TopDeck 是做什么的？
MTGA 上线 LCI 系列之后，限制环境是怎么样呢？发生着什么样的演变呢？我们尝试结合 17 Lands 的数据，以及深度挖掘和 AI 模型一点一点的挖（po）掘（jie）环境。
好啦，介绍完我们的工作，就开始进入正题吧~欢迎关注我们的公众号：TopDeck，我们会时刻关注限制环境，带来更多的分享。
本期目录
01. Sierkovitz 大神介绍
02. 文章概览
03. Speed Impacts Plans
04. Fast Decks
05. Midrange and Control
06. Conclusions
这是一篇搬运 + 翻译稿。原文是 Sierkovitz 在 12.20 发布于 MTG Arena Zone 上的《Finding Plans in a Draft Format Using Data》。
特别声明：刚开始尝试做 Pro 内容翻译，难免错误和理解不到位，特别欢迎指正 & 讨论。
01 Sierkovitz 大神介绍
Sierkovitz 是一位万智牌数据处理方面的大神，他从 2022.7 开始在 MTG Arena Zone 上发表一些基于数据分析推导出来的万智牌思路或攻略。
Sierkovitz 除了在 MTG Arena Zone 上发表文章，还在 Youtube 上有自己的播客（Sierkovitz）。Marshall 和 LSV 主理的播客 Limited Resources 从 2021.6 开始会定期邀请 Sierkovitz 参与他们的节目。
Sierkovitz 自己的播客和 Limited Resources 最新一期的节目都对本篇文章的内容做了更加细致的讲解。有兴趣的朋友请自行前往。
那现在让我们回到这篇文章中来：《Finding Plans in a Draft Format Using Data》
02 文章概览
The success of your deck in drafting depends on whether it can achieve its intended game plan! Make sure your decks reach its full potential with this Limited strategy guide.
轮抓套牌的成功依赖于你是否能执行游戏计划。希望这篇文章的思路能帮助你把套牌的全部潜力发挥出来。
One of the important skills in Limited is realizing what a given color pair wants to be doing and then figuring out if it is good in doing it. Some color pairs want to be more aggressive, some are more midrange and some do better in a control setup. Most of the time realizing it is intuitive in broad strokes – we know that Izzet (UR) deck in The Lost Caverns of Ixalan (LCI) tends to be on the aggressive side. But broad strokes will miss some subtleties that diving deeper into data can answer.
Let’s look at the LCI format more in detail and based on 17Lands.com data, try to see which decks do better when they are aggressive, which decks prefer to play a long game and – probably most importantly – check if they can successfully play the long game.
在限制环境中，重要的技能之一是意识到不同色组想要做什么，然后判断它在这方面是否真的擅长。一些色组倾向于快攻，一些更倾向于中速，而另一些则偏控制。大多数情况下，这种判断来自直觉，例如：我们知道伊捷（蓝红）套牌在 LCI 中往往更偏向快攻。但是，粗略的判断可能会忽略一些细微之处，而深入数据分析则能解答这些问题。
让我们通过 17Lands 的数据仔细研究一下 LCI 环境，试图找出哪些套牌更适合快攻，哪些更倾向于打后期。最重要的是看看它们是否能成功地打后期。
03 Speed Impacts Plans
速度影响计划
Speed of the format is a term that is used frequently but from my experience, it is often thrown around in a way that is not very meaningful. Just saying that the format is fast or slow may seem like informative, but it can be more detrimental to the understanding of the game than helpful. There are many reasons why games in a format last only a few turns. One, perhaps the most obvious one, is that the cards are statted to be aggressive, thus leading to early start of the game, racing and shorter games.
根据我的经验，「环境速度」这个词经常被使用的比较随意，但缺乏实际意义。仅仅说环境的“快或慢”可能看起来很有信息量，但实际上可能对游戏的理解有害而不是有益。有很多原因会导致某种环境的对局仅只持续较少的几个回合。其中，最明显的原因之一是卡牌被设计的具有进攻性，因此导致战局展开的更早、对着抢血，以及较少的对局回合数。
But aggressively statted creatures if they all trade early can also lead to slow games, where both players lose all their resources early and spend mid to late game topdecking 2/2s and taking ages to win the game. What makes formats fast is a snowball nature of board states. A situation where once ahead, opponent is on the backfoot and has to trade inefficiently just to stay alive, meaning they can never get good value from their cards, while the advantaged player – does.
但是即便如此，如果所有攻击性较强的生物在早期都交换掉了，也可能导致对局进展缓慢。双方玩家都会在早期耗尽资源，并在中后期不断抽到 2/2 生物，然后需要很长时间才能赢得比赛。
所以，使得游戏快节奏的根本因素是「滚雪球效应」。一旦领先，对手就会陷入被动，不得不进行亏本的交换来维持生存。这意味着他们永远无法从自己的牌中获得良好的回报，而处于优势的玩家则可以。
This is best seen in graphs that show the impact of spending more mana than the opponent. In the one below you can see the win rates of 1000s of games in LTR format, where I looked at how much more/less mana did the player who was on the play spend until a given turn than the player on the draw did:
在以下图表中，你可以看到 LTR（魔戒）环境的数千场比赛的胜率，我观察了先手玩家在某个回合之前比后手玩家多/少花费了多少法术力。这最能体现在图表中，展示了比对手花费更多法术力的影响。
If the player spent the same amount of mana until the end of the full turn cycle (so both players finished their Xth turn), in the early game, till turn 7, the player on the play has a small advantage in those games (they win 1-2% points more than on average). After turn 7, that advantage disappears. The advantage of being on the play is covered with abundance by the fact that the player on the draw had access to one extra card.
假设对局双方都进行完了相同的回合（即双方都完成了他们的第X个回合），且都花费了相同数量的法术力。那么在第 7 回合之前的游戏中，先手玩家会稍微占据优势（胜率比平均水平高 1-2%）。第 7 回合时基本持平。第 7 回合之后，这种优势就会消失，先手的优势会被后手玩家多抓了一张牌所弥补。
But the interesting part of this analysis is looking at the win rate differences when one of the players outspent the other. Even as early as turn 2, if the player on the play spent 1 mana more, they have 4.1%p higher win rate than if they were in parity. And conversely, if they spent 1 mana less, their win rate is 4.9%p lower than the average. Those numbers are specific for LTR, but the trends are true for every other format I analysed with this method. And same trend as for mana is seen for life and cards drawn.
However, the datasets I analysed were all contemporary sets. And the longer I look at those results, the more I think that they are a symptom of the modern set design. Cards do a lot for their mana cost. 2-drops are relevant threats. 1-drops are best they ever have been. Every creature has some additional effect or is an absolute stat monster. This means that felling behind early is hard to overcome and leads to the snowball effect I was describing.
这个分析的有趣之处在于观察玩家在法术力花费上的差异对胜率的影响。即便在 2 回合，如果先手玩家多花费1点法术力，他们的胜率比使用同等法术力时高出 4.1%。相反的，如果他们少花费 1 点法力，胜率则比平均值低 4.9%。这些数据是针对 LTR 的，但这种趋势在我用这种方法分析的每个环境都是一致的。而且，对于生命值和抓牌数量，也存在与法术力相同的趋势。
我研究的数据集都是关于现代万智牌环境的。而且越是深入研究这些结果，我越觉得它们是现代万智牌设计的一种表现。卡牌的价值与法术力费用成正比。2 费的卡牌构成了实在的威胁。1 费的卡牌则达到了前所未有的强度。每个生物都带有附加效果，或者具有超模的身材。这意味着在早期落后，就很难逆转局面，也就是我所说的滚雪球效应。
This is my hypothesis, so treat it with appropriate caution, but I also think that this causes control decks to be less viable in the recent sets. Snowbally nature of cards, multiple effects cards have on ETB, all those factors cause removal to be less efficient in controlling the board state completely – it is very hard to survive the early onslaught of your opponent and survive till the late game in a good enough shape to take over the game. And because so many cards are value generating machines, playing a slower game is not necessarily the best plan for those cards leading to conflicts between individual card power and deck’s meta-strategy.
这是我的假设，所以请谨慎对待，但我也认为这导致控制思路在最近的系列中变得不那么可行。卡牌的雪球效应，以及它们在进场时产生的多重效果（ETB），都导致控制套牌中的去除变得不那么高效——在对手的早期猛攻中很难生存下来，并且在良好的状态下坚持到对局后期以扭转局势。由于许多卡牌都能产生价值，玩一个更后期的对局并不一定是这些卡牌的最佳策略，这会导致单卡强度与套牌的策略之间的冲突。
To put it simply – we moved from times where aggressive creatures were weak and missing one step meant aggressive deck was far behind to the times where aggressive creatures are very strong and this means that now control can’t cope if they have a stumble. But the most important take home message here is that speed is not linked to aggression of the format but to the value accumulation through playing modern design cards.
简而言之，我们已经不再处于快攻生物较弱、落后一步就会导致快攻套牌远远落后的时代。现在是进攻型生物非常强大，控制套牌一旦展开不顺就会难以应对。以及最重要的，速度与使用现代万智牌产生的价值积累有关，而不是与环境侵略性相关。
The game durations and their distributions are essential to figure out what does a given archetype want to achieve. In this table you can see the win rate of each LCI color pair by the turn on which the game ends.
在这张表中，你可以看到每个 LCI 色组在游戏结束时的回合数与胜率的关系。这些对局回合数及其分布对于了解色组的目标至关重要。
04 Fast Decks
快速套牌
As you can see there are some decks that do well only if they end the game very early. The most extreme example is Izzet (UR). In games ending on turns 4-6 it has a ~70% win rate, but the longer the game lasts, the lower the win rate. 63.8% if the game ends on turn 7 is still good, even on turn 8 it has a credible 58% win rate but on turns 11-16 it only wins 40-45% of its games, showing that Izzet decks don’t have a late game in this format. And this is a key information to you as a player.
正如你所见，有些套牌只有在对局非常早期结束时才能表现出色。其中最极端的例子就是伊捷（红蓝）套牌。对局在第 4-6 回合结束时，它的胜率约为 70%，对局持续时间越长，胜率就越低。如果对局在第 7 回合结束，胜率仍然达到 63.8%，甚至在第 8 回合也有可观的 58%，但在第 11-16 回合，它只能赢得 40-45% 的比赛，这表明伊捷套牌在 LCI 缺乏后期对局能力。作为玩家，这是对你非常重要的信息。
Given a choice of a very aggressive creature and a mid range one, you should lean aggressive while drafting Izzet decks. Because doing so taps into the decks natural power – being very fast and very good at being fast. Of course, once in a blue moon you will manage to draft some sort of Izzet control but vast majority of the time, you should try to maximise the power of the color pair to the full extent.
假如要在快攻生物和中速的生物中选择一个的时候，如果你是玩红蓝色组，你应该倾向于选择具有攻击性的生物。因为这样做可以充分发挥色组的天然优势——即非常快速且擅长快速发动攻击。当然，偶尔你可能会成功抓到一些红蓝控制套牌，但绝大多数情况下，你应该尽量最大化这一色组的力量。
Izzet being aggressive should be no surprise to those who drafted LCI, but WR might be. People try to draft WR as an aggressive leaning mid-range deck, but numbers show that this might be a bad strategy. WR decks that can close the game early are wildly powerful, with a ~68% win rate on turns 4-6, very similar to Izzet. And further, Boros decks have an almost identical win rate progression as the UR decks. This shows you should do all you can to slant your Boros deck as aggressively as you can.
在 LCI 环境，红蓝以快攻为主并不令人意外，但红白色组却有些出乎意料。玩家们尝试将红白打造成一个快攻偏中速的色组，但数据显示这可能是一个不太好的策略。能够在早期结束对局的红白套牌非常强大，在 4-6 回合时的胜率约为 68%，与红蓝非常相似。此外，红白的胜率变化也几乎与蓝红卡组完全相同。这表明你应该尽可能地使你的红白色组倾向于快攻。
But a careful observer will see one more thing. If the UR and WR win rates over time look almost identical – why does UR have a much better win rate than WR? They should be the same, no? Here we come to a key point – “wanting” to play fast games is not the same as “being able” to play fast games. It all breaks down on the capability of a color pair to finish games early. If a deck wind 70% of its games that end on turn 5, but only 2% of its games end that early, and 80% of its games end on turn 10 or later – this is not a good sign for its generic win rate.
但是，细心的观察者会发现更多。如果红蓝和红白的胜率随时间变化看起来几乎相同——为什么红蓝的胜率比红白要高得多呢？它们应该是一样的，不是吗？这里涉及到一个关键问题——“想要”玩的快并不等同于“能够”玩的快。这一切都取决于一种颜色组合早早结束游戏的能力。
如果一个套牌在 5 回合结束的对局中胜率是 70% ，但只有 2% 的对局能这么早结束，而 80% 的对局要打到 10 个回合或更久——这对于它的整体胜率来说并不是一个好消息。
It is one thing to understand what the plan is and the other to be able to implement it successfully. Izzet decks win better because they are capable to end games early more often. As simple as that. If you see at the distributions of the game lengths of both color pairs, you will see that on each of the early turns, Izzet decks end a couple percent more games than Boros and on longer games, the situation is reversed. And it is those small differences that eventually contribute to a much higher win rate of Izzet decks. Izzet decks managed to end 42.2% of their games by turn 8 and Boros managed only 37%.
了解计划和成功实施计划是两码事。红蓝套牌更容易获胜，因为它更擅长早期结束对局。就是这么简单。比较两种色组的对局回合数分布，你会发现在头几个回合结束的对局，红蓝比红白的胜率高几个百分点，而在较长回合的对局中，情况则相反。正是这些微小的差异最终导致了红蓝具有更高的胜率。红蓝套牌在第 8 回合前结束了 42.2% 的游戏，而红白套牌只有 37%。
TopDeck Tips：这里作者想说的应该是红蓝和红白都是在 8 回合之前是强势期，而红蓝对比红白也有能力将更多的对局在 8 回合之前结束。所以最后算完总账红蓝的胜率高于红白，哪怕过了 8 回合之后红白的表现更好。
Another deck that does very well in its fast version is RG Dinosaurs. I thought it is a classic mid range deck but the data suggests that RG does much better in its more aggressively slanted version. And this is something to keep in consideration. Through all the talk of Jeskai color pairs being the best thing to do in the format, it has escaped many ****** that RG has been the most winning deck for a couple of weeks running already.
RG is not as spectacular in the early games as WR and UR, but it has a slightly better staying power. And I could gamble a guess that the lower win rate in the early turns is caused by bad builds that are light on early drops and cheap interaction being overrun by other aggro decks. This makes me think that in order to win more with Dinos, you need to prioritise those. 2 drops, cheap removal are key so you don’t end up on the back foot – and keep in mind that your late game is most likely not as good as you think it is.
红绿恐龙是另一个在其快攻版本中表现非常出色的套牌。我原以为它是一个经典的中速套牌，但数据表明红绿在更具侵略性的版本中表现更好。这是很有意思的发现。在讨论洁斯凯（红白蓝）在这个环境中强势的时候，很多人忽略了红绿已经连续几周成为胜场最多的色组。
红绿在早期对局中并不像红白和红蓝那样引人注目，但它的持久力稍微更好一些。我敢打赌，短回合对局中胜率较低是由于构筑不佳，缺乏低费用牌和低费用互动，被其他快攻套牌压制所致。这让我认为，要想用恐龙胜率更高，就需要优先考虑这些方面。2费曲线和廉价去除是关键，这样你就不会陷入被动局面。另外请记住，你的后期对局可能并不像你想象的那么好。
05 Midrange and Control
中速和控制套牌
The other deck that does reasonably well in shorter games is WU (Azorius), but this archetype has a twist. Or even two. Firstly, it is not very good at finishing games early. Only 27% of games end before turn 8, comparing with 42% in UR. But that issue is mitigated by the fact, Azorius is actually good in the later game with win rates in 54-56% range on turns 10-14, when Izzet decks at the same time are below 45% win rate.
This staying power suggests to me that there are two types of Azorius decks – a small fraction of them will be good aggressive decks. And if you end up drafting one of those – your plan is very aggressive in the air. But those decks were most likely only available reliably in the early format. As it progressed, Azorius looks more like an aggro control – capable of wins but not through sheer aggression but by combining some early aggressive threats with control cards that let them last long and use their not-as-aggressive threats as the ones UR has, finish the job.
白蓝是另一套在短回合对局中表现不错的套牌，但它也有点门道。首先，它并不擅长早期结束对局。只有27%的游戏在第8回合前结束，而红蓝则有42%。但这个问题被事实所缓解，因为白蓝在后期对局中表现出色，在第 10-14 回合的胜率达到了 54-56%，而此时红蓝套牌的胜率却低于45%。
这种后期对局能力让我觉得在白蓝牌组中有两种类型——一小部分会是出色的快攻牌组。如果你最终抓到了这样的牌组，你的策略就是用飞机进行非常激进的快攻。但这样的套牌，可能只在环境早期比较可靠。随着环境的演化，白蓝看起来更像是一个进攻+控制型牌组——能够取得胜利，但并非通过直线的进攻，而是通过将一些早期的进攻威胁与控制牌相结合，让它们持续存在造成威胁来完成任务。
This can be seen in the comparison with UR:
这是白蓝和红蓝的对比：
The distribution of Azorius game length is much flatter. You can also see big differences in the number of games that end very early, and the significant % of games that end past turn 10 compared to Izzet. This means to me that drafting Azorius you should probably default to a slower build, as the very aggressive ones don’t seem to come together very often. This can make great use of the fact that white and WU double-sided uncommons, which are very control-slanted are more frequent than the regular uncommons in the set. Cards like Clay-Fired Kilns, Spring-Loaded Sawblades and Master's Guide-Mural are the reasons for that staying power and why despite having a high win rate in short games, Azorius looks like the best mid-range deck in the format.
白蓝的对局回合长度分布更加平滑。你可以看到白蓝相比红蓝，在短回合结束的对局少了很多，有相当比例的对局是在第 10 回合之后结束的。这对我来说意味着，在抓白蓝的牌时，会更倾向于选择一个更慢的构筑，因为非常具有侵略性的构筑似乎并不经常出现。
可以更好的利用白色和白蓝双面银牌，这些卡更偏向于控制，也比其他银牌更常见。土煅砖块、压簧锯轮和大师装配画，正是白蓝套牌的持久力所在。尽管在短回合内的胜率很高，但白蓝看起来更像是这个环境中最好的中速套牌。
There are 2 color pairs in the format that do much better in the late game Dimir (UB) and Golgari (BG). They both have low win rates in the early game – in the 40% range on turns 4-7. This means they need to focus their early turns on stemming the bleeding. Making sure they can survive till the late game when they excel. They both don’t have the tools to win early and you should not focus on that plan, rather than that, think how to make sure you manage to get the game past turn 10 in a good shape, which will put you as a clear favourite.
在这个系列中，底密尔（蓝黑）和葛加理（黑绿）在后期对局中表现更出色。它们在前期的胜率都很低，在 4-7 回合时只有 40% 左右。这意味着它们需要把前几回合的重点放在苟住上，确保能够存活到后期对局，那就是它们的天下了。它们都没有在早期获胜的工具，所以你不应该专注于这个计划，而是要考虑如何确保你能在良好的状态下将对局进行到第 10 回合以后，这将使你成为明显的优势方。
The distribution of the both decks’ game lengths is very similar.
两套牌的游戏长度分布非常相似。
They both manage to make the games longer – only 22% of the games end before turn 8 – this is exactly what they want to achieve. But it is still not enough. They would need to prolong the games by a whole turn to become tier one candidates. But keep in mind – we are looking at the data from average 17Lands.com user, who, while still above average a player, is not at the expert level. So my guess will be that a part of it is related to non-optimal deck building. The best of the best will be able to make the games last longer and – as a result – win more.
Of the two decks, BG has slightly better win rate stats. Why? The main difference is – it is slightly better early. Especially that small win rate difference in favor of BG on turns 8-10 is essential, since over 40% of the games for those decks end in that timeframe. What is worth noting, while still good into a very late game, BG drops off past turn 14, while UB stays pretty good even after that, making me think BG is a bit more midrange on the deck spectrum than UB, which is almost purely a control strategy.
他们都设法让对局回合数更长，只有 22% 的对局在第 8 回合前结束，这正是他们的目标。但这还不够。他们需要再延长一个回合的游戏时间才能成为顶级选手。但要记住，我们观察的数据来自平均水平的 17Lands 用户，虽然水平不错，但并非专家。因此，我猜部分原因可能与不够好的套牌构筑有关。最顶尖的玩家将能够让对局回合数更长，从而获得更多胜利。
在这个色组中，黑绿的胜率略高一些。为什么呢？主要是因为它在早期稍微强一些。特别是在 8-10 回合，黑绿小幅领先的胜率至关重要，因为这两套套牌中超过 40% 的对局在这个回合数结束。值得注意的是，虽然在非常后期的对局中仍然表现良好，但黑绿在第 14 回合后胜率会下降，而蓝黑即使在那之后仍然保持良好状态，这让我觉得黑绿更偏向中速，而蓝黑几乎纯粹是一种控制策略。
06 Conclusions
总结
I hope that the article prepares you to look at the format speed data and how to look at it in the context of game plans, but it also should be helpful if you are struggling with LCI. I know that thinking about those results I came to several useful conclusions of the formats and found several spots where my drafting style was misaligned with the color pair plans.
希望这篇文章不仅能帮你了解 LCI 系列的速度数据，还能在对局计划中应用。同时，如果你在面对 LCI 时感到困惑，它也会给予帮助。我意识到通过思考这些数据，我对于这个系列的牌有了一些有益的结论，以及我在组牌时的策略与数据呈现不一致的地方。
There is also a large scale observation from those data. Modern design is aimed at making formats fast. Games last shorter and controlling archetypes have problems in dealing with that. Threats are too good to be able to reliably extend the games long enough for controls to gain an edge, and I think in the long run this may lead to a certain tiredness with how different limited formats play out. As I am writing this piece, Khans of Tarkir is on Arena and even if it is not my favourite format and card power level is not to my liking, some design features from KTK would be interesting to apply from time to time. Once in a while we can have a set with 1- and 2-drops that are slightly weaker. Or fewer. Streets of New Capenna had a chance to be a format like that but it failed to deliver – mainly because the mono-color spells and powerful 2-drops were so much better than what 3-color cards had to offer, that it removed some incentives to explore the full solution space in the format.
根据这些数据，还可以进行大规模观察。现代万智牌设计旨在加快对局节奏。对局回合更短，而控制类套牌在应对这一挑战时遇到了问题。威胁实在太强大了，以至于无法可靠的延长对局回合，来让控制类玩家获得优势。
我觉得从长远来看，这可能会导致对不同的系列的限制赛玩法雷同感到有些厌倦。就在我写这篇文章的时候，KTK 正在 MTGA 上线。即使这不是我最喜欢的环境，而且卡牌强度也不符合我的口味，但偶尔感受一些 KTK 的设计特点还挺有趣的。偶尔我们可以推出一些稍弱的1费和2费卡牌，或者数量更少的这类卡牌。
新卡佩纳之本有可能成为这样的一个环境，但它未能达到预期——主要是因为单色法术和强大的2费卡牌要比三色卡提供的效果好得多，这使得探索该环境的动力减弱了。
Once in a while a set with a stacked 3-drop slot and bombs that start at 5 mana, with tools that enable longer games and give control decks a fighting chance, would be a breath of fresh air. And I hope that the brief revival of KTK and the generally positive response to it from the players will give a hint to the design team. And if not – make sure to point them to my merry rumblings. Maybe some data will be an eye-opener.
如果有一个系列从 3 费开始用堆叠，从 5 费开始扔炸弹，有办法能延长对局回合数，让控制套牌有机会支楞起来，那可能会是一股清新的空气。希望KTK 的短暂复苏和玩家们的积极的反馈能给设计团队一些启发。不行的话，就让他们听听我愉快抱怨吧，也许数据会让他们有所觉悟。
PS：Sierkovitz 这篇文章对应的原始播客有 90 分钟长，包含了更多内容，文本量极大。但 TopDeck 最近感冒了，法术力明显不足，如果你对原始播客仍然很有兴趣，想一探究竟，请在这篇文章下方点火、评论。我们会根据反馈的数量看看要不要早点做出字幕版昂~
再次特别声明：刚开始尝试做 Pro 内容翻译，难免错误和理解不到位，特别欢迎指正 & 讨论。
今天就写到这里，欢迎关注我们的公众号：TopDeck
如果觉得这篇内容有用，也非常欢迎转发给你的牌手朋友。
TopDeck 会持续关注 LCI 限制环境变化~祝各位 7 胜席卷~
