First, let me start off by saying, "whew". That was a lot of work
Just finished entering all of the matches from pierakor's "mono best tron" playlist, plus SwissRolls' games, into the sheet. We now have 960 games of Utron analyzed. You can view the spreadsheet here. Some key takeaways that I've noticed:
Best scoring three cards: The tron lands
Gemstone Caverns gets us just under a percent increase in win percentages when we're on the draw.
Best nonland cards to have in the opening hand are, in order, Repeal, Condescend, Supreme Will, Chalice of the Void, Anticipate, Thirst for Knowledge, and Remand.
Worst cards to have in the opening hand, in order from worst to best, are Treasure Mage, Ugin, and Wurmcoil Engine. Expedition Map is the next worst after that, but actually has a positive correlation to win percentages so long as there are no multiples. Ugin and Wurmcoil seem pretty self-explanatory for not being so great in the opener, but I'm a little more surprised at Treasure Mage. I suppose it makes sense, though. The body is fine, but what we get is useless if we don't live long enough to cast it. Having something that actually disrupts the opponent seems to perform much better for us.
The deck mulligans relatively well, going from a 69.41% win percentage with a full opener to 63.59% with six cards.
The deck performs relatively well with varied numbers of lands. Four is best (74.89%), followed by three (68.38%), five (64.15%), and two (63.12%).
Quite a bit more information to be gleaned there. Feel free to check it out, I figure it could help lead to some good conversation about deck tweaks.
Hi barrin_master_wizard It took a while, but I already had the template pretty much made thanks to my previous work on Lantern.
So, breaking down the data, this is what's going on (please forgive me if you have already figured some of this out, I don't know what parts you don't know, and I'd rather be thorough in the explanation):
Opening Hand Data tab: This is where all of the data from the gameplay videos is entered. I enter in if it's preboard or postboard, what deck the opponent is playing, if the Utron pilot is on the play or draw, if the game is won or lost, if the match is won or lost, who the pilot was, and how many of each card are in the opening hand. The columns on the far right automatically calculate various things, like how many cards are in the opening hand, how many lands are in the opening hand, if there is a blue source in the opening hand, if the hand contains natural tron, if the hand contains two of the three tron lands, etc.
The data in the Opening Hand Data tab is used to create pivot tables that show win rates with various numbers of each card in the opening hand, or comparing statistics based on the columns on the far right (natural tron, etc.).
Weighted Data Trends tab:
Column A has each card name, referenced from the associated pivot table for that card.
Column B shows the number of wins with no copies of the associated card in the opening hand.
Column C shows the number of games with no copies of the associated card in the opening hand.
Column D calculates that win rate, using the numbers from columns B and C.
Column E shows the number of wins with one copy of the associated card in the opening hand.
Column F shows the number of games with one copy of the associated card in the opening hand.
Column G calculates that win rate using the numbers from columns E and F.
Column H finds the difference in win rates between having one copy (column G) and zero copies (column D).
Column I is used to weigh the data. There will be different numbers of games to be compared for each card, since the cards are randomized in the deck. The issue before weighing the data was that something may have a great increase or decrease in win rates when the card is in the opener, but the sample size is too small to be reliable. Thus, I use a function in column I to weigh the data according to sample size. The way the function works is that I divide the number of games the card was in the opener by the total number of games. The more games that the card was in the opener, the larger this fraction will be, and the greater the weight is that will be applied to column G.
Column J is the product of the difference in win rates from zero to one (column H) and the weight (column I).
I also wanted to be aware of diminishing returns, though. Sometimes a card will be great in the opener, but having multiples is not so great for us. So the block of columns K through O do the same steps for the difference between one and two copies, the blocks of columns P through T do the same steps for the difference between two and three copies, and so on.
Columns AE and AF are used to find the overall weighted data trends with diminishing returns considered. If there is no data on having multiples of a card, then column AF will display "No Data". If the overall diminishing returns is negative, column AF will display "Negative". If the overall diminishing returns is positive, the a value (the product of column J and column AE) will be displayed. The function isn't perfect, but it's what I've been using for now.
After I've entered the gameplay data into the Opening Hand Data tab, I update the Weighted Data Trends tab to see if there is new data for having multiples of a card and then sort the rows by columns AF, AE, and J, in that order. Those columns are also color-coded to help read the cards that correlate with wins better or worse.
Additional Data Points tab: Some decks have specific data points that make it unique from others. For example, with this deck we might want to compare hands that have natural tron to hands that do not. Or, we might want to compare win rates of being on the draw naturally versus being on the draw with Gemstone Caverns in hand (putting us "on the play", in a sense). It gives us win rates for the different numbers of lands in the opener, number of cards in the opener (mulligan trends), win rates if we have a blue source in our opener, etc. There's plenty more data points we can look at, but that's where I ask for suggestions. What specific combinations of cards do people want to see? I can set up the spreadsheet to look at exactly that, usually.
As for using the spreadsheet for deckbuilding and tweaking, I need to reiterate what I've posted before (and on reddit).
First, I should probably explain the approach. Using the idea that every competitive Modern deck is designed to either deny the opponent the ability to have significant interaction with the gamestate, or to minimize what interaction they are allowed to have with the gamestate, the first few turns of the game are key to how each Modern deck will successfully accomplish this task. The phrase "Modern is a turn four format" is probably not new to anyone reading this, and pretty much aligns with this concept. If a deck isn't trying to either win as quickly as possible, denying the opponent the "time" (future turns) to make significant interaction, then it's probably trying to stop the opponent from achieving a quick win. Our goal in Mono Blue Tron is to either deny the opponent from making the plays they need to win as soon as possible or to delay them until we can make more significant plays. In those first four turns each player will have access to ten cards on the play or 11 cards on the draw, assuming neither player takes a mulligan. Thus, the opening hand will account for between 64% to 70% of the resources available to maintain control of the direction of the game during those crucial turns.
So, with the idea that our ability to maintain some control over the first few turns of the game is (or just survive that long) is going to be crucial to winning the game, I focus on the opening hand. It is very important to pay attention to more than just columns AF, AE, and J! For example, it may look like Island is the "worst card correlating with win percentages in the deck". But this does not take into account the context! What kind of opening hands would an experienced Utron pilot keep that did not have an Island in the opening hand? Well, a hand with natural tron and a Wurmcoil is a great example. That's quite a good opening hand. If we actually look at the win rates of opening hands with an Island, we'll see that the win rate is 66.46% (315/474 games). That's still quite an impressive win rate. It just happens to be lower than the win rate of hands that are kept without one Island. Paying careful attention to each number is very important here, otherwise we could easily be misguided or misinterpret the numbers, as you mention above.
What I've been doing to use this data to tweak my list is to pay attention to the cards that correlate the best with increased win percentages, but also pay attention to the diminishing returns. Here are some examples:
Supreme Will: This card seems to correlating quite well with increased wins, but the diminishing returns shows a sharp decline. The sample size on the diminishing returns is also quite small, though. Thus, for now, I know that I would love to have one in the opening hand, but I'm not ready to risk increasing the chance of having two. Therefore, I run three in my main.
Chalice of the Void: Nearly the exact same scenario as Supreme Will, above. I run three of these as well.
Thirst for Knowledge: Yes, it's pretty much ubiquitous that we run a full playset of this card. For me, though, it's comforting to see the numbers support this decision. I don't want to simply believe something is true because it's popular and everyone agrees that it's true, I want to know what is true.
Remand: See Supreme Will and Chalice of the Void explanations Same scenario, I run three.
Now, I've skipped over Condescend and Anticipate. It appears that it might be better if we cut a Condescend for the same reason that I only run three Will, Chalice, and Remand. I just haven't personally pulled the trigger on that yet. I'm inclined to cut the fourth Condescend for an Anticipate, which scores great. These are my flex numbers for now.
The wincons (Mindslaver, Wurmcoil Engine, Ugin, etc.) are obviously not the best to have in the opener. It doesn't take a rocket scientist to realize that having a 6+ drop in the opener isn't going to get us far in the first few turns without somehow assembling a fast tron. This does not mean we cut these cards, it just means that we need to be aware of how these cards are going to affect our chances at winning when they're in the opening hand. Most of that's relatively common sense, though.
I also use the data to figure out what the best performing sideboard cards are. For example, Summary Dismissal seems to perform very well, as does Spreading Seas, Negate, Silent Arbiter, and Filigree Familiar. Each of these cards are great at multiple matchups (Silent Arbiter being amazing in one of our worst matchups, Merfolk!).
So that's the spreadsheet, explained, and how I prefer to use the data. When combined with pierakor's MORT data (which gives better data on which wincons and singletons work the best, but isn't so good at finding correlating wins for cards that are run in multiples), we can use actual empirical data to tweak and build the deck.
So after a little bit of a break from the deck I'm back to playing in order to test for GP Atlanta. I'm only 3 leagues in with this version but so far I have two 4-1s and a 3-2. I like this list a lot and it feels pretty solid against the current field.
For the main deck, my version differs from most in that I'm playing myr battlesphere, opt, and spatial. The sphere is much better than wurm #2 in a lot of matchups now such as humans and any white deck. Wurmcoil is still strong against some decks but I have been far and away liking the 1-1 split right now. Also plat angel is back in the main thanks to aggro picking up again. Opt main is the other large change. I had tried it out before as a 4 of and wasnt impressed but now with only 3 it somehow feels much better, allowing you to setup tron easier and find those good one of inclusions. Spatial in the main is a concession to how the aggro creature decks are doing and it frees up sb slots.
In the sb my list differs significantly from what I see most often. I have a full 4 slots devoted to graveyard hate bc I believe dredge to be extremely scary with its new tools. The surgicals also come in vs tron and combo but crypt is more backbreaking and faster I find. Plus can recure it with ruins and tutor it with t-west. I see most people play relic or grafdiggers cage but i think these are a bit slow. Also I play aetherize for the go wide aggro decks like humans and dredge. This card is very strong and I think you should consider it. Thought-knot is another card I dont see much of and I'm confused as to why bc it is excellent against control, tron, and burn (some of the most popular decks in the format right now). Nezehal is my concession to U/W and cryptic command decks in general. You get to 7 mana cast it and win the game everytime. I think its worth 1 slot for a hard shutdown card against a really popular strategy even if the matchup is slightly favored.
Anyway this is the list I am most likely taking to the GP with maybe some minor tweaks here and there. Let me know what you think of my card choices, I have a reason for every inclusion and have considered pretty much any possible card you can play at one point and can explain why I decided against including it in this version. As always U-tron is a deck you can tune to beat what you want to beat and I think the most popular decks at the gp will be dredge, U/W control, humans, G-tron and so thats what I'm tuning against.
I had tried builds with high numbers of chalice and 2-3 karn and I found that while you do smash U/W with them, chalice is too weak against humans,dredge, other aether vial decks, ect which means you are extremely matchup dependendent g1 and you will lose a lot of game 1s to the chalice. Even in the matchups where is it good like tron and burn it doesnt win the game right out I find. Burn has about half 1 drops and half two drops and can draw the right half, tron can get natural tron ect. Nezahal, Primal Tide does win the U/W matchup right out. By the time you cast it you often have double blue because they have probably field of ruined you as well and it only takes up one slot in the board as chalice in order to be effective needs 2 slots in the 75 or more.
I havn't tested with Hieroglyphic Illumination but they both see two cards when you are digging for an answer or a threat which is what you do most of the time but opt does it on the one mana side and Illumination only sees 1 on the one mana side. If i were playing challice I would like Illumination more, I dont think you can play 7 one drops in a high challice list but I'm not playing that card. Being able to counter something on t3 and opt feels really strong.
For graveyard hate I think some surgical should be in every list as I find its the best way to beat g-tron, tutor a Field of Ruin or Tectonic Edge and surgical is usually GG. I don't bring in any gy hate vs the b/g/x decks, those are pretty good matchups anyway. I like to bring in dismembers, thought-knot, maybe spatial or crucible depending on what I see and that usually wins a lot of the time.
I think you make some decent points about Chalice, but the same arguments could be said about Condescend and/or Remand. These cards do little to nothing against Humans if they have Vial or Cavern of Souls out. Tron can usually just get Tron online (faster and more reliable than we can) and pay the extra mana. I don't think that means we should cut either of those cards, though. It just means that Humans will be a tough matchup that we need a prepared sideboard for. Summary Dismissal is great for Tron, and Silent Arbiter is great for the Humans matchup while also being strong against Dredge, Merfolk, Spirits, etc.
We can't be perfectly prepared for every other deck, and we will have bad matchups. It just so happens that we are better (preboard) in a meta that is full of decks that Chalice and our counters is good against. As for Chalice vs. Burn, it is very effective in cutting them off a lot of spells, and two Chalice nearly wins the get outright (see pierakor's most recent Burn video to see this in action).
Looking at it, it seems pretty bad. Its cmc is high and its blue cost is high for its cmc. It doesn't draw. What are we planning to counter with this? Ulamog or World Breaker seem like obvious targets, but outside of those, I really don't know. Storm can remand grapeshots (or just play more/empty), UW has better cheap counterspells and will probably win a counter war if we play a 4cmc counterspell... I don't even know what else this could come in against.
I guess it's pretty good as I've seen a lot of people running it, I'd just like to know why.
I can't definitely say why the card seems to correlate so well with win percentages, according to the data. We could go back and watch the videos on pierakor's channel and watch how it's used to see for ourselves. The spreadsheet is set up so that you can filter to see the games which had Dismissal in the opener, find what was played against and when, to make it easier to pick out the videos to watch. Otherwise, we're just running on conjecture.
If you take into considerations matches of 2-3 people, you’re running on conjectures too. Because that is data... based on a particular decklist and/or playstyle. We need better coverage, otherwise such a project can be certainly useful, but just until a certain extent.
I think for "unique" cards, you're correct. By this, I'm talking about cards like Torrential Gearhulk, Sundering Titan, Platinum Angel, Trinket Mage, etc. Cards that folks swap in/out based on personal preference and are the more toolbox-y cards, as opposed to the standard cards seen in every deck, with multiple copies (e.g. Remand, Condescend, Repeal, etc.).
This data, I believe, is more useful for those multi-copy cards. The one-of cards, not so much.
Reason being, as Bloody notes, play style and decklist more drastically affect the utility of those cards than any other, as well as matchups. Against a mono color deck, for example, Sundering Titan is less good and likely to get pitched to Thirst, not played at all, or played with little impact. Platinum Angel, likewise, against a removal heavy deck is far less good and likely to have less of an impact. For those toolbox, one-of cards, this data is less useful, because variables like: play style, meta,etc. play a greater impact on these cards than others. You're running only 1 copy, playing it less, and playing it primarily for niche situations, which don't trend towards quality statistics, without a massive effort to identify and account for far too many variables.
YES! Remand and other counters are awful against Vial / Cavern decks, but the deck runs 7+ counters, typically, and is a primary tool, regardless of decklist nuance. We can spin our wheels on arguing about specific decks and cards that are terrible against them, but that's what the sideboard is for. Every main deck has weaknesses and the sideboard is meant to augment those.
Regardless, because of all the variables (opponent decks, play style, human error, etc.), this data will never flawlessly depict what cards are "the best" and what the optimal decklist is going to be. However, I think the data is generally useful for showing us which commonly main-decked cards are better performers than others. Which we can then extrapolate to how many copies should we play, if any, and what gaps still exist with those winners played (i.e. How do you close the gap between the 80% win rate of card XYZ and 100%?)
@BloodyRabbit_01, I'm not sure that you're looking at the same spreadsheet that I am. Can you tell me what significant variance in playstyle might exist when 963 of the 970 games entered are by the same player? There are currently two people whose games are entered, and one of them has a total of seven games on there. Additionally, a change in playstyle seems a bit of a stretch. How many "playstyles" does a control deck that primarily controls the game through Condescend, Remand, Repeal, and Thirst for Knowledge have? There is some variance in pierakor's lists, but nothing significant enough to change the playstyle of the entire deck.
Now, if this is all a response to Summary Dismissal, did you do what I suggested and look through pierakor's videos and actually try to see how the card helped in the matches it was played? Or did you decide that your initial opinion on the card was sufficient and that seeing actual gameplay is inferior to your assumptions about how the card would perform?
As far as the data method itself, I'm not sure why anyone would feel the need to argue with it. First, it's been demonstrated as effective. The first deck that this method was used with ended up winning the Pro Tour. Second, even if there are some unknowns, we can compare the two methods available to us. Either we can use data and attempt to adjust our decks accordingly, trying to find out why some cards correlate better with wins when they're in the opening hand. Or we can assume that our conjecture and anecdotal evidence is enough and tweak our decks accordingly.
As for sample size, I suppose I should point out that each card's data is weighted for sample size. That problem was already considered and a solution was built in to solve it.
If you watch Pierakor v. Shoktroopa, there are significant differences to their respective play styles. This is something both have acknowledged in their own various videos and/or posts.
Nobody's attacking the overall data, as I believe we've all mentioned, it's extremely thorough and excellent to have. However, as with any data, it needs to be properly scrutinized in order to be properly applied.
For example, your Delta on Summary Dismissal is 4.84%. Meanwhile, the Deltas on both Karn, Scion of Urza and Gifts Ungiven are 32.92%, with positive Weighted %. If we use just that information, we should be jamming multiple copies of Karn and Gifts. Yet, your own deck list in your signature contains 0 copies of either. Similarly, Spatial Contortion shows a -0.03% Delta and 0 Weighted %, yet you run 3 between main and side. The data also indicate that Island is negative Delta and Weighted %, while many of the blue spells on the list are positives...
Data is great, but it doesn't provide a 100% clear picture of everything that's going on with a deck and/or the meta it's competing with.
The point with Summary Dismissal is, there are matches it can be excellent in and matches it can be totally useless in, just like many other cards. However, the data reflect this, imo, given that the Total Games with 0 copies is over 900, while the wins with 1 and total games with 1 are both under 40. It's used in niche situations against specific matchups. If you know which matchups those are and your particular play style and deck list have a high win % against them already, while your matchups against other archetypes is weak, you're going to want a different card.
In regards to summary dismissal it is a powerful card against green tron. The ability to cleanly answer ulamog and worldbreaker can make your post board games much more winnable. Its fine against storm but really its probably our single best card against tron. I've moved to a new graveyard hate package of 2 surgical 2 tormods and 1 cage. So far I have been having a lot of success against dredge. I agree with Rjawas that tormods is well positioned at the moment and should not be overlooked. I'm playing this 75.
I too am on the tormod's crypt plan, my paper meta has evolved to have a ton of dredge and the power to tutor crypt with tolaria west makes most of us have at least 4-5 virtual copies of it as well as our other graveyard hate (map tutoring tolaria west). It also having the ability to be recurred with academy ruins.
It seems quite locking, Givin us the time to set up our finishers. When it cracks hopefully we have plenty of mana to do whatever we want. What you think?
Too many risks that are not outweighed by the upside. On T3, you're looking at 3 lands, if not 4 on the other side depending on who was on the play, by the time this card comes into play. Even if the opponent only plays a land every 3rd turn after that, you're still not doing enough to interfere with their game plan using this card.
Remember, Modern is a T4/5 format and you just tapped out on T3 to play this. Leaving them open to play whatever they want on T3/4 (depending on who's on the play). Then, if you're playing against another blue deck, this has limited impact.
This card, imo, is neither fast enough nor unique enough to warrant any kind of inclusion or consideration. By unique, I mean it doesn't add an element of disruption that we can already more effectively/efficiently utilize through other cards or a method of disruption that is more efficient than what we already use. Spreading Seas, GQ, Field, Tec Edge, etc. provide better methods for disrupting the opponents mana. Additionally, our counters provide a more efficient and one-sided means for disrupting their spell casting ability too.
I took a break from U Tron after Spirits becoming a thing, and it seems like I was correct: it’s kinda of an abysmal mu for this archetype.
What do you think the optimal win condition is against Spirits? I notice you don't run Mindlsaver in your posted 75. Do you think that defeating Spirits on the creature plane is the appropriate path for our deck? Or, does it make more sense to utilize a Slaver Lock and Ugin, while controlling the board state with a heavier counters and removal package sided in?
It seems that the fundamental idea of UTron and its many one ofs is that it is a jack of all trades deck that can answer anything. In modern, that can be a good thing, as there is a very wide meta. However, as many of us have experienced, I think, that also means we're just too slow in certain matchups and couldn't get to those many silver-bullet one ofs in time to make them a worthwhile inclusion in the deck.
Looking at MTGGoldfish figures, Dredge, Humans, and Bant Spirits occupy over 18% of the meta and are the top 3 decks for overall representation. Grixis Death's Shadow, Hardened Scales Modular, and Infect represent another 8% of the meta. This means, you have a significant chance of seeing a fast, aggro-esque deck whenever you sit down at the table. These decks are the ones UTron typically struggles with because they move so quickly and utilize pieces that are tough to counter quickly enough to make a difference, thus relegating us to rely on our one of silver-bullets like Platinum Angel.
Therefore, some folks have split away from the toolbox style and moved towards a different version running a higher volume of what are perceived as the cards that provide the highest advantage against our weakest matchups, the decks I've listed that also happen to represent the highest share of the prevailing meta at this time. Hence, 3-4 Walking Ballista, multiple Wurmcoil Engine, and more board cleanup in the form of Repeal and Spatial Contortion, instead of Dismember which expedite us getting killed by a turn or 2, against decks that don't need any help with that.
The fun thing about UTron is that it can be tailored to your meta because it has about 15-20 flex slots, between the main and side, that you can tweak as you need.
Well after... I don't know how many years... I finally own a Snapcaster mage and also picked up a Karn, Scion of Urza and a crucible of worlds while I was at it. Still lack explosives (It's in UMA right?) and still only have one chalice. Its good having a stockpile of shocklands from the last time they were in standard but not playing a shockland deck in modern... lots of trade fodder.
I have been a bit out of the loop... whats this! data sweet data... *digs in*
How many pages back is an example list that was used to generate this? I see you had a few different versions during the process.
Well it looks like we really suck against the caverns tribal aggro decks (and hollow one decks).... Kind of average against easy to hate Aggro, Tron and Scapeshift. Mostly positive against midrange and crush control. Basically what we intuitively know...
It is pretty much a tenant of modern that you choose what you lose to.. its pretty clear that Piektor has chosen not to lose to affinity/Burn/dredge with the sideboard and control/midrange with the main. While Bloodyrabbit seems to be completely focused on not losing to fast aggro. I do worry he might have gone too far though, I am not sure its possible to rely on the deck's inherent advantages against slower decks you still need to interact with them.
What happens if we focus on beating caverns decks... does it seriously require the full 18 cards to beat them that Bloodyrabbit uses?
Modern sideboards typically are a whole bunch of one-ofs with overlapping uses for different decks to try to tackle the diversity in modern. Now Jeff Hoogland (who plays more modern than anyone else regardless of what you think of him) to effectively hate a deck you need at least 4 pieces of hate, the faster the deck the more hate you need to have a hope of finding them. I am going to accept that we do lack the early cantripping of Blue control decks and only power away with card advantage once we start casting thirsts. Which makes finding hate when it matters really hard, so we can't play a bunch of 1 of hate cards but I feel our 6 mana cards are pretty interchangeable because it's really quite easy to pull away on card advantage post 3/4 mana.