turn 1 difficulty consists of playing a land, tapped/untapped, fetch, basic/dual, tap land, and a turn 1 play. As turns advance and options increase, it branches out. However, at some point before the end of the game usually, the number of branches from that game state node begins to narrow until the difficulty falls to a large extent.
Narrowing those decision trees for either yourself or your opponent is a very valid strategy.
there's been a slight uptick in death and taxes around here, is it that good in the change in the meta lately?
also there's one kid that used to be on chord but took maybe a year or s break from it due to it not being good is back on it, with another guy building it. is there much logic to it being fine right now?
there's been a slight uptick in death and taxes around here, is it that good in the change in the meta lately?
also there's one kid that used to be on chord but took maybe a year or s break from it due to it not being good is back on it, with another guy building it. is there much logic to it being fine right now?
maybe damping sphere and Shalai, Voice of Plenty being released? i mean those decks were alright but im not sure anything in the meta points to those decks being particularly good. creature decks playing to the board arent bad vs humans, tron hasnt performed well in tournaments but its still a major player.
elves has been doing well too.
Private Mod Note
():
Rollback Post to RevisionRollBack
Modern: UWGSnow-Bant Control BURGrixis Death's Shadow GWBCoCo Elves WCDeath and Taxes (sold)
KTK, thinking about this. For simplicity should creatures (french/vanilla) just be considered to interact with 'combat' so value 1, rather than inflating the numbers saying opponents creatures, lifetotal and planeswalker?
I think we all agree that creatures are interaction but probably amongst the least interactive interactive things.
Creatures would then gain additional points for zones they interact with, opponents hand, deck, graveyard.
Would also want a negative value for cards preventing interaction eg TNN or Scion of Oona
Thoughts? (edit might be on a totally different more computationally intense plan now, oh well)
EDIT: @tronix I think that's a misconstrued version of what I'm saying. I'm not saying any quality that makes a card or deck "good" is a form of interaction. I think it would be more precise to say that what reduces the amount of interaction that a potential opponent might have is what makes it "good".
do you have an example of one but not the other?
Ya. You might notice the order in which I talk about the three main concepts on which a competitive deck is built upon. I always list "minimize the amount/significance of the opponent's interaction" before "maximize it's own amount/significance of interaction with the gamestate". The mini comes before the max.
An example would be something that draws lots of cards. Sure, drawing cards is very useful. It's strictly increasing the number of options available, maximizing possible decisions. However, minimizing the opponent's options is consistently more impactful when it comes to controlling the game.
couldnt it be said that having more cards improves your ability to minimize your opponnents decisions?
the same goes for other strategic enablers like card selection or ramp spells. they are a part of deck construction used to strengthen your strategy, and are thus instrumental in your ability to impact the decision trees.
why is the level of impact considered when both effects have the same objective. cutting off one branch should be interaction the same way cutting off ten is.
For your first question, yes, but only under specific circumstances. First, the opponent's decisions must already be minimized to the point where spending a resource (time) to cast that card and draw more cards won't allow them an opportunity to have a counterplay that just wins. Second, the success of that card that lets us draw more cards is very dependent on whether we do draw cards that continue to allow us to control the game, to continue to minimize the opponent's branches.
If you read the rest of my post that you quoted...
An extreme example of this, to point out how much of a difference this makes, is if we have two decks. One can draw tons of cards.. Tons of options. The other deck, however, doesn't draw extra cards, but instead makes it so the opponent cannot make any relevant choices with the cards drawn. Thus, even if the opponent draws any single card in their deck, if there is no coherent branch from having that card in the hand that leads to a winning node on the decision tree, then it just doesn't matter.
The quality of a card is primarily defined by how much it prunes branches off of the opponent's decision tree. Second to that is maximizing it's own branches (and the "health" of those branches), and third to that is whether it helps do the first two things consistently.
...you'll see that I already explained this. Having cards that maximize our options by pure card draw in no way minimize the opponent's actions. How does Thirst for Knowledge answer anything on it's own? There must be some other number of cards in that same deck to make Thirst for Knowledge worth it. It'd be like increasing a country's military budget, but then never spending that money on improving the military's power.
Trying to assign a numerical value to a deck based on interactivity is really an effort in futility. Say my opponent has Bolt and Push in hand. His interactive score is +2. I play a Hollow One, which is interacting with the Bolt and Push in my opponent's hand by negating their usefulness. Does this make Hollow One's interactive score +2 for me then? Perhaps -2 for my opponent? You can't assign numbers to this type of value gained in an "This card can target X things" kind of way.
Playing a card that effectively blanks your opponents cards is interaction. I am using one card in my deck to negate X cards in my opponents deck. Lightning Bolt does the same thing to Wild Nacatl, playing one card to negate an opponents card. My Primeval Titan makes all my Jund opponent's creatures bad. Is Primeval Titan +18 or so interactivity?
However,
The whole discussion smells like veiled "my midrange/control pile lost to the faster/bigger/more explosive pile and I need to quantify my rage/disappointment somewhere on the internet but people don't like outright whining" and I don't feel like there is any value for anyone to be gained by re-re-re-re-retreading the feelsbads of midrange/control players any further.
Why are their feelings given such recognition? Where are the Affinity players complaining about Pyroclasm effects or the Scapeshift players complaining about counterspell effects? The whole conversation reeks of bias and entitlement.
THIS 100X!!! If you don't agree with this then there is no amount of logic that will ever convince you that good, non-oppressive, combos should be allowed. If you don't agree with it then just don't play this game, and you certainly shouldn't feel entitled to make any comment on ban lists ever.
For your first question, yes, but only under specific circumstances. First, the opponent's decisions must already be minimized to the point where spending a resource (time) to cast that card and draw more cards won't allow them an opportunity to have a counterplay that just wins. Second, the success of that card that lets us draw more cards is very dependent on whether we do draw cards that continue to allow us to control the game, to continue to minimize the opponent's branches.
If you read the rest of my post that you quoted...
...you'll see that I already explained this. Having cards that maximize our options by pure card draw in no way minimize the opponent's actions. How does Thirst for Knowledge answer anything on it's own? There must be some other number of cards in that same deck to make Thirst for Knowledge worth it. It'd be like increasing a country's military budget, but then never spending that money on improving the military's power.
so you only count cards as interactive that have a direct impact on minimizing decisions. you brought up deck construction as a point of interacting therefore i just followed that line of reasoning. if i build my deck to be full of redundant 1 for 1 trades (decision limiters) then card selection and draw to accumulate more of those influences the decks total ability to interact by your definition. its the same concept as including more copies of a card like blood moon so you have a higher chance for it to be a factor in a game. since you are a lantern player whir is another example. as a card by itself it does nothing, but because of the other components of the deck you can use it to limit decisions. it just happens to have a 100 percent success rate whereas card draw and selection is lower.
also does your definition of interaction means how interactive a deck is variable depending on the matchup? you used the hexproof creatures example as a form of interaction, but that is contingent on the opponent caring about removing creatures. therefore bogles has basically zero interaction against storm. or is it that a card has some potential value? even if the card in application versus a matchup or metagame may not serve the intended purpose.
note that im not heckling you or anything. im just trying to draw out your reasoning because i dont agree with how you define interactivity. as i understand it right now it is too broad and contains a certain amount of circularity, but im one to keep an open mind.
Private Mod Note
():
Rollback Post to RevisionRollBack
Modern: UWGSnow-Bant Control BURGrixis Death's Shadow GWBCoCo Elves WCDeath and Taxes (sold)
To post a comment, please login or register a new account.
That's why my main focus is a simulation based approach that's driven by AI, capable of playing decks against each other very quickly and generating data sets on the matchups. My end goal is to take that approach and merge it with auto generated card evaluation to solve metagames... but that's currently a long ways off.
My labels of interactive and non interactive in the previous post were referring to the pilots decisions. Lantern and Bogles are similar in that they seek to reduce the opponents ability to interact, but on the pilot side they're very different. Lantern cares about what the opponent can potentially do, Bogles doesn't. This also has nothing to do with decision trees, Affinity is a deck that's low on interactivity but very high on decision trees.
Edit: There are also many cards that do interact with Bogles, one cannot make a non interactive deck, they can only play cards that dodge the popular cards in the format. Hollow One is an example of this, their whole deck is resistant to the popular cards in the meta. If Barter in Blood were a popular card, Bogles would lose to it. Engineered Explosives can disrupt them. Settle the Wreckage gets them. And so on... there's more to it than simply trying to play cards your opponents can't answer, because given a stable metagame those answers will appear.
Link
and specific quote:
It was specifically used, and is currently used, for designing AI's to properly evaluate gamestates and decisions. I think what many people might be missing is that gamestates aren't solely determined after the opening hands are kept in Magic. They are primarily influenced, above and beyond all else, by the construction of the decks.
Lantern Control
(with videos)
Uc Tron
Netdecking explained
Netdecking explained, Part 2
On speculators and counterfeits
On Interaction
Every single competitive deck in existence is designed to limit the opponent's ability to interact in a meaningful way.
Record number of exclamation points on SCG homepage: 71 (6 January, 2018)
"I don't want to believe, I want to know."
-Carl Sagan
I've read about it, I don't think it's the right approach for Magic because variance is too high. I also think it's unrealistic to base expectations on making the best possible play because on average any given player will not make the optimal play.
Between these two factors you wind up with a situation where the AI begins to make non obvious plays that the player can't learn from and that the player won't make. You also wind up with a situation where games are played very slowly. If I remember the paper I read on this a couple years ago correctly, it was taking over an hour per game. This doesn't work in Magic because with such a small data set you can't explore all of the possible outcomes.
The approach I've taken instead, which has worked with several decks at this point is to seek to make a good move based on known information rather than simply the most optimal. While it can take upwards of a minute to determine the optimal play it only takes a small fraction of a second to determine a good one. As a result, you can make up for quality with quantity. This quantity allows for the rapid testing of multiple deck configurations and matchups. The heuristic I've been using is player clock speed with the primary goal being to keep your clock speed below your opponents, and secondary goals being to lengthen your opponents clock or speed up your own. This approach is highly effective for aggro, midrange, and control, but doesn't work for combo.
These particular sentences are interesting to me, because that's where we can come to appreciate the application of expectiminimax.
I understand that it's a lot of work, but despite the excessive amount of work to apply it, it seems to be the best method. Just because something is difficult doesn't mean it's not correct.
Lantern Control
(with videos)
Uc Tron
Netdecking explained
Netdecking explained, Part 2
On speculators and counterfeits
On Interaction
Every single competitive deck in existence is designed to limit the opponent's ability to interact in a meaningful way.
Record number of exclamation points on SCG homepage: 71 (6 January, 2018)
"I don't want to believe, I want to know."
-Carl Sagan
Processing resources are a thing. The way I like to sum it up is that Magic has something of an uncertainty principle in it. If you seek to get too detailed in your move analysis you wind up with fewer moves and fall victim to variance, making your data unreliable. If you leave out too many details in favor of more speed, you fail to capture necessary information making your data unreliable.
The best approach excepting access to a supercomputer is to strike a balance between the two. One big factor here besides variance is that Magic is always evolving, new cards enter the Modern card pool every 3 months, new cards are added to decks shifting meta balance every week. In order to have any meaningful results that can be applied to the current metagame, processing time is a very large factor.
Did you not look at what expectiminimax is? Or read the article I linked, in which they explain how to reduce the necessary processing power required while applying minimax?
Lantern Control
(with videos)
Uc Tron
Netdecking explained
Netdecking explained, Part 2
On speculators and counterfeits
On Interaction
Every single competitive deck in existence is designed to limit the opponent's ability to interact in a meaningful way.
Record number of exclamation points on SCG homepage: 71 (6 January, 2018)
"I don't want to believe, I want to know."
-Carl Sagan
The approach they use relies on perfect information. Magic does not have perfect information in most situations. Not only do you not know what is in your opponents hand, but you cannot even calculate those percentages because you don't know what is in the opponents deck. As an example, lets say you're on some Eldrazi deck and your opponent is on Grixis Control. You have the option of playing a Mind Stone or a Matter Reshaper. How do you make a distinction between the two when you don't know your opponents hand, can reasonably assume they play Kolaghan's Command in their deck, but have no idea if they play Magma Spray? So how do you weigh the two cards to make a play?
does not rely on perfect informationrelies on "perfect but incomplete information". That's why I linked it, and that's why I'm confused as to why you are saying it doesn't apply.EDIT: To expound on this, in your example, you use expectiminimax to calculate the chances of your opponent having any specific card in their hand. No, you don't know every single card in the opponent's deck, nor do you have to. Are you at least familiar with the metagame and statistically probable decklists? If not, then we have much more basic problems to solve to improve, long before we get to game theory.
You then use that calculation to determine what is the best play to minimize the opponent's decision tree branches, while maximizing our own.
Lantern Control
(with videos)
Uc Tron
Netdecking explained
Netdecking explained, Part 2
On speculators and counterfeits
On Interaction
Every single competitive deck in existence is designed to limit the opponent's ability to interact in a meaningful way.
Record number of exclamation points on SCG homepage: 71 (6 January, 2018)
"I don't want to believe, I want to know."
-Carl Sagan
I'm trying to answer a much broader question. You're starting from the position that a metagame is predetermined and that there's only a handful of cards each deck plays. My end goal, is that I want to take any arbitrary draft format, set up 8 bots to draft and build optimal decks, and solve each of those formats. Something like Modern is only a stepping stone.
I want to take into account that the opponent could be playing anything that's generating positive results for them, not just statistically likely cards in any given archetype.
Lantern Control
(with videos)
Uc Tron
Netdecking explained
Netdecking explained, Part 2
On speculators and counterfeits
On Interaction
Every single competitive deck in existence is designed to limit the opponent's ability to interact in a meaningful way.
Record number of exclamation points on SCG homepage: 71 (6 January, 2018)
"I don't want to believe, I want to know."
-Carl Sagan
do you have an example of one but not the other?
UWGSnow-Bant Control
BURGrixis Death's Shadow
GWBCoCo Elves
WCDeath and Taxes(sold)It's certainly, not the most interactive, as some of it's nuts draws are combo like, but you are talking about a deck that has multiple counter spells, sweepers, and dismembers in the 75, as well as EE and Ballista as control pieces, the ability to flash in Bojuka Bogs or Ghost Quarters... Bogles it is not.
Can any deck that's not U/R meet the demand for interactivity that some people are clamoring for?
Ya. You might notice the order in which I talk about the three main concepts on which a competitive deck is built upon. I always list "minimize the amount/significance of the opponent's interaction" before "maximize it's own amount/significance of interaction with the gamestate". The mini comes before the max.
An example would be something that draws lots of cards. Sure, drawing cards is very useful. It's strictly increasing the number of options available, maximizing possible decisions. However, minimizing the opponent's options is consistently more impactful when it comes to controlling the game.
An extreme example of this, to point out how much of a difference this makes, is if we have two decks. One can draw tons of cards.. Tons of options. The other deck, however, doesn't draw extra cards, but instead makes it so the opponent cannot make any relevant choices with the cards drawn. Thus, even if the opponent draws any single card in their deck, if there is no coherent branch from having that card in the hand that leads to a winning node on the decision tree, then it just doesn't matter.
The quality of a card is primarily defined by how much it prunes branches off of the opponent's decision tree. Second to that is maximizing it's own branches (and the "health" of those branches), and third to that is whether it helps do the first two things consistently.
EDIT: I do think that it's probably important to point out - The imaginary decision trees that exist during the game are part of the gamestate, whether we can physically perceive them or not.
EDIT 2: And by "during the game", I'm including the deck design/choice phase of the game.
Lantern Control
(with videos)
Uc Tron
Netdecking explained
Netdecking explained, Part 2
On speculators and counterfeits
On Interaction
Every single competitive deck in existence is designed to limit the opponent's ability to interact in a meaningful way.
Record number of exclamation points on SCG homepage: 71 (6 January, 2018)
"I don't want to believe, I want to know."
-Carl Sagan
UWGSnow-Bant Control
BURGrixis Death's Shadow
GWBCoCo Elves
WCDeath and Taxes(sold)couldnt it be said that having more cards improves your ability to minimize your opponnents decisions?
the same goes for other strategic enablers like card selection or ramp spells. they are a part of deck construction used to strengthen your strategy, and are thus instrumental in your ability to impact the decision trees.
why is the level of impact considered when both effects have the same objective. cutting off one branch should be interaction the same way cutting off ten is.
UWGSnow-Bant Control
BURGrixis Death's Shadow
GWBCoCo Elves
WCDeath and Taxes(sold)I definitely agree on this part. My little software project has significantly changed how I approach deck construction. Deck slots are a huge resource in attempting to answer your opponents cards. Sometimes it involves something you actively want to play, other times it involves denying a cards usefulness.
When it comes to your first edit, which I didn't quote, I like to think of the branching of a decision tree as difficulty. At the start there is very little difficulty, in a typical deck your turn 1 difficulty consists of playing a land, tapped/untapped, fetch, basic/dual, tap land, and a turn 1 play. As turns advance and options increase, it branches out. However, at some point before the end of the game usually, the number of branches from that game state node begins to narrow until the difficulty falls to a large extent.
Narrowing those decision trees for either yourself or your opponent is a very valid strategy.
also there's one kid that used to be on chord but took maybe a year or s break from it due to it not being good is back on it, with another guy building it. is there much logic to it being fine right now?
maybe damping sphere and Shalai, Voice of Plenty being released? i mean those decks were alright but im not sure anything in the meta points to those decks being particularly good. creature decks playing to the board arent bad vs humans, tron hasnt performed well in tournaments but its still a major player.
elves has been doing well too.
UWGSnow-Bant Control
BURGrixis Death's Shadow
GWBCoCo Elves
WCDeath and Taxes(sold)Faithless Looting is the Brainstorm of Modern.
Modern: Storm
Legacy: ANT
UWGSnow-Bant Control
BURGrixis Death's Shadow
GWBCoCo Elves
WCDeath and Taxes(sold)I think we all agree that creatures are interaction but probably amongst the least interactive interactive things.
Creatures would then gain additional points for zones they interact with, opponents hand, deck, graveyard.
Would also want a negative value for cards preventing interaction eg TNN or Scion of Oona
Thoughts? (edit might be on a totally different more computationally intense plan now, oh well)
Legacy - LED Dredge, ANT & WDnT
For your first question, yes, but only under specific circumstances. First, the opponent's decisions must already be minimized to the point where spending a resource (time) to cast that card and draw more cards won't allow them an opportunity to have a counterplay that just wins. Second, the success of that card that lets us draw more cards is very dependent on whether we do draw cards that continue to allow us to control the game, to continue to minimize the opponent's branches.
If you read the rest of my post that you quoted...
...you'll see that I already explained this. Having cards that maximize our options by pure card draw in no way minimize the opponent's actions. How does Thirst for Knowledge answer anything on it's own? There must be some other number of cards in that same deck to make Thirst for Knowledge worth it. It'd be like increasing a country's military budget, but then never spending that money on improving the military's power.
Lantern Control
(with videos)
Uc Tron
Netdecking explained
Netdecking explained, Part 2
On speculators and counterfeits
On Interaction
Every single competitive deck in existence is designed to limit the opponent's ability to interact in a meaningful way.
Record number of exclamation points on SCG homepage: 71 (6 January, 2018)
"I don't want to believe, I want to know."
-Carl Sagan
Trying to assign a numerical value to a deck based on interactivity is really an effort in futility. Say my opponent has Bolt and Push in hand. His interactive score is +2. I play a Hollow One, which is interacting with the Bolt and Push in my opponent's hand by negating their usefulness. Does this make Hollow One's interactive score +2 for me then? Perhaps -2 for my opponent? You can't assign numbers to this type of value gained in an "This card can target X things" kind of way.
Playing a card that effectively blanks your opponents cards is interaction. I am using one card in my deck to negate X cards in my opponents deck. Lightning Bolt does the same thing to Wild Nacatl, playing one card to negate an opponents card. My Primeval Titan makes all my Jund opponent's creatures bad. Is Primeval Titan +18 or so interactivity?
However,
The whole discussion smells like veiled "my midrange/control pile lost to the faster/bigger/more explosive pile and I need to quantify my rage/disappointment somewhere on the internet but people don't like outright whining" and I don't feel like there is any value for anyone to be gained by re-re-re-re-retreading the feelsbads of midrange/control players any further.
Why are their feelings given such recognition? Where are the Affinity players complaining about Pyroclasm effects or the Scapeshift players complaining about counterspell effects? The whole conversation reeks of bias and entitlement.
so you only count cards as interactive that have a direct impact on minimizing decisions. you brought up deck construction as a point of interacting therefore i just followed that line of reasoning. if i build my deck to be full of redundant 1 for 1 trades (decision limiters) then card selection and draw to accumulate more of those influences the decks total ability to interact by your definition. its the same concept as including more copies of a card like blood moon so you have a higher chance for it to be a factor in a game. since you are a lantern player whir is another example. as a card by itself it does nothing, but because of the other components of the deck you can use it to limit decisions. it just happens to have a 100 percent success rate whereas card draw and selection is lower.
also does your definition of interaction means how interactive a deck is variable depending on the matchup? you used the hexproof creatures example as a form of interaction, but that is contingent on the opponent caring about removing creatures. therefore bogles has basically zero interaction against storm. or is it that a card has some potential value? even if the card in application versus a matchup or metagame may not serve the intended purpose.
note that im not heckling you or anything. im just trying to draw out your reasoning because i dont agree with how you define interactivity. as i understand it right now it is too broad and contains a certain amount of circularity, but im one to keep an open mind.
UWGSnow-Bant Control
BURGrixis Death's Shadow
GWBCoCo Elves
WCDeath and Taxes(sold)