theory//practice

[mmf-cash]

feel free to support the text by buying a print copy or ebook: https://www.amazon.com/G-Y-G-Pharaoh-Games/dp/1536983241

[mmf-cash]

GoatStats I guess I should go ahead and put a thread up for this. I'm working on compiling a database of Goat Format tournament statistics, in t...

[mmf-cash]

It's been more than two years since my last serious tournament report. Let's get it. I did notoriously terribly in season 1/Summer 2017 and neve...

[mmf-cash]

a brief history of chaos control

in the early Revival era, back when everyone played with Exarion Universe and most deckbuilding efforts were coming out of the Nostalgia Duelist Facebook group, there were two decks at the forefront of the metagame: traditional Goat Control, which had remained more or less the same since US Nationals all the way back in 2005, and Chaos Turbo, an almost entirely new deck featuring Thunder Dragon, Card Destruction, Dekoichi, Night Assailant, and multiple Chaos monsters (typically BLS and three Chaos Sorcerer). none of these cards were unheard of in 2005, but they were not streamlined into this kind of shell until the Revival.

[10:28 AM] 🐸: a fundamental of goat used to be taht thunder dragon was a -1

[10:28 AM] 🐸: nice fundamental

while Chaos Turbo was popular, it was not actually very good. it played a lot like a more linear Goat deck, maximizing the use of Tsukuyomi + Flip effect locks and the infamous pieces of the Trinity, and cutting clunky stuff like Airknight Parshath entirely, but it had a much lower ceiling than Goat Control itself. If you were bad at playing Goat Control, you could increase your winrate by playing Chaos Turbo, but if you were good at playing Goat Control, the deck did not offer many theoretical matchup advantages.

so, when the first DGz Goat format war league took place, the final matches were, predictably, all Goat Control mirrors, and from there, the rest is history: from the dust emerged Kris Perovic’s list, which became the golden standard for years to come.

in 2017, the second generation of DGz Goat warring breathed new life into this deck, and the efforts of a few duelists brought it up to speed with the new pre-Exarion landscape. gone was the linear reliance on cards like Card Destruction and Phoenix Wing Wind Blast. the new lists, dubbed Chaos Control, had adapted by adding Scapegoats and Metamorphoses of their own, but the new developments did not stop there. videos from the previous year of Kris showing off his spiffy Recruiter Chaos deck in casual duels with Allen Pennington led many to adapt to the new prevalence of Dekoichi with their own Shining Angels and Mystic Tomatos. Chaos Control lists branched off into two directions: one retaining the Thunder Dragons of old (colloquially named Thunder Dragon Chaos), and one replacing them with the new hotness in Shining Angel (called Angel Chaos).

[6:56 PM] kperovic: This is more of a guide than a rule because it's possible to construct a deck on the edges of these definitions but in general IMO re Sorc:

If you have Sorc and Meta, it's Chaos Control whether or not you have Recruiters. If you have Sorc and Recruiters without Meta, it's Chaos Recruiter. If either deck runs more Return than Meta, I'd probably call it Chaos Return.

If you have Sorc, Thunder Dragon, and no Meta then it's Turbo Chaos. If you have Sorc and Meta then it's most likely Chaos Control whether you have Thunder Dragon or not.

If you have just a tech Sorc in an otherwise standard Goat deck with multiple Goats and Meta then I'd probably call it Goat Control still. Bump that to 2 Sorc and I'd probably call it Chaos Control.

when two of the top three players in the league were on some version of Recruiter or Angel Chaos at the end of the season, it was clear that something had changed in the format. Recruiters, and particularly Shining Angel, were now legitimate metagame threats that demanded to be answered by the other decks, whereas previously they would have been seen simply as food for Thousand-Eyes Restrict. Thunder Dragon Chaos had tanked in popularity while maintaining respectable results in the hands of the few competent players that stuck to it, but Shining Angel was undoubtedly the main character and star of the show in season 1 of the new DGz Goat format war league.

[3:48 PM] Ynusgridorh: Turbo's goal is to recycle Graceful and Card Destruction with Faith and recycle Faith with Night Assailant. It also maxes on Raigeki Break or PWWB to make good use of Night Assailant and Thunder Dragon. Control plays like a regular Goat Control deck by tributing Sinister or tokens for TER and using Tsuk to set up soft locks.

each of the new Chaos branches came with a unique set of pros and cons. With Shining Angel, the name of the game was versatility, while for Thunder Dragon, it was raw power. Shining Angel can become a pseudo-Pot of Greed in almost any situation if used correctly, either by gaining advantage through battle then forcing an answer or simply baiting the opponent into doing something bad, but winning a game off of a single Graceful Charity activation while blanking opposing Duos until said Charity activation is a privilege unique in the format to Thunder Dragon and Thunder Dragon alone.

[2:15 PM] ACP: the reason why angel chaos is so good is because its early game options are the most numerous in the format

[4:54 PM] ACP: shining angel attacks well, defends well, can be a meta target for TER, trades very well with dekoichi and other recruiters, fills up your grave, has insane combos and mind games with tsuku/book

[4:54 PM] ACP: the card does actual everything

[10:57 AM] Ynusgridorh: Advantages of TD: 1) nerfs Duo 2) combo with Graceful 3) combo with Tribe 4) can tribute a snatched monster 5) can tribute TER 6) more reliable light fodder

[1:43 AM] MMF: angel rewards good reads and as allen would put it, "mindgames"

[1:44 AM] MMF: thunder dragon doesn't make you do any of that shit

[1:44 AM] MMF: to get your benefits from tdrag, you just draw it and discard it from your hand, done

[1:44 AM] Jazz: right

[1:44 AM] MMF: you sit on it forever, or you dump it for the charity you already have

[1:44 AM] MMF: no reads required, no effort required

[1:44 AM] MMF: allen sees this as a bug while i see it as a feature

the original hyperaggressive style of Recruiter lists initially showcased by Kris is, ironically enough, used exclusively by members of ALL THE OUTS today, first Morphing Jar, and then Chevalier de Fromage. these lists are completely distinct from both variants of Chaos Control, and are outside the scope of this post, but are still worth mentioning as further proof that Chaos Sorcerer is the new face of Goat format. some standard Goat lists are even playing a single copy at this point. if you aren't playing Chaos Sorcerer, what exactly are you doing with your life?

[mmf-cash]

soul control in goat format

it’s no secret that i think soul control is one of the coolest ygo decks of all time. it’s also no secret that it’s not exactly popular in goat format; not then, not now. 

soul control is something of an antithesis to goat control. before the battle position ruling change that pushed scapegoat and metamorphosis over the top, soul control was a cutting-edge answer to stuff like zombies and chaos control (the proto-goat decks that played similar monster cores, typically with 2 scapegoat and 1 metamorphosis, not to be confused with the decks we now call chaos control that typically play multiple copies of chaos sorcerer), but when multiple mained tsukuyomis became the standard, playing a monarch-based deck meant that you were weeks behind the curve.

so, how does the deck survive in this new environment? well, for a while, it didn’t. sandtrap himself switched off the deck sometime after nationals; by SJC Indy in August, he was experimenting with gravekeepers, resulting in this tragic/hilarious match. 

things only got worse for soul control during the exarion-based goat format war league in 2014, when thunder dragon chaos decks began to appear (these decks were completely nonexistent in 2005). in this matchup, the core combo of thestalos+soul exchange going second would meet an opposing thunder dragon or sinister serpent in approximately half of your games. suddenly, the two most popular decks were incidentally soul control’s two worst matchups.

when jazz made his format library website in 2016, the hot new idea for monarchs was “aggro monarch,” a deck that cut soul exchange entirely in favor of brain controls (which weren’t legal when sandtrap topped with the deck in April). further additions included three trap dustshoot to combat tsukuyomi, a standard suite of 3 scapegoat and 3 metamorphosis, and a shift from thestalos to mobius and granmarg, the latter being somewhat of a trademark in “modern goat monarch” decks. 

during the 2017 reboot of the dgz war league, another new variant based on stray lambs and guardian sphinx began making sporadic appearances. this version, in my opinion, dealt with the “tsukuyomi problem” much more eloquently than the hamhanded use of dustshoot in “aggro monarch,” but it did so at the cost of introducing new weaknesses to nobleman of crossout (which can be a perfectly acceptable tradeoff).

detox and espn independently tested monarchs as a direct counter to morphing jar’s hyperaggressive recruiter chaos deck in the later weeks of the war league season. three sakuretsu was key here, as much for tsukuyomi as for shining angels, mystic tomatos, and pretty much anything else that can attack. as recruiter-based chaos decks effectively supplanted the thunder dragon decks in the new warring metagame, soul control swapped out one of its worst matchups for a solidly positive one.

so, can soul control compete in goats in 2017? there’s no way around it: in the majority of your games, you’re going to be playing a losing matchup, and you’re going to have to accept it. but, if this recruiter shit ends up not being a fad, this should be in the back of your mind when you’re going to the drawing board on how to combat recruiter’s latest incarnation.

a question wrestled with in both lines of development is the proper ratio of monarchs to soul exchanges.

the original sandtrap list played 4 monarchs and 2 soul exchange, for the following turn 1 odds:

12.37% to open with at least 1 soul exchange and at least 1 monarch

7.89% to open with no soul exchanges and at least 2 monarchs

15.71% to open with at least 1 soul exchange and no monarchs

we want to maximize the first number and minimize the second and third. with these numbers, we’re opening with the combo about half as often as we’re opening with dead pieces (we can add the second and third numbers together, since they cannot occur at the same time, to get a general idea of “how often we brick”). if we follow the trend in the format library list and move up to 7 monarchs, but retain soul exchange over brain control, we have the following odds:

26.05% to open with at least 1 soul exchange and at least 1 monarch

19.11% to open with no soul exchanges and at least 2 monarchs + 13.39% to open with at least 1 soul exchange and no monarchs = 32.50% to “brick”

if we remove a single monarch from this point, we end up with...

23.58% to open with at least 1 soul exchange and at least 1 monarch

14.82% to open with no soul exchanges and at least 2 monarchs + 15.86% to open with at least 1 soul exchange and no monarchs = 30.68% to “brick”

... while cutting a single copy of soul exchange instead leaves us at...

18.40% to open with at least 1 soul exchange and at least 1 monarch

21.75% to open with no soul exchanges and at least 2 monarchs + 9.67% to open with at least 1 soul exchange and no monarchs = 31.42% to “brick”

in retrospect, it’s clear that any less than the full set of soul exchange is unacceptable. next question: can we do better than 3 soul exchange and 6 monarchs by cutting yet another monarch?

20.75% to open with at least 1 soul exchange and at least 1 monarch

10.73% to open with no soul exchanges and at least 2 monarchs + 18.86% to open with at least 1 soul exchange and no monarchs = 29.59% to “brick”

the answer: a resounding no. in my opinion, the best “ratio” here is 3 soul exchange and 5 monarchs. this leaves us with the question of which monarchs we want to play. thunder dragon has become virtually nonexistent in favor of shining angel, so thestalos seems like an auto-include at 3. this leaves us to fill the remaining 2 slots with some split of mobius, granmarg, and zaborg, unless we want to look into techy stuff like jinzo or guardian sphinx.

[mmf-cash]

on style

suppose that we have two strategies to choose from. the first has an expected return of 100 with probability 0.5. the second has an expected return of 50 with probability 1.

in theory, we are indifferent between these two options. in practice, we are not. an individual irrational preference for one of these strategies over the other can be stated as a “style” of the individual in question.

[mmf-cash]

dsadfhds

1. the good, e.g. utility, is that which reduces the opponent’s options.

2. from 1, it is not necessarily evident what we should call the “bad.” we will inevitably assume, however, that some “bad” does indeed exist.

3. theory can be stated as the sum of all human attempts to determine what this “bad” is for a given ruleset or game.

[mmf-cash]

lots of rust, we both suck, etc

1. edgeguarding not consistent enough. sliphogging and rollhogging should be more than enough for every conceivable situation

2. get comfortable with pivot desync in neutral again. get comfortable doing things other than dsmash with ledge invincibility

3. need to revamp my approach to platforms. my uairs are not smart and my platform movement is not confident

4. allen has a good understanding of CC/ASDI’s impact in the matchup and doesn’t give me very many easy CC->punishes, but he doesn’t quite have the shield pressure part down, which leads to me getting a few randy grabs-that-shouldn’t-really-be-grabs off of super early aerials

5. allen needs to go for popo instead of nana when he’s down 2 stocks imo

[mmf-cash]

10 things not to do with solemn in goat format

don’t use solemn on an even board as a bottomless trap hole/sakuretsu armor with LP cost attached. if they’re adding the first monster to the field and your only answer is solemn, you should be thinking about how you will push back with your own monsters+solemn protection, not solemning the monster and hoping your own summon goes unanswered.

don’t use solemn on cards like breaker, noc, or faith. corollary: try your hardest not to put yourself in a position where you would ever need to use solemn on breaker/noc/etc.

the PEOPLE’S SOLEMN is when you solemn their metamorphosis tributing sinister serpent, then kill them or put yourself very far ahead on your following turn. the PEOPLE’S SOLEMN is a storied technique of high risk and high reward that is neither for the light of heart nor the faint of butt. go for the PEOPLE’S SOLEMN when you have something for a followup chaos monster (snatch, BLS, meta) and you’re putting them on light/useless backrow (nothing that you would need the solemn for pushing snatch/BLS/meta etc thru).

early solemns are best with perfect information (dustshoot) and aggressive hands that you can tell are aggressive from the beginning (multiple normal summons, nothing like goat/meta or tsuk/flip eff in sight). if you solemn in the first few turns, try to at least have an idea of how you want the next few turns to play out and what you’re going to do to make that happen.

opening with solemn and multiple backrow does not mean that you should automatically set solemn along with all of those backrows on the first turn. don’t give your opponent more information about your options than you need to. giving your opponent an easy trade of heavy storm for 1 solemn and 4000LP is a great way to lose games without giving yourself a chance to play.

if you see your opponent playing solemns and you have one of your own while they have backrow, try to get in the habit of saving a solemn for one of theirs. it won’t always be the right play, but it’s an important thing to pay attention to on both sides of the board.

familiarize yourself with the best and most popular counterplay options to solemn. try not to get caught in games 2 or 3 with solemns in deck against their noblemen of extermination/decrees. corollary: realize that you might not even want solemns against these decks in the first place, regardless of the presence of extermination/decree.

eat, breathe, and shit the game 1s of this feature match and this feature match. read them in the morning before breakfast. read them before you go to bed at night. try to understand every possible thought and consideration that went into the solemns played in these game. these are the feature matches that defined an entire generation of competitive YGO.

the best way to mitigate the life cost inherent in solemn is to wait as long as possible before playing it. the closer to the end of the game that you activate solemn, the less the LP payment will “matter” on average.

stop adding 3 solemn judgment and 3 trap dustshoot to standard goat lists. if you’ve processed all of the other nine points above, this should require no further argument.

[mmf-cash]

dgz goat warring: understanding the recruiter meta

the most base-level interpretation of the data shown here is that chaos recruiter is the best deck in goat format. however, we know that this is not actually the case. in fact, since cards are costless in warring, all matchups are theoretically even in the long run.

i feel like i’ve repeated this story ad nauseam across various discord chats, but i’ll summarize it here for good measure: recruiter originally arose in the dgz warring metagame as an answer to the triple-dekoichi triple-thunder dragon chaos decks popularized by Jazz and circulated mostly within smaller social media groups like nostalgia duelist. the idea was that a recruiter was the best answer to a flipped dekoichi, since you would crash your tomato into the dekoichi, get a spirit reaper from your deck, attack directly and rip a card out of their hand.

it quickly became the de facto “deck to beat” in warring when the two players shown below rose to the top of the ranks with it:

two of the top 5 players in the rankings almost exclusively play chaos recruiter. their results are shown above. matches shown are only those that were played with chaos recruiter (in both cases, the vast majority). against chaos and beatdown, these two are completely undefeated, and against goats, they have a single loss between the two of them.

coincidentally, that loss was to one of the two “goat mains” in the top 5. while recruiter is currently “beating” goat if you look at results across all levels of play, it is only going even with goat players in the top 5. in other words, the best goat players are capable of beating recruiter consistently enough, but the majority of goat players are not.

in my opinion, recruiter’s worst matchup in theory should be chaos control (aka thunder dragon chaos), and i believe this is also reflected in the data. between the two of them, ACP and Mustang have only played against this deck a single time, and a single win for Allen over Silver doesn’t make a very strong argument for recruiter in the matchup. these decks adapted quickly to the threat of tomato/reaper by shaving copies of dekoichi in favor of dark mimic lv1, which creates the same combos with tsukuyomi as dekoichi without being a liability against recruiters (and makes ter with metamorphosis, go figure). with the reliance on dekoichi gone, all that is left is a deck that is completely comfortable with drawing and passing for the first 20 turns of the game against chaos recruiter while the opponent simply digs their own grave deeper and deeper.

here is a screenshot of one of the early lists from kris perovic that eventually evolved into the lists that are prevalent on dgz now (watch the deck in action here!):

for most of the first month of warring, recruiter lists shifted quickly from hyper-aggressive triple-solemn-triple-dustshoot lists in the vein of perovic’s to warrior beatdown lists with blade knight and don zaloog before finally settling on goat control hybrids with multiple copies of metamorphosis and scapegoat. another consistent trend is found in the average number of recruiters played per list: around may 1st, this number hovered between 4 and 6, but it has since dwindled all the way down to a consistent 3. creature swap has become almost entirely nonexistent. “wombo combos” with swap and/or reaper have become far less common as the goat and chaos decks have adapted; since around the end of May the recruiter decks have had to play a far more honest game against these decks, mainly with scapegoats and metas of their own.

an example of the latest incarnations of the dgz recruiter deck might look like this (disclaimer: i have played zero games with this list):

when the recruiter deck is framed in this way, the crucial nature of the relationship between recruiter decks and what is being called “chaos control” becomes obvious. if Allen is right that recruiter is a midrange deck, thunder dragon chaos is the pure control deck.

the dynamic between thunder dragon and shining angel will be a core feature of the metagame to keep an eye on in weeks to come.

[mmf-cash]

dgz goat warring: last week tonight (1)

i compiled this data towards the end of may and came up with some interesting results. please note that the data shown is by no means up-to-date, but most of the trends shown here have remained more or less constant since compilation on 5/23.

the first thing to note is the extreme overperformance of chaos recruiter. this will be explored in more depth in future posts, but for now, we will simply say that the best two players in the league right now have both almost exclusively played recruiter, and the results are skewed accordingly.

including burn was a mistake due to the lack of data. all that is proven here about burn is that its first five opponents were relatively unprepared.

early data suggests that chaos turbo/control variants have the best matchup against recruiter. chaos variants on the whole still need far more data points in order for us to be able to draw substantial conclusions about their relative metagame positions.

[mmf-cash]

performance taken personally

i think ascribing a "natural" quality to the phenomenon we typically call "talent" is making the same mistake as people who consider themselves "lucky" or "unlucky" in poker or some other variance-inclusive game. if luck is "probability taken personally," as we like to say in poker, i think talent is something like "performance taken personally."

luck does not exist; the perception of luck exists. talent does not exist; the perception of talent exists.

why, then, do some always tend to rise to the top of the pack in competitive activities?

the answer is clear: they have to.

[mmf-cash]

let’s talk about ice climbers vs jigglypuff

why do hungrybox and chu dat camp the shit out of each other?

let’s start more generally. why does any character camp another character at all? what does it mean for it to be optimal for one character to commit to not approaching the other?

we refer first to the landmark don’t approach:

not approaching is the biggest short-term improvement you can make to your game. this strategy is THE REASON most top placers at nationals are where they are; it's that good. i firmly believe that theoretical top level melee is a game of first hit wins, just because whoever lands the first hit never has to approach their opponent.

approaching becomes bad when it becomes difficult or impossible to ensure that an approach option will be “safe” against the diverse defensive toolkit available to the player getting hit in Melee (DI/SDI, CC/ASDI, fast OOS options).

we can optimally camp a certain character when all of this character’s “approach options” are one or more of the following things:

consistently slow or otherwise reactable to the point that “stuffing” their approaches can, with some amount of practice, be made into a matter of pure execution as opposed to interaction

excessively frame-negative on shield to the point that a blocked approach will always equate to a punish for the opponent

vulnerable to crouch cancel/ASDI down such that the character cannot effectively deal with a character holding both sticks down

now we have explained what makes camping possible in a particular matchup, but we have yet to speak on exactly why it works. the short answer is as follows: camping works because if performed correctly, the camping player will only ever need to win one neutral interaction to guarantee a kill.

to the vacuum theorist in melee, all interactions between characters can be reduced to the neutral/punish dichotomy. in this situation it is as clear as ever that there is a fundamental “trade-off” or deadlock of sorts between frequency of neutral interactions and magnitude of punishes (openings per kill).

when Puff is floating in the air, Ice Climbers’ “approach options” are limited pretty much to the aerials that can hit them with about a fullhop’s worth of distance. while these are not particularly slow aerials (nair/uair min f9) they involve a large vertical motion which is easy for Jigglypuff to recognize and react to for the same reason that Falcon’s nair across the stage on BF is typically easy to react to and stuff.

Puff’s “approach options” are not particularly weak to crouch cancels (bair breaks ICs CC at 76%) or shields (most of her aerials are -3 if performed late and lcancelled). However, for Puff, the very act of landing entails a rather significant penalty of around 10 frames of “down-time” in which she cannot put out a non-rest hitbox (4f normal landing lag + 5f jumpsquat, 5f before jab/dair hits, 6f before nair hits, 9f before bair/uair hit). if Puff runs out of jumps and must land during an approach sequence, she is therefore at a significant disadvantage (think about how annoyingly fast ICs can cover horizontal space).

in a “terminal case of campiness” such as this, the the goal of the game is no longer to reduce your opponents stocks to zero, but to reduce the clock to zero while you are ahead in stocks/damage.

ergo, hungrybox elects to spend as much time as possible (on PS) vs chu dat

ahead of chu dat in stocks/damage

unthreatened by a wobble

if hungrybox maximizes the amount of time that these two things are true, he should be most favored to win against chu.

do you need to ledgecamp with puff in every matchup? no. should you? no, actually. i actually think that ledgecamping against fox hurts you in the matchup. i think ledecamping against almost every character... maybe except falcon, but even that’s iffy territory... If you ledgecamp against Marth, it’‘s not good. ledgecamping against Sheik isn’t good. ledgecamping against Fox and Falco isn’t actually good... it’s beatable, and it’s not a full exploit...

hungrybox expanded a lot on his mindset against chu dat in grand finals of DreamHack Austin on his stream. on the one hand, we have chu dat hugging the space underneath side platforms, which hungrybox identifies as correct, since puff is easiest to grab and least safe when approaching from cardinal horizontal directions.

you see how whenever chu hits me he goes and retreats to the platform underneath it?... he’s forcing my approach because i’m losing and i am weak when i am on the ground...chu dat stays underneath the platform because when puff approaches, she’s weaker... she’s easier to grab. that’s exactly what you should do...

and then on the other hand, we have hungrybox playing to the timer by keeping sopo alive for as long as possible. this is not the first time that this has come up in the matchup, even in recent memory; any IC that has had the poor misfortune of running into Michael on netplay should be able to attest to this.

expect this aspect of the matchup to become even more prevalent at tournaments that experiment with lower game timers such as 6 or 7 minutes.

[mmf-cash]

first impressions on rivalries data

we’re a couple weeks on the tail of yahoo’s smash rivalries event, and for those not in the know, the game statistics shown on stream in between sets have been made available thanks to fizzi the stats guy.

i am still working on compiling matchup-specific data, but there are already a couple of things that stand out to me. the first is that jigglypuff is almost impossible to edgeguard (per the accepted definition). she almost always floats back to the stage outside of hitstun and has to win some kind of drifting/spacing mixup or whatever other kind of “keep-out” bair-based line of play that the spacie will substitute for a traditional edgeguard. it is worth pointing out that hungrybox played almost exclusively against Fox for the entire tournament, but we can imagine the same would probably have been true had he played against Sheiks, Marths, and Falcos instead.

despite his placement, mang0 represented falco well, showing that the character’s core neutral and punish tools can still keep up in the contemporary metagame. unsurprisingly, chu dat had by far the highest average punish rate in the entire tournament, but his neutral game fell just a little short of the top 4 characters, with a neutral win % more comparable to Falcon or Samus than Fox or Falco. once again, the data for Jigglypuff is warped by virtue of hungrybox being so much better than everyone else, but another thing stands out in spite of this: not only is his character essentially immune to edgeguarding in the traditional sense, but his edgeguards were also far and away the best and most consistent of the entire tournament.

it is interesting that Falcon and Ice Climbers were the only characters that could contest Fox’s overall rate of center stage control.

[mmf-cash]

The end is important, the means is the means. Combo trees that don't end with a stock can be uprooted.

Alex’s Puff Stuff

[mmf-cash]

tier lists and the average prize model

i have written elsewhere on the application of the average prize model (APM) to the notion of tier lists in melee. these articles provide a framework for tier lists which are deductive, objective and verifiable relative to the following factors:

a complete set of theoretical matchup data between every character and every stage

complete information on the prize structure of tournaments and the character demographics of each tournament

the smash.gg year-of-end reviews gave me an opportunity to prove the viability of this method. i plugged every single matchup ratio between top 8 characters presented in those articles, along with a few that i had to find on my own, into the APM along with the following data:

average character usage at top 100/all levels in 2016: the average national-level event in 2016 had 680 entrants. APM simulations with data across all levels used character usage stats pulled from smashboards, while the simulations with top 100-level data used character presence proportionate to SSBMRank 2016 representation.

average prize structures: i used a rough estimate of $13,572.52 as the average total prize pot of melee tournaments in 2016. this yielded the following prize structure: 1st = 6146.25, 2nd = 2673.84, 3rd = 1464.45, 4th = 861.01, and so on down to 7th/8th place.

since the APM uses swiss/single-elimination top cut as opposed to double elimination all the way through, and since the smash.gg data comes with obvious grains of salt, we must restrain ourselves from drawing hasty conclusions from the following results. that said, the top charts below display average and weighted average top 8 matchups for every character, and the bottom charts show each character’s average wins and cash in a 680-person 16-round single-elimination top 8 cut tournament:

we may split hairs over whether or not Puff is in a “tier of her own.” if she is, then Peach, Fox, and Ice Climbers clearly constitute a tier below her across all levels, while Fox is the only other viable character at top 100-level, where Peach simply cannot contest Puff’s 100% winrate.

while we may be inclined to say that Puff's results are artificially inflated by Hungrybox overperforming relative to the character, this does not give the full picture. we should instead say that Puff's metagame is overcentralized on a single player, and this comes with both upsides and downsides for Puff as a character; in an organic metagame, an underplayed character is punished more for individual losing matchups with a lack of 50-50s to "even out" their results, but in turn enjoys more polarized matchups at higher levels as the players of other characters suffer from a relative lack of preparation. it is difficult to contest Puff's dominance in these metagames since it is worth so comparatively little for a character to have a good matchup against her (as opposed to a good matchup against Fox, Falco, Sheik, Marth, etc, all of whom everyone plays against in tournament far more often than Puff).

[mmf-cash]

mythbusters: “which matchups in the top 8 are ice climbers winning” edition

it is a common misconception that ice climbers go even or worse with sheik and lose solidly to everyone else in the top 8. as if to add insult to injury, some will even include samus in these 8.

these win ratios are pulled from smash.gg’s 2016 year-end reviews, which can be found on their blog here. i have filled out the top 100-level stage data for falcon and jigglypuff myself, as these were not shared in their articles. “practical” refers to the recorded winrates on smash.gg. “theoretical” refers to the ideal ratio relative to the recorded stage data calculated here.

the most jarring piece of data is peach’s overall losing winrate (not everyone can be armada). even against top 100 opponents, ice climbers had positive matchups against fox, falco, sheik, and marth in 2016. not exactly a stellar showing for what was supposed to be the year everyone learned their ice climbers matchup.

note that ice climbers did not lose a single matchup in this list on final destination, and on the flip side, they did not go better than even with anyone on yoshi’s story. i think the dynamic between these two stages plays a large role in the perceived “polarizing” nature of most matchups involving an ice climbers player. peach provides an extreme example: ICs won a third of their reported games against top 100 peaches, but lost every single time to the same peaches on yoshi’s story.

by the data, sheik’s matchup is paradoxically the most even. ICs and sheik split the six stages almost completely equally: two 80-20s, two 50-50s, and two ~75-25s. the only reason sheik loses in this case is because her margin of victory on BF/DL is smaller than ICs’ on FD/PS.

i don’t mean to imply that ice climbers actualy have theoretically winning matchups with all of these characters. i suspect that sheik and marth are both even in the long run, for example, along with captain falcon and falco. but the data alone should at least be sufficient to get people to stop talking about ice climbers like a low tier with no matchups they could possibly be winning. in fact, they are winning quite a few of them right now.