I played Shard Volley back in the day and it can bite you in the ass. The lower the land-count and the more copies of it the more likely the ass-biting becomes. I went down to one or two copies pretty quickly. I don't actually remember how many. I think I am going to try a copy or two in place of the Gonti's along with the second Grim Lavamancer and a card to be named later.

2 boros with splash for destructive revelry in the side (and both had 3 bridges in the side)

1 nactal burn (that's been a while!)

Also 2 players were 8-0 after day 1 (1 was one of the boros list that made the top 32):

2 boros with splash for destructive revelry in the side (and both had 3 bridges in the side)

So for the past few big tournament, it seems like the best performing list have the splash of green for destructive revelry

Thanks for reporting these lists. I think GP Lyon is only good for entertainment value. The meta should be drastically shifting now. I don't know how the prevalence of enchantments will change moving forward, but we should see more Jund (so Nacatl will be worse) and more aggressive decks (so Stomping Ground is worse).

Do you have an easy way, elcon, do determine the EV of opener Vexing Devils in the current meta? I.e., integrate its damage over all decks in your database? I know you've posted that it's worse than Shock if the opposing deck has 4 T1 answers and 8 T2 answers, but this still doesn't convince people for some reason. We need something like: opener VDs average 2, first draw VDs average 1.9, later VDs average 0 given the actual meta game to really try to convince people.

That is a glaring omission to that section of the first post. I'll figure out a way of modeling it more rigorously and add something to the primer about it soon.

The blurb I posted the other day after someone on Reddit asked me about the Elementals and Vexing Devil "back of the envelope"-ish, based on the assumption that Devil will never enter the combat step. That's where I got the worse than Shock number.

Spark Elemental should be relatively easy as well. Hellspark is complicated, but I think I did an acceptable estimate of it on Reddit. I'll add all 3 to the primer.

Awesome! It doesn't have to be super involved. Something like your back of the envelope calculation, but averaged over the field with their actual number of T1 and T2 answers. You can basically assume its EV is 0 after the first draw step or so since there are so many ways it's not relevant after that.

Do you have an easy way, elcon, do determine the EV of opener Vexing Devils in the current meta? I.e., integrate its damage over all decks in your database? I know you've posted that it's worse than Shock if the opposing deck has 4 T1 answers and 8 T2 answers, but this still doesn't convince people for some reason. We need something like: opener VDs average 2, first draw VDs average 1.9, later VDs average 0 given the actual meta game to really try to convince people.

I've been jamming 4 MB Molten Rain on Cockatrice and it's lols against all of the BBE, JTMS, and Tron decks. Guide -> Eidolon -> Molten Rain beats lots of decks, even if Molten Rain is just Stone Rain haha.

Pr(keep) = OneLander*Pr(keep one-lander) + TwoToThree*Pr(keep a two- to three-lander).

You don't have to do this kind of problem using a simulation. Pr(Keep a two- to three-lander) is probably close to 1. A consensus considering one-landers hasn't been met, but I'd guess it's roughly 1/3.

Under these assumptions, the optimal (w.r.t. keep on 7 probability) land count for a 60-62 land deck is 20/60 with Pr(keep)=65.97%.

If your Pr(keep one-lander) > 0.338 and your criterion for optimality is probability for keeping opening 7s, then 19 lands is better.

If you want 6 spells and 3 lands, you should go with 20/60 lands, 7 spells and 3 lands, you should go with 18 lands. Etc. There are a lot of different criteria, and they all point to 17-21 lands, with most point to 19-20. With the number of games most players play, you're unlikely to ever witness the differences between these land counts, though.

That deck's not popular, so don't change your decklist to try to beat it. Within the game, you probably want extra Lavamancers and Searing Bloods. You may not want Paths since they're already going to be trading counterspells for some of your cards, so you don't want your other draws to also trade with their cards. Also DRev hits Courser.

Another question: why do we play Destructive Revelry over Shattering Spree? Yes Revelry hits enchantments an deals 2 but doesn't a Chalice of the Void shut us DOWN? Wouldn't it be better to have a playset of Shattering Spree to potentially destroy Chalices? Also Spree seems better against Affinity too, just blow up their whole board.

Searing effects are good enough against Affinity (Blood, Blaze, Revelry, Smithereens). Yes, we lose to Chalice on 2 on turn 3-4 IF we don't already have them very low or have a bunch of 1-cost spells, but we lose to Leyline on turn 0 unless we have a creature nut draw opener.

You do need creatures to get utility out of the pump mode, but I view the pump mode as a perk on top of the skullcrack mode.

When I play skullcrack, it's either player only lightning strike or player lightning strike that shuts down life gain. When I play Atarka's Command, it's either the same as skullcrack (short of damage can't be prevented, which I don't value highly) or player lightning strike that produces a back being amount of damage via creatures. While I need creatures to get that, I don't feel compelled to play more of them because if I have none then AC is almost the same as skullcrack.

The floor is skullcrack, and the ceiling is sky high. The fact that the probability of getting more than skullcrack out of it is non-zero and positive, so it's better than skullcrack to me. I'm also willing to pay life for green in order to get that ceiling.

If I can choose between mashed potatoes and mashed potatoes with gravy, I'm picking gravy every time.

The floor if you draw Atarka's Command is "Skullcrack; you lose 2 life." Though, there are many games where you don't draw it and you had to pay the 2 life anyway. I'm still on the fence if it's worth it or not.

According to this, as long as you have at least 14 sources of a color, you should be pretty statistically fine. Maybe one more is ideal, because he does his calculations under the assumption of 24 lands and we use 18-20.

I wouldn't fudge Karsten's numbers when making deck-building considerations. Instead, use his methodology with your land count to see how many colored sources you need.

Also, note that 14 G and 14 W is not adequate even in a 24-land deck since his guarantees are one-at-a-time. I.e., if 14 G gives 90% to find it in your opener (or whatever % and whatever metric he used), then 14 G and 14 W does not give 90% to find W and 90% to find G. In fact, I think he has a table for double color requirements, so if you want a R/W land and a R/G land by turn 2, I think he recommends 20 total sources of those colors, which is many more than the 16 you'd have separately (assuming 12 fetches, 2 of each shock, say).

I don't know if this card is good or not, but you have to consider that PTE is a 1-for-1 that Evolving Wildses them, but this card isn't a trade that does so. You're up on (V)CA until they remove it or outclass it.

I think your algorithm gives a pretty high 1-lander keep rate, elcon. You might want to try fewer lands. Ignoring mulligans, the likelihood that you have exactly 3 lands in your top 10 cards is maximized by having L/60 = 3/10 (as you'd expect!). Similarly, to maximize the probability of having 3 lands in your top 9 cards, you should have solve(L/60=3/9) lands in your 60-card deck.

To add to the mathematics, if your mulligan criteria are: (1) mulligan all sevens, sixes, and fives with 0 or more than 3 lands, (2) mulligan all fives with more than 2 lands, and (3) mulligan all sevens, sixes, and fives with 1 land with probability p, then--to optimize your average starting hand size--you should run 22/21/20/19/18/17/16 lands if p is less than the following values (but not less than each of the previous): 0.12/0.30/0.48/0.65/0.82/0.99/1.01. To estimate how many lands you should run, draw 6 cards from a 41 non-land deck many times and determine the proportion of keeps, then look to the above table.

Does anyone have an alternative mulligan criterion set? I enjoy computing these things.

Yeah, Spark Elemental!

4 Guide

4 Swift

4 Spark

2 Grim

4 Bolt

4 Rift

4 Spike

4 Skewer

4 Bump

4 Gonti

4 Volley

18 lands

It's super fun! The idea is to kill T3/4 as often as possible.

Thanks for reporting these lists. I think GP Lyon is only good for entertainment value. The meta should be drastically shifting now. I don't know how the prevalence of enchantments will change moving forward, but we should see more Jund (so Nacatl will be worse) and more aggressive decks (so Stomping Ground is worse).

Awesome! It doesn't have to be super involved. Something like your back of the envelope calculation, but averaged over the field with their actual number of T1 and T2 answers. You can basically assume its EV is 0 after the first draw step or so since there are so many ways it's not relevant after that.

OneLander = (l, 1)*(n-l, 6)/(n, 7)

TwoToThree = (l, 2)*(n-l, 5)/(n, 7) + (l, 3)*(n-l, 4)/(n, 7)

Pr(keep) = OneLander*Pr(keep one-lander) + TwoToThree*Pr(keep a two- to three-lander).

You don't have to do this kind of problem using a simulation. Pr(Keep a two- to three-lander) is probably close to 1. A consensus considering one-landers hasn't been met, but I'd guess it's roughly 1/3.

Under these assumptions, the optimal (w.r.t. keep on 7 probability) land count for a 60-62 land deck is 20/60 with Pr(keep)=65.97%.

If your Pr(keep one-lander) > 0.338 and your criterion for optimality is probability for keeping opening 7s, then 19 lands is better.

If you want 6 spells and 3 lands, you should go with 20/60 lands, 7 spells and 3 lands, you should go with 18 lands. Etc. There are a lot of different criteria, and they all point to 17-21 lands, with most point to 19-20. With the number of games most players play, you're unlikely to ever witness the differences between these land counts, though.

Searing effects are good enough against Affinity (Blood, Blaze, Revelry, Smithereens). Yes, we lose to Chalice on 2 on turn 3-4 IF we don't already have them very low or have a bunch of 1-cost spells, but we lose to Leyline on turn 0 unless we have a creature nut draw opener.

The floor if you draw Atarka's Command is "Skullcrack; you lose 2 life." Though, there are many games where you don't draw it and you had to pay the 2 life anyway. I'm still on the fence if it's worth it or not.

I wouldn't fudge Karsten's numbers when making deck-building considerations. Instead, use his methodology with your land count to see how many colored sources you need.

Also, note that 14 G and 14 W is not adequate even in a 24-land deck since his guarantees are one-at-a-time. I.e., if 14 G gives 90% to find it in your opener (or whatever % and whatever metric he used), then 14 G and 14 W does not give 90% to find W and 90% to find G. In fact, I think he has a table for double color requirements, so if you want a R/W land and a R/G land by turn 2, I think he recommends 20 total sources of those colors, which is many more than the 16 you'd have separately (assuming 12 fetches, 2 of each shock, say).

Does anyone have an alternative mulligan criterion set? I enjoy computing these things.