It purely depends on the deck you want to play them. In some decks, digging deeper and being a Sorcery spell is more important than the (worse) Instant speed Opt.
However, I in general I see Vision > Opt and Opt = Sleight of Hands.
Greetings,
Kathal
Private Mod Note
():
Rollback Post to RevisionRollBack
What I play or have:
Modern/Legacy
either funpolice (Delver, Deathcloud, UW Control) or the fun decks (especially those ft. Griselbrand)
It solely depends on the Control deck. For example, UW Control would rather play Visions, cause of the rather awkward mana curve the deck has and thus wants to hit its silver bullets more reliant. On the other hand, Esper Control might want Opt more, cause the deck is a classic draw and go one.
Greetings,
Kathal
Private Mod Note
():
Rollback Post to RevisionRollBack
What I play or have:
Modern/Legacy
either funpolice (Delver, Deathcloud, UW Control) or the fun decks (especially those ft. Griselbrand)
Serum Visions is better than Opt. Most of the time you have at least one draw step to spare (i.e. you're not just going to die on your opponent's turn). If you're digging for something important, SV pushes up to three irrelevant cards out of the way, versus Opt or Sleight of Hand's two.
Opt vs Sleight of Hand is a more pertinent question. I'll make some simplifying assumptions for the following analysis. Suppose your deck is comprised solely of good cards with value 1 and bad cards with value 0, and a random draw (without knowing if it's good or bad) has probability 0<p<1 of having value 1 and 1-p of having value 0. Also suppose the following sensible strategies for playing Opt and Sleight: with Opt, if the scryed card has value 1, you leave it on top and proceed to draw it, otherwise you bottom it and draw a random card. With Sleight, you always take a value 1 card if you see it, otherwise you take either of the two 0s.
Consider all possible permutations of good/bad cards in the top two. Additionally, like with Serum Visions, I'll assume that you have one draw step to spare. This assumption has the effect of leveling out the field - no matter what the top 2 cards were, what you chose to do with the cantrip, or even which cantrip it was, after cantrip + draw, the very next card you draw is random. Example: if both cards on top are 1s, you play Opt, see a 1, keep it on top, and draw it with Opt. Then your draw step yields another 1, and the one after that yields a random card with EV p. For each permutation, what is the probability of it happening, and what is the value of Opt+draw compared to Sleight+draw?
Opt + draw
11: prob p^2, value 2
10: prob p(1-p), value 1
01: prob p(1-p), value 1+p
00: prob (1-p)^2, value p
Sleight + draw
11: prob p^2, value 1+p
10: prob p(1-p), value 1+p
01: prob p(1-p), value 1+p
00: prob (1-p)^2, value p
The expected value calculation shows that both cantrips are dead even for any p. However, note that Sleight is more consistent. With Opt you have two events (11, 10) that have values 1 and 2, whereas with Sleight, the same events have value 1+p each. You won't hit the max of 2 with Sleight, but you won't hit the low of 1 either.
Opt does have one advantage which the above analysis doesn't put a number on: being an instant. A deck with many instant-speed counters/removal spells can pass the turn without committing to anything, then killing whatever the opponent plays, or cantripping with Opt if they don't play anything worth killing.
In short: Sleight and Opt are really close, close enough that although the numbers say Sleight is better (in the second-order stochastic sense), the outcome could easily be reversed if you made some other sensible assumptions, or simply if you value instant speed more than lower variance. Serum Visions is still the boss though.
Exercise for the reader: perform the above analysis for Serum Visions vs Opt or Serum Visions vs Sleight of Hand, but considering two draw steps and the top three cards instead (can you see why?). What strategy do you use for Serum Visions?
Yeah, Opt bring stiff competition with Serum Visions, but I don't think it'll replace it.
Opt is at its best in more draw-go, controlling, no-silver-bullets decks like UWR variants or UW Control. That's a fair few decks already.
Serum Visions is still better at digging for specific cards than Opt is, as it goes through 3 cards instead of 2. This means I'd go for the full playset of Serum Visions first in combo decks, Blue Moon (digging for Blood Moon and Roast-level red removal is that good), likely Scapeshift (I aggressively keep 2-land hands with a Serum Visions in them, and I'm not sure I'd do the same if Serum Visions were Opt instead), and tempo decks such as Grixis Death's Shadow (digging for creatures is swell).
Opt might not even unseat Sleight of Hand. Assuming more than half your deck is dreck at any given time (so in izzetmage's example, p < 0.5), the chance that you bottom a bad card and make your next draw a teeny bit better is greater with Sleight than with Opt. After all, if the top card is good, you picked it with Opt and didn't bother checking whether the card second from the top is bad.
With that being said, unless self-milling (or milling your opponent just in case you're still playing Mill) brings significant benefits (such as Delve or Lingering Souls), I think Opt replaces Thought Scour.
With that being said, unless self-milling (or milling your opponent just in case you're still playing Mill) brings significant benefits (such as Delve or Lingering Souls), I think Opt replaces Thought Scour.
What deck is running Thought Scourt that isn't also designed to utilize the graveyard?
With that being said, unless self-milling (or milling your opponent just in case you're still playing Mill) brings significant benefits (such as Delve or Lingering Souls), I think Opt replaces Thought Scour.
What deck is running Thought Scourt that isn't also designed to utilize the graveyard?
Some versions of storm were running Thought Scour. And while Storm does use the graveyard, it's probably better for that deck to replace it with Opt as the card selection is far more valuable than milling over 2 unknowns.
SV > opt > slight
Pretty much is how it's going to go. There are a few cases where opt could be better than SV but that's going to be rare. Opt is better than SV in the early turns in control builds, but mid to late game you want to be casting SV. So just keep your SV in there and maybe take something else out for opt.
I'd actually rather have peek over opt btw. The information you get from peek is better than scry 1.
Opt seems good in flash creature decks. Clique, Spell Queller, Remand and such are cards to play with opt. Every other deck likely sticks with sleight or thought scour.
Opt is at its best in more draw-go, controlling, no-silver-bullets decks like UWR variants or UW Control. That's a fair few decks already.
In those decks I think Opt has to fight Think Twice to even get above 0 copies. Pure control doesn't even play 4 Serum Visions all the time.
If you're talking about UWx, I would venture to agree with Jordan Boisvert that most competitive decks run 4x Serum Visions and have come around to the strength of the card, and most lists that don't aren't optimized.
As far as Opt vs Serum Visions, I plan on testing a playset of both in UWR/Jeskai Control just to see if the turbo xerox rule can hold up without brainstorm/ponder (cutting 2 lands and 2 spells for the 4 opt, for a total of 22 lands and 8 hard cantrips + 2 remand/2 electrolyze). Should be interesting.
Is Opt the new default cantrip? I can only see Serum Visions being Opt 5-8 now, or do you think I'm wrong?
Serum Visions is better than Opt. Opt is nearly unplayable. The amount a card digs matters a lot. Dig 1 cantrips are unplayable (even though some people still play them), Dig 2's have corner case uses but are still a hard sell (see Sleight of Hand). Opt is effectively a dig 2. Few decks are going to want it. In order for a cantrip to really pass the bar, it needs to be a dig 3 like Serum Visions is.
The amount a cantrip digs is basically the most important aspect of it. Instant/Sorcery matters little.
Exercise for the reader: perform the above analysis for Serum Visions vs Opt or Serum Visions vs Sleight of Hand, but considering two draw steps and the top three cards instead (can you see why?). What strategy do you use for Serum Visions?
I would do it myself, but I don't have the time right now. Can you perform this analysis for Serum Visions vs Preordain? All the logic I've looked at says that Preordain should be the better card, but every time I've run it through my MTG simulator, Serum Visions has slightly outperformed, and we're talking over millions of games. My explanation has been that Serum Visions is always going 3 deep, so it can line up longer streaks of cards with a value of 1, while Preordain only goes 2 deep any time it sees a card with a value of 1, so Serum Visions ends up giving you better draws... assuming you have a turn to make use of them... but I haven't really set that up in a formula to mathematically prove it.
I would do it myself, but I don't have the time right now. Can you perform this analysis for Serum Visions vs Preordain? All the logic I've looked at says that Preordain should be the better card, but every time I've run it through my MTG simulator, Serum Visions has slightly outperformed, and we're talking over millions of games. My explanation has been that Serum Visions is always going 3 deep, so it can line up longer streaks of cards with a value of 1, while Preordain only goes 2 deep any time it sees a card with a value of 1, so Serum Visions ends up giving you better draws... assuming you have a turn to make use of them... but I haven't really set that up in a formula to mathematically prove it.
Why would you need to run simulations when an exact polynomial solution exists?
Considering the top 3 cards of the library and 2 extra draws in addition to the cantrip, SV and Preordain perform the same. It gets interesting when you consider the top 3 cards and 1 extra draw instead. In this case Preordain wins by a single case, 011. With SV you'd draw the 0 and scry either 1 on top for the next draw, giving you a score of 1. With Preordain you'd bottom the 0, leave the 1 on top and draw it with Preordain, then draw the 1 with the next draw for a score of 2.
Why would you need to run simulations when an exact polynomial solution exists?
The simulation was actually built for another purpose but can be used for this. So I ran it out of curiosity, and while it's slight, Serum Visions has performed better and I'm trying to explain why. It could be an error in my simulation approach, or it could be an error in the logic for the polynomial solution.
In the 011 case for example, I'm not sure that your approach is actually correct. I agree with the Serum Visions part, but Preordain should never have a 011 case, because that final 1 is random. It will only have a 01X case where you scry to the bottom, draw the 1, and then have a random card. As a result Preordain only goes 2 deep. Maybe that third card is good, and maybe it isn't, but that's going to be random and outside of the scope of what Preordain does for you. Where Preordain wins is speed, but it doesn't set things up as well (though it's close).
I consider every possible ordering of the top 3 cards (along with their probabilities) and compare how SV and Preordain would fare in each case. You are choosing to condition your cases on the cards that you see with SV/Preordain, which I don't think is a smart way of doing it. Go read my first post again, and try to understand why I have 11 and 10 for Opt instead of 1x.
This is what it looks like for p=0.7, assuming 1 extra draw.
outcome, prob, sv, pr
111, 0.343, 2, 2
110, 0.147, 2, 2
101, 0.147, 2, 2
011, 0.147, 1, 2
100, 0.063, 1.7, 1
010, 0.063, 1, 1
001, 0.063, 1, 1.7
000, 0.027, 0.7, 0.7
I understand your logic, I'm just not convinced that it's right because I've observed otherwise over a huge number of games.
I'll tell you what, when I get home tonight I'll run my simulator with a deck using Opt, Serum Visions, and Preordain, at 250,000 games per (I would do longer, but there's time constraints with the hardware I have) and compare the results, I'll try to host the database too so that anyone can look at the results. I'll keep it to a simple deck to remove variables.
I understand your logic, I'm just not convinced that it's right because I've observed otherwise over a huge number of games.
I'll tell you what, when I get home tonight I'll run my simulator with a deck using Opt, Serum Visions, and Preordain, at 250,000 games per (I would do longer, but there's time constraints with the hardware I have) and compare the results, I'll try to host the database too so that anyone can look at the results. I'll keep it to a simple deck to remove variables.
You can do that for SV/Preordain because they function relatively similarly (though I'd argue if your simulation took into account fetchlands and shuffle effects or not which make SV much weaker...), but Opt is entirely different due to its instant speed utility which is very much an important factor for control and tempo-oriented decks (like say Geist decks). How are you going to determine through a simulation which one is more effective. Can your program determine if you'd have lost the game by not keeping a counterspell up by casting SV/Preordain (or losing tempo by not casting and keeping instant speed interaction up?), but not losing by keeping Opt/Counter up? In other words is your simulation entirely deterministic, or does it accurately replicate a real game of magic?
I forgot to set it up tonight, so I'll run it overnight. I haven't actually run it for 6 months or so, so I don't remember how long it takes to play games. Seeing as how it's overnight though, I can probably up it from 250k games.
You can do that for SV/Preordain because they function relatively similarly (though I'd argue if your simulation took into account fetchlands and shuffle effects or not which make SV much weaker...), but Opt is entirely different due to its instant speed utility which is very much an important factor for control and tempo-oriented decks (like say Geist decks). How are you going to determine through a simulation which one is more effective. Can your program determine if you'd have lost the game by not keeping a counterspell up by casting SV/Preordain (or losing tempo by not casting and keeping instant speed interaction up?), but not losing by keeping Opt/Counter up? In other words is your simulation entirely deterministic, or does it accurately replicate a real game of magic?
It actually does take into account fetchlands with known information, though fetches tend to not work well with a lot of scry, because the fetches undo your previous scry. It's also capable of playing draw go but not as effectively as it is at playing aggro and midrange. The basic algorithm that it functions on (at least for what I'll be running here) is that it rates cards in terms of clock speed or expected clock speed, prioritizes, maximizes mana, etc... and plays cards out. It was originally designed for burn decks which are basically the very definition of a card being a 0 or a 1, so when there's a clear cut good/bad play the card selection aspects work best (it was primarily built to prove SDT before it's banning in Legacy).
Anyways, what that all means is that while it's capable of doing things like keeping up counterspell mana and playing Opt at the end of the turn vs playing Serum Visions on that turn, the algorithms aren't really optimized to do so. In my spare time I've tried to tweak that... but it's not easy, and a lot of what is and isn't optimal comes down to the specific matchup and the potential/actual clock you're facing down.
Basically, it adds too many variables. You're better off just looking at how each one changes the clock speed, and then making decisions from there.
What it looks for isn't really whether a line would have won or lost any specific game. Instead it plays each game to the best of it's ability and records game information in a database. From there I have a set of database queries that can find general trends, such as which cards lead to faster games, which cards are more common in games won or lost, and so on.
It's not perfect, but it has been very good at identifying optimal deck skeletons, curves, and evaluating which cards are better than others for certain types of decks.
Private Mod Note
():
Rollback Post to RevisionRollBack
To post a comment, please login or register a new account.
However, I in general I see Vision > Opt and Opt = Sleight of Hands.
Greetings,
Kathal
Modern/Legacy
either funpolice (Delver, Deathcloud, UW Control) or the fun decks (especially those ft. Griselbrand)
Greetings,
Kathal
Modern/Legacy
either funpolice (Delver, Deathcloud, UW Control) or the fun decks (especially those ft. Griselbrand)
Opt vs Sleight of Hand is a more pertinent question. I'll make some simplifying assumptions for the following analysis. Suppose your deck is comprised solely of good cards with value 1 and bad cards with value 0, and a random draw (without knowing if it's good or bad) has probability 0<p<1 of having value 1 and 1-p of having value 0. Also suppose the following sensible strategies for playing Opt and Sleight: with Opt, if the scryed card has value 1, you leave it on top and proceed to draw it, otherwise you bottom it and draw a random card. With Sleight, you always take a value 1 card if you see it, otherwise you take either of the two 0s.
Consider all possible permutations of good/bad cards in the top two. Additionally, like with Serum Visions, I'll assume that you have one draw step to spare. This assumption has the effect of leveling out the field - no matter what the top 2 cards were, what you chose to do with the cantrip, or even which cantrip it was, after cantrip + draw, the very next card you draw is random. Example: if both cards on top are 1s, you play Opt, see a 1, keep it on top, and draw it with Opt. Then your draw step yields another 1, and the one after that yields a random card with EV p. For each permutation, what is the probability of it happening, and what is the value of Opt+draw compared to Sleight+draw?
Opt + draw
11: prob p^2, value 2
10: prob p(1-p), value 1
01: prob p(1-p), value 1+p
00: prob (1-p)^2, value p
Sleight + draw
11: prob p^2, value 1+p
10: prob p(1-p), value 1+p
01: prob p(1-p), value 1+p
00: prob (1-p)^2, value p
The expected value calculation shows that both cantrips are dead even for any p. However, note that Sleight is more consistent. With Opt you have two events (11, 10) that have values 1 and 2, whereas with Sleight, the same events have value 1+p each. You won't hit the max of 2 with Sleight, but you won't hit the low of 1 either.
Opt does have one advantage which the above analysis doesn't put a number on: being an instant. A deck with many instant-speed counters/removal spells can pass the turn without committing to anything, then killing whatever the opponent plays, or cantripping with Opt if they don't play anything worth killing.
In short: Sleight and Opt are really close, close enough that although the numbers say Sleight is better (in the second-order stochastic sense), the outcome could easily be reversed if you made some other sensible assumptions, or simply if you value instant speed more than lower variance. Serum Visions is still the boss though.
Exercise for the reader: perform the above analysis for Serum Visions vs Opt or Serum Visions vs Sleight of Hand, but considering two draw steps and the top three cards instead (can you see why?). What strategy do you use for Serum Visions?
| Ad Nauseam
| Infect
Big Johnny.
Opt is at its best in more draw-go, controlling, no-silver-bullets decks like UWR variants or UW Control. That's a fair few decks already.
Serum Visions is still better at digging for specific cards than Opt is, as it goes through 3 cards instead of 2. This means I'd go for the full playset of Serum Visions first in combo decks, Blue Moon (digging for Blood Moon and Roast-level red removal is that good), likely Scapeshift (I aggressively keep 2-land hands with a Serum Visions in them, and I'm not sure I'd do the same if Serum Visions were Opt instead), and tempo decks such as Grixis Death's Shadow (digging for creatures is swell).
Opt might not even unseat Sleight of Hand. Assuming more than half your deck is dreck at any given time (so in izzetmage's example, p < 0.5), the chance that you bottom a bad card and make your next draw a teeny bit better is greater with Sleight than with Opt. After all, if the top card is good, you picked it with Opt and didn't bother checking whether the card second from the top is bad.
With that being said, unless self-milling (or milling your opponent just in case you're still playing Mill) brings significant benefits (such as Delve or Lingering Souls), I think Opt replaces Thought Scour.
What deck is running Thought Scourt that isn't also designed to utilize the graveyard?
Some versions of storm were running Thought Scour. And while Storm does use the graveyard, it's probably better for that deck to replace it with Opt as the card selection is far more valuable than milling over 2 unknowns.
In those decks I think Opt has to fight Think Twice to even get above 0 copies. Pure control doesn't even play 4 Serum Visions all the time.
I see several UW Control decks with the full playset of Serum Visions (and no Think Twice). I think they want Opt instead.
Pretty much is how it's going to go. There are a few cases where opt could be better than SV but that's going to be rare. Opt is better than SV in the early turns in control builds, but mid to late game you want to be casting SV. So just keep your SV in there and maybe take something else out for opt.
I'd actually rather have peek over opt btw. The information you get from peek is better than scry 1.
If you're talking about UWx, I would venture to agree with Jordan Boisvert that most competitive decks run 4x Serum Visions and have come around to the strength of the card, and most lists that don't aren't optimized.
As far as Opt vs Serum Visions, I plan on testing a playset of both in UWR/Jeskai Control just to see if the turbo xerox rule can hold up without brainstorm/ponder (cutting 2 lands and 2 spells for the 4 opt, for a total of 22 lands and 8 hard cantrips + 2 remand/2 electrolyze). Should be interesting.
Modern Tallowisp Spirits - A Modern Tallowisp Deck UW
Eldrazi Ninjas - Summoning Octopus Jutsu YYYYAAAHHHH!
STANDARD
Naban Wizards
Serum Visions is better than Opt. Opt is nearly unplayable. The amount a card digs matters a lot. Dig 1 cantrips are unplayable (even though some people still play them), Dig 2's have corner case uses but are still a hard sell (see Sleight of Hand). Opt is effectively a dig 2. Few decks are going to want it. In order for a cantrip to really pass the bar, it needs to be a dig 3 like Serum Visions is.
The amount a cantrip digs is basically the most important aspect of it. Instant/Sorcery matters little.
I would do it myself, but I don't have the time right now. Can you perform this analysis for Serum Visions vs Preordain? All the logic I've looked at says that Preordain should be the better card, but every time I've run it through my MTG simulator, Serum Visions has slightly outperformed, and we're talking over millions of games. My explanation has been that Serum Visions is always going 3 deep, so it can line up longer streaks of cards with a value of 1, while Preordain only goes 2 deep any time it sees a card with a value of 1, so Serum Visions ends up giving you better draws... assuming you have a turn to make use of them... but I haven't really set that up in a formula to mathematically prove it.
Considering the top 3 cards of the library and 2 extra draws in addition to the cantrip, SV and Preordain perform the same. It gets interesting when you consider the top 3 cards and 1 extra draw instead. In this case Preordain wins by a single case, 011. With SV you'd draw the 0 and scry either 1 on top for the next draw, giving you a score of 1. With Preordain you'd bottom the 0, leave the 1 on top and draw it with Preordain, then draw the 1 with the next draw for a score of 2.
| Ad Nauseam
| Infect
Big Johnny.
The simulation was actually built for another purpose but can be used for this. So I ran it out of curiosity, and while it's slight, Serum Visions has performed better and I'm trying to explain why. It could be an error in my simulation approach, or it could be an error in the logic for the polynomial solution.
In the 011 case for example, I'm not sure that your approach is actually correct. I agree with the Serum Visions part, but Preordain should never have a 011 case, because that final 1 is random. It will only have a 01X case where you scry to the bottom, draw the 1, and then have a random card. As a result Preordain only goes 2 deep. Maybe that third card is good, and maybe it isn't, but that's going to be random and outside of the scope of what Preordain does for you. Where Preordain wins is speed, but it doesn't set things up as well (though it's close).
This is what it looks like for p=0.7, assuming 1 extra draw.
outcome, prob, sv, pr
111, 0.343, 2, 2
110, 0.147, 2, 2
101, 0.147, 2, 2
011, 0.147, 1, 2
100, 0.063, 1.7, 1
010, 0.063, 1, 1
001, 0.063, 1, 1.7
000, 0.027, 0.7, 0.7
| Ad Nauseam
| Infect
Big Johnny.
I'll tell you what, when I get home tonight I'll run my simulator with a deck using Opt, Serum Visions, and Preordain, at 250,000 games per (I would do longer, but there's time constraints with the hardware I have) and compare the results, I'll try to host the database too so that anyone can look at the results. I'll keep it to a simple deck to remove variables.
You can do that for SV/Preordain because they function relatively similarly (though I'd argue if your simulation took into account fetchlands and shuffle effects or not which make SV much weaker...), but Opt is entirely different due to its instant speed utility which is very much an important factor for control and tempo-oriented decks (like say Geist decks). How are you going to determine through a simulation which one is more effective. Can your program determine if you'd have lost the game by not keeping a counterspell up by casting SV/Preordain (or losing tempo by not casting and keeping instant speed interaction up?), but not losing by keeping Opt/Counter up? In other words is your simulation entirely deterministic, or does it accurately replicate a real game of magic?
It actually does take into account fetchlands with known information, though fetches tend to not work well with a lot of scry, because the fetches undo your previous scry. It's also capable of playing draw go but not as effectively as it is at playing aggro and midrange. The basic algorithm that it functions on (at least for what I'll be running here) is that it rates cards in terms of clock speed or expected clock speed, prioritizes, maximizes mana, etc... and plays cards out. It was originally designed for burn decks which are basically the very definition of a card being a 0 or a 1, so when there's a clear cut good/bad play the card selection aspects work best (it was primarily built to prove SDT before it's banning in Legacy).
Anyways, what that all means is that while it's capable of doing things like keeping up counterspell mana and playing Opt at the end of the turn vs playing Serum Visions on that turn, the algorithms aren't really optimized to do so. In my spare time I've tried to tweak that... but it's not easy, and a lot of what is and isn't optimal comes down to the specific matchup and the potential/actual clock you're facing down.
Basically, it adds too many variables. You're better off just looking at how each one changes the clock speed, and then making decisions from there.
What it looks for isn't really whether a line would have won or lost any specific game. Instead it plays each game to the best of it's ability and records game information in a database. From there I have a set of database queries that can find general trends, such as which cards lead to faster games, which cards are more common in games won or lost, and so on.
It's not perfect, but it has been very good at identifying optimal deck skeletons, curves, and evaluating which cards are better than others for certain types of decks.