[Free Podcast] So Many Insane Plays # 43: Gitaxian Probe and Interview with RayR

[Smmenen]

We've heard your request for more shows, and we've produced another one for you. 

http://www.eternalcentral.com/so-many-insane-plays-podcast-episode-43-gitaxian-probillard/

Kevin Cron and Steve Menendian discuss the numerous layers of tactics of Gitaxian Probe, and interview vintage Vintage player and tournament
organizer Ray Robillard (aka iamfishman).

Podcast (somanyinsaneplays): Download (Duration: 1:49:18 — 74.7MB)

0:01:00: Announcements
0:02:20: Gitaxian Probe
0:59:55: Interview with Ray Robillard
Total runtime: 1:49:18
Show Notes

– Vintage Super League
– Chris Pikula on Gitaxian Probe

Contact us at @ManyInsanePlays on Twitter or e-mail us at SoManyInsanePlaysPodcast@gmail.com

[diophan]

Thanks guys for making another episode. I'd like to give a few counterpoints to some of the probe discussion:

I felt like several of your arguments against probe leaned too heavily on people making bad decisions. For example, your "0 hand 1-2 probes" mulligan question. Most people are not cutting land for probes, so that hand is an automulligan without probe. With probe in your deck, you are given the opportunity to keep the hand. Whether you make the right decision is another question. Additionally, if you've mulliganed down to say 5 having probes make you more likely to hit your first or second land drop. My experience playing delver is that once you start mulliganing below 6 you run the danger of mulliganing into oblivion because it is difficult to get the first mana source with small hands.

To take your "information can be bad" argument to a silly extreme, it's potentially bad to know how many cards your opponent has in hand or how many untapped mana they have, but no one lowers their blast shield and uses the force to play better Magic without knowledge of such things.

In the game 3 against Randy I don't agree that the extra information made Kai take a worse line. If he would have jammed he would have gotten his Academy wastelanded. Furthermore, I don't see any point in that game where Kai could have held onto Academy and played out threats earlier that would have resulted in a win. I did just quickly glance through the game again so I could have certainly missed something.

I don't agree that one necessarily cuts countermagic in a delver shell for probes. For instance, here are the consequential differences between Steve's last VSL deck and the one I got 17th with at EE2:

Mine:
+2 probe
+1 LoA

Steve's:
+1 pyromancer
+1 gush
+1 misdirection

My list balances playing 1 fewer gush by getting digs online earlier. Although you did mention this point, for me a huge chunk of the reason for playing probe in a delver style deck is to enable your delve spells. I don't think the thinning aspect is a primary motivator for most people. Although I am playing one fewer threat, I would argue that each of my token generators is slightly more powerful and can be cast sooner, since the monk token from a mentor off a probe is still a formidable threat in the right matchup.

A final note: I think a more interesting statistic than how many probes were played in a given time period would be the average number of probes played in a delver style deck. A reasonable definition of a delver style deck might be something like 4 preordain, 1 ponder, 2-4 gush, 2-4 dig, and some number of pyromancer/mentors.

[nedleeds]

Interesting cast, I'm a little older and actually used to (drive almost an hour) to work at the comic book store in the Danbury Mall in Connecticut circa 1994.

On the mulligan discussion you can quantify the chances of drawing a land pretty easily

Using Steve's 14 land Mentor deck as an example, lets say we replaced the Pyroblasts with Probes. We are looking at a no lander with a probe on the play. Discarding the chance of your Probe getting Misstepped (you could probably develop a heuristic for factoring this in given the % of blue decks in a field with the average missteps per deck being 3-4) the chance of the top card of your library being one of those 14 is 26.4%. Pretty horrific. The chance of finding 1 and only 1 land on a 6 card mulligan is 38.3%. I'd argue you are always better off taking 6 if your goal is to obtain a land. You can of course also calculate any number of scenarios using failure scenarios and the multivariate hypergeometric distribution.

Another thought experiment is to discard the almost unquantifiable parts of Gitaxian Probes ephemeral value: being blue, costing 1, requiring 2 life, or in the case of Street Wraith being a black card, being subject to Phyrexian Revoker, costing 2 life, having a converted mana cost of 5. You can just ask yourself how would you construct U/r/x Delver if you were allowed to play 56 cards. You certainly wouldn't argue that a 56 card deck construction limit is superior to 60, right?

Another interesting experiment in the same vein as the probe argument was born of boredom with a magic playing buddy one evening at Gencon, we had come back (a non 0 amount of alcohol had been consumed) to the hall and were waiting for some mutual friends to finish up a board game session (so we could leave and continue consuming). We had no cards on us, we were at paper covered tables waiting for these folks to finish. We had pens and copious amounts of writing space. We invented a new kind of magic called 'Perfect Ultimate Vintage'. In PUV you write your opening hand down, hidden from your opponent. The 'deck' construction rules follow the vintage B&R list. After writing down your hidden hand you roll for first (adding a little twist). Each card you would draw, or reveal off you library can be any card, at runtime. Anyway, its an interesting exercise in magic. Cards like Mindbreak Trap, blue leyline, Chancellor of the Annex, force of will, are all very important to not get blown out. But the one card that instantly became ubiquitous and had to be banned was ... Street Wraith.

Thanks for taking the time to record.

-Sean O'Brien

[DaveKap]

Thanks for putting together an analysis of Probe and it was great to hear about the origins of Waterbury.

I would be interested in hearing more about the following of Probe:

1) Value of blue card to Pitch vs. say a Pyroblast
2) Value of being an extra card in the gy for Delve
3) Value of token generation
4) Psychological value of your opponent's resources being exposed. I have found that after casting Gitaxian Probe, my opponents often feel "naked" and somewhat dejected.

[thecrav]

Some feedback on the cast itself: When you mention cards that aren't currently seeing a fair amount of play, it would be helpful if you'd have a quick mention of what those cards do. Newer players may not know what these previously-played cards do and people like me who drive and listen aren't able to look them up.

For the discussion of Street Wraith vs Gitaxian probe, there was one point that I was surprised you didn't bring up: Contribution to Storm count.

Some simple examples:
* If I'm playing a storm deck, I probably want the Probe to fuel my Tendrils.
* If I'm looking to resolve some other big bomb-y sorcery such as Tinker or Show and Tell, I'm probably wanting to play Street Wraith instead so I'm not fueling your Flusterstorm.
And then all that changes up again if I've got a (castable) Flusterstorm of my own!

[diopter]

I'm quite intrigued when you say you play Delver like a hard control deck. I feel the same way in a sense about Delverless Mentor builds - BKelly's two-Mentor innovation now seems perfectly natural as a result.

But, Delver the card allows Delver the deck to play soft control so effectively that I probably would be inclined to Probe in that shell just to support that line.

Also, even in the controlly 2-Mentor builds I still play a few Probes so I can get a feel for whether I can Gush or Ancestral safely to advance my control position.

One thing I really really agree with is that no-land probe hands are death. Probe is not a cantrip in the practical sense of the term - it has zero digging power!!! Mana bases should be built stably irrespective of Probe counts, only the variance-reducing Preordain should be allowed to shave your land base

[mueller]

glad that you touched on among the most important points of probe: it is sorcery speed and takes the place of countermagic. If you draw one of a main phase Gush, you have to use it immediately to gain it's filtering value, otherwise, it's a dead card until your next turn. If you are digging for countermagic at instant speed, Probe is completely dead. If you Preordain and see Probe, you are actually losing card selection. The inclusion of Probe over oft-omitted Mystical Tutor is difficult for me to understand.

[gribdogs]

Another fun listen.  I really liked the interview at the end.  I would suggest future podcasts continue in that vein by chatting with Vintage notables.

As for the main point of Probing, I have to say you guys have put way more thought into that card than I thought possible.  To be fair I'm a terrible Magic player, so I really don't dig too deep into these sorts of issues with a card.  Having said that I just dumped a playset of Probe in a casual deck to play in the near future.   

[MaximumCDawg]

A++ would eat again. 

Thanks for making a new cast so soon!  You successfully made me yell at the radio this time, so well done.  The particular thing I was yelling occurred when you discussed how cutting countermagic for Probes was bad in Delver.  Your point, and it is correct, is that by replacing Countermagic with Probes, you are lowering the overall percentage of Countermagic to non-Probe cards in your deck.  You make a 56 card deck, sure, but you've lost 4 counterspells out of it and so the overall percentage is lower.  As far as that goes, it's fine.

HOWEVA

The natural response is, "why the heck not cut one of everything to keep the proportions the same?"  In other words, cut one Delver, cut one Preordain, cut one land, cut one counterspell.  Baddaboom, baddabing, same percentages and smaller deck.

Now, I get that you sometimes plain don't want a Probe for other reasons you mentioned -- sorcery speed and mulligan problems and so forth -- but this seems like an obvious solution and I was not satisfied by your discussion about how it was "obvious" that you don't cut a Preordain for a probe.

[diopter]

CDawg: that does not work in practice (the shaving of different classes of cards).

It comes down to the tension between the mulligan and the late game. Certain classes of cards just make opening hands hum, you cannot just shave them.

For instance, in current iterations of UR Delver I would never cut a Delver. I want better odds of having turn 1 Delver, that's part of why I want to play Probe in the first place.

Cutting Preordain for Probe is just silliness squared. That's cutting off the nose to spite the face. Preordain is one of the most pivotal cards in modern Vintage for reducing variance, I can't imagine playing this style of blue deck today without the playset. It's possible you want to play more, I've dabbled with Sleight of Hand.

I'll tell you this, Smennen has been beating the quad laser Preordain drum for 4 years now. I used to scoff too, but infi play testing hours later, I conclude that he is completely correct.

In fact, Probe is AWESOME at defending Delvers and Preordains because they attract Missteps. And Probe gives you amazing info for Preordaining. You want to play them aside each other.

And, of course, the usual math of 2 Cantrips to remove 1 land doesn't really apply to Probe. That will lead to a lot of heartbreaking mulligans.

---

However, if you cut late game stuff like Dack or DTT,or Mentor ,you just disproportionately skew the ratios of those critical bombs. A late game DTT whiffing on finding e.g. Mentor is really annoying. And there are many situations where only a cantrip chain into DTT into Bolts can win, and often Probe will just break that chain.

---

I am playing 3 Probes right now in my Mentor build but a lot of thought went into A.) the 3-count, and B.) the overall big-picture and little-detail changes required to bring the shell to 57 cards.

[John Cox]

I think you guys have this all wrong. You shouldn't cut cards to add probe, you should build around a 56 card deck. We don't make a 64 card deck and then say "what 4 cards am I going to cut to reduce variance?" We go for 60 from the start. If 4 of probe is included they should be the first cards in the deck.

[diopter]

John Cox: this is a contradictory and impossible approach due to mulligans.

[John Cox]

Having played with 4 probe in belcher (and street wraith before probe was printed) I can easily say that knowing when to mulligan with a probe in hand becomes second nature. If anything it amplifies choices making the questionable more questionable and the keepable more keepable.

[diopter]

John Cox: could you expand on that a bit?

My rule of thumb with Belcher was always that if it couldn't deploy Belcher or Empty, mulligan. I believe this rule is shared by other Belcher players like Nat Moes. Obv there are exceptions but they are few.

From this perspective, I am not sure how Probe makes keepable hands more so. The Belcher/Empty-less hands definitely are still auto-mulls with limited exceptions, but suddenly there are a whole class of hands where Belcher or Empty are present but Probe obscures whether there is enough mana to go off.

[MaximumCDawg]

I think you guys have this all wrong. You shouldn't cut cards to add probe, you should build around a 56 card deck. We don't make a 64 card deck and then say "what 4 cards am I going to cut to reduce variance?" We go for 60 from the start. If 4 of probe is included they should be the first cards in the deck.

That's another way of putting the question I posed earlier.  You design your deck to have 56 cards in it and build the proportion of cards in it to maximize that.  Now, obviously you can't overcome the fact that Probe is Sorcery Speed draw and it "masks" the card it is going to draw you into.  Those limitations are insurmountable.  However, we can mathematically address the probability of having that first turn Delver if we go to four Probes and drop one Delver.

In a 60 card deck without a Probes and four Delvers, you have about a 56% chance to draw at least one in your opening hand.  (4/60 + 4/59 ... + 4/53).  So, that's the baseline.  

Say you cut one Delver (among other things) and go up to four Probes.  We gotta measure two things: the likelihood of finding a Delver naturally, AS WELL AS the likelihood of drawing more cards with Probes and finding a Delver that way.  Initially, we have (3/60 +.... + 3/53), or 43% chance of drawing at least one Delver naturally if we only run three of them.  Big drop from running four.  

Next, let's account for Probe's drawing effect.  Here, we (regrettably) need to account for the chance that you draw multiple Probes and thus have multiple extra draws.  Ugh.  Okay, well, you have a 43% chance of drawing AT LEAST ONE Probe in your opening hand.  In some of those cases, you'll draw more than one Probe.  How likely is that?  This gets messy because now you have a nested series of probabilities dependent on what you drew; if you drew one probe initially, your subsequent chances of drawing probes change.  By my reckoning, the chances of having TWO probes is about 17%.  I would not be surprised if I made a math error here, but this seems intuitively right anyway.  Going up to the case of drawing at least THREE probes is more than I want to tackle at the moment; someone with MATLAB might wanna give it a go, I've gotta do it by hand.

So, you have a 43% chance of having AT LEAST ONE probe in your hand.  And a 17% chance of having AT LEAST TWO probes in your hand.   We can incorporate these into our chances of the first turn Delver like so.

57% of the time we have no probes, leaving us with a 43% chance of drawing at least one Delver.
26% of the time we have exactly one Probe, meaning we can draw an eighth card.  This adds another 3/52 chance to draw at least one Delver, which is about 5%.
17% of the time we have at least two Probes, meaning we can draw an eighth and a ninth card.  This adds another 3/52 + 3/51 to the chance to draw at least one Delver, which is about 11%.

Now, I THINK those cases are all independent.  (Zero probes) + (one probe) + (two or more probes) should never overlap.  That means we add the percentages together to get a 59% chance of drawing a Delver on turn 1 with 3 Delvers and 4 Probes.  Mind you, I didn't add the extra probabilities contributed by drawing three or more Delvers, meaning the real percentage is probably slightly HIGHER than this.

So, my back-of-the-envelope math suggests that:

Chances of drawing a Delver in your opening 7 with 4 Delvers and no Probes: 56%
Chances of drawing a Delver in your opening 7 with 3 Delvers and 4 Probes: greater than 59%

On the other hand, if you just use a 56 card deck and calculate the chances of drawing a 3-of in your first 7, you only get 45%.  So, maybe my math is wrong, maybe I hit a rounding error, or maybe there's some subtle probability difference between having a smaller deck versus having Probes.  Probably the math error; I think you can safely equate Probes and smaller decks.  I dunno.

My brain hurts.

[wappla]

Your math is not right. Drawing an opening hand does not behave statistically how you presented it. One needs to use a stats function called hypergeometric distribution probability. Most spreadsheet programs have this built in. The probability of drawing at least one of a four-of in an opening hand of seven cards is 39.95%.

[MaximumCDawg]

I don't see how.

So when you draw one card from a perfectly random deck, you have a 4/60 chance of drawing one of your 4-ofs.  Assuming you do not actually draw it the first time, you now have a 4/59 chance of drawing it on your second draw.  And this continues.  What's wrong with this analysis?

[Smmenen]

The 39% figure is well established in magic strategy for your chances of drawing a 4-of in your opening hand of 7.  Use the Hypergeometric distribution and it will tell you that.

[wappla]

You can do it by hand, but be careful how you actually formulate it.

To find the number that have at least one Delver, we want to know how many combinations of draws have zero. That means every draw is not a Delver, so:

56/60*55/59*54/58*53/57*52/56*51/55*50/54 = 60.05%, which is the probability of drawing 0 Delvers. Note that above I stated the probability of drawing at least one (non-zero Delvers) was 39.95%, which is reassuringly 1-.6005.

[MaximumCDawg]

Okay, I get your analysis.  Can we apply something similar to determine how dropping a Delver but maxing out your probes affects the chances of an initial Delver?  Do we just drop the size of the deck and redo the analysis?

EDIT:  For those watching from home, when these cats toss around the term" hypergeometric distribution" they're talking about a special case in combinatorics / probability.  http://en.wikipedia.org/wiki/Hypergeometric_distribution

Basically, it looks like The Answer for problems exactly as we're posing here:
"The hypergeometric distribution applies to sampling without replacement from a finite population whose elements can be classified into two mutually exclusive categories like Pass/Fail, Female/Male or Employed/Unemployed. As random selections are made from the population, each subsequent draw decreases the population causing the probability of success to change with each draw."

So that's fine and good but how do we apply Gitaxian Probe when we're looking at this kind of analysis?

[wappla]

I wouldn't get too hung up on the math. Even with fancy stats functions, we are ignoring too much for it to be all that useful beyond the big picture, i.e. that you will draw a 4-of more often than a 3-of.

It gets very complicated very fast if you want a complete picture. The basics are fine and pretty easy to figure out, but the tree of possibilities grows quickly. Probe into Preordain, for example, or Probe into Sapphire letting us cast a previously drawn Ponder, or Probe into Brainstorm with a Sapphire and a fetchland in hand, and a second cantrip. The scenarios for even just the opening hand are quite large.

If we were so inclined, you could figure out probabilities for each relevant scenario and get an answer. Of course, for a true comparison, you'd also have to do the same for the non-Probe deck. Opening hand with no Delvers but a Sapphire and Ponder/Preordain/Brainstorm etc. What about 1-3 Probe decks? Etc.

But, to answer the simplest version of the question, we want to compare the probability of at least one Delver in a zero Probe deck where Delver is a four-of compared to the ability to find one solely via Probes turn 1 in a deck with three Delvers and Probe as a four-of. Let's assume we will only probe once on our first turn, as we value subsequent Probes for information in the midgame. We also are ignoring mulligan decisions altogether, which would significantly change probabilities. A lot of hands that factor into these odds would be snap mulligans. All in all, statistical analysis is not useful for analyzing the card: the value of the information gain is nebulous and the value of deck-thinning is highly matchup specific and depends highly on what the card is replacing in the deck. This notion that you start with a 56 card deck and add four probes is misleading because it obscures the opportunity cost of those four slots. Instead of adding four Probes, you could add, for example, Mystical Tutor, Merchant Scroll and two Probes, or countless other combinations of cards.

Anyway:

On the play:
4x Delver, 0x Probe: 39.95%
3x Delver, 4x Probe, using one Probe Turn 1: 33.09%

showing work:
31.54% chance of natural Delver(s)
+ 68.46% no Delvers * 39.95% hands with Probe * 5.66% chance of drawing Delver with first Probe = 1.55%

[ajfirecracker]

I'm listening to this right now, about an hour in, and WOW does your discussion of Gitaxian Probe mirror my thoughts about Serum Powder.

[Smmenen]

Weird.  Serum Powder is much more of a deterministic card.

In any case, I'll try to see if we can respond to most of the comments here in another podcast.

[Protoaddict]

2 comments to add to the discussion:

1 - I think that the 2 life we are paying to cast probe (not always but in like 99% of the examples given above it is the case) is being drastically under recognized in these arguments. The more probes in a deck, the more likely you are to see more probes, the more likely you are to have to cast them, the more the hit to your life total matters, especially when you consider fetches. Playing 4 probes does not give you a 56 card deck, it gives you a 56 card deck with the possibility of you starting with 18-12 life instead of 20.

We are kinda in a post combo world right now. Life totals matter, especially when a large chunk of the format is playing tempo based damage decks (not poison that is) and some other lists are playing cards that require 7 life chunks. I think we really need to start looking at the actual cost of cards like this, because much like delves relationship to decksize, life is typically a finite number as most decks do not run life gain cards in any sort of number. You probably would not run 4 TC + 4 DTT + 4 Tasigur in a deck even if you could, and likewise you may not have the resources to run 4 probe, 8 fetch, etc in the same list.

2 - I still hold that probes value is largely untapped in decks that fully utilize its other attributes. I was running probes in Dredge for a time, and the synergies that spell has with the list is actually pretty crazy. When you consider it is a free Dredge trigger, plus tells you what to name for Cabal therapy, Plus tells you if you need to go for broke right away, the card is pretty amazing. If I thought that deck was playable right now (too much grave hate, so I don't) I would likely run a full compliment of Probes, Bazaars, Therapies, and try to establish board turn one. I did have a game with probe dredge not that long ago that was something like turn one Bazaar activation, Discard 2 trolls, Double probe dredge, gets some naros and bridges, decimate opponents hand and have enough for game.

[ajfirecracker]

Re:Serum Powder, I think it has a powerful niche use that Gitaxian Probe lacks (finding Bazaar of Baghdad) but otherwise it's quite similar. The same tensions regarding the value of thinning, the density of effects in the deck, and the tendency to draw into the right sort of thing (Mana for Delver, business for Tendrils) all work against Powder in a very similar way.

One thing I find really interesting about Dredge (especially versions which do not intend to flash back creatures or spells game 2) is that Serum Powder transforms into a Gitaxian Probe-like element used to thin pre-game and to construct more effective hands rather than as a 0-mana tutor effect.

I think a lot of the tensions regarding Gitaxian Probe in otherwise capable blue decks undergird the use of Serum Powder in "normal" decks like Espresso Shops. All of the same arguments about effect density and mulligan (/ deck manipulation) information begin to take shape.

[MaximumCDawg]

Proto, I don't think anyone is neglecting the disadvantages to Probe.  We were just noodling out whether or not you can preserve your percentages of key cards by just swapping one from each category out and putting in a playset of Probes.  It was an isolated point in the larger picture of the pros and cons of Probe.

Re:Serum Powder, I think it has a powerful niche use that Gitaxian Probe lacks (finding Bazaar of Baghdad) but otherwise it's quite similar. The same tensions regarding the value of thinning, the density of effects in the deck, and the tendency to draw into the right sort of thing (Mana for Delver, business for Tendrils) all work against Powder in a very similar way.

There's a superficial similarity, but there are such drastically different impacts of using those cards that I don't know if it's a fair comparison.  Powder is like having a free, one-sided, Wheel of Fortune; it's an ASTOUNDINGLY powerful effect.  The trade-off is that late game (defined as turn 1) Powders are neigh-worthless.  Probe doesn't have that problem.  It offers a smaller gain and a smaller downside.

[Islandswamp]

I haven't finished the episode yet, but I have to say, the discussion on Probe was the best thing that I've heard in a long time. That is the kind of depth that I love. When non-players ask me why I love magic, I mention the deep levels of strategy that exist. Articles and Podcasts such as this one are what I think the game needs, as they illustrate just how incredibly deep this game really is.

Thanks guys, for doing such amazing work. I've been doing my little articles on Vintage and Legacy, and I owe a lot to listening to this cast and to reading your articles. I've learned quite a bit.

Keep up the good work!

[ajfirecracker]

There's a superficial similarity, but there are such drastically different impacts of using those cards that I don't know if it's a fair comparison.  Powder is like having a free, one-sided, Wheel of Fortune; it's an ASTOUNDINGLY powerful effect.  The trade-off is that late game (defined as turn 1) Powders are neigh-worthless.  Probe doesn't have that problem.  It offers a smaller gain and a smaller downside.

Fair enough.

Basically I just wanted to say I enjoy the podcast and this episode got me thinking

[Varal]

From a theoretical standpoint, if you always cast all your probe on your first turn, including those you draw with probe, you effectively have a 56 cards deck for all probability calculations. Obviously not working on the draw and computing probabilities of Force of Will and Mental Misstep. Also, not true for mulligan, brainstorm, ponder, preordain.

Some people seems to have misconceptions with probability. You add probabilities when computing the probability that one of multiple independent events happen and multiply them when computing the probability that multiple independent events happen simultaneously.

Quick example with dice:
Probability to roll a 6,6: (Probability to roll 6 on first die)(Probability to roll 6 on second die)=1/36
Probability to roll a pair: P(66)+P(55)+P(44)+P(33)+P(22)+P(11)=6/36

[WhiteLotus]

Great podcast, thanks for the effort you guys put in .

However i'd would like to hear more about the "optimal" number of probes in different archetypes. I for one think that 4 probes in any deck is incorrect and would lean towards 3 for combo decks and 2 in Gush aggro (you probably shouldn't be playing any in other strategies). I think the card needs a critical mass of utily/synergies to become playable, your deck has to be able to leverage most aspects of the card whether it be the information given, the free storm, fueling the grave, Digging.
The inclusion of probe will always be in tension with a number of key aspects of the format, so in the end I don't think there is a correct answer on probe, it boils down to player preference/playstyle.