Battles in March of the Machine Limited: How to Best Use the New Card Type Using Data
Table of Contents
March of the Machine is looking like one of the all time great formats in Limited. Great mixture of powerful cards and efficient interaction, interesting play pattern and being able to play almost any color combination makes it a treat for limited players. It also gives the players feel good mechanics that make even some losses feel less painful. Companioning a Yorion, Sky Nomad, building around Omen Hawker, are side quests that will generate that all important dopamine even if the end result of the game will be a loss. Set designers tapped into the same mechanism that made Commander so successful: ability to enjoy the game independently of the final result, and flipping Battles is an important part of that strategy of increasing out enjoyment. Playing a Battle creates a small sub-game within the game and flipping it means you won the (yup, I went there) battle, without necessarily winning the war.
But Battles themselves are completely novel to the game. This means there are no heuristics linked to them, and even the most hardcore grinders had to discover how to play Battles from scratch. In this article I will tap in to the pool of 300,000 limited games aggregated by 17Lands.com and look through what makes the Battles successful, how many should you play and which colors benefit from the new mechanic the most.
Before I go into that, current 17Lands dataset contains 750k games. Yet I chose to pick only less than a half of them. the reason for it? Many people think that the more data, the better. But that is not always the case. Datasets can be curated to improve the quality of the data and it is worth the effort even if you loser out on the quantity. Since Battles are a new thing, it is reasonable to assume that in the early format they were played incorrectly. In general, the quality of early format data is lower as people play weak cards, some softer opponents populate the pods. On top of that, some color combinations are criminally open, skewing the results and making the data not really helpful in later stages of the format where those factors play a lesser role. That is why I selected only 300k most recent games in the data set covering roughly last 2 weeks. This way we see the trends that are more relevant for current metagame and based on a more informed drafting and deck building from the 17Lands.com users.
Who doesn’t like a long method-related introduction? But on to the main topic. A good place to start looking at the general trends of Battles, is through the lens of their individual win rates. There are 36 Battles in the set, and all of them are on Uncommon or higher rarity. This makes me expect a lot from them, as they will have to be picked relatively high, and I want the cards that I pick highly to have equally high win rates. But that is not really the case. Looking at the Game in Hand Win Rate (GiHWR), so what percent of games you win if you have a card in your opening hand or if you draw it at any later stage of the game, the battles look relatively modest.
The top 12 of them fall in the 63% to 54% range. 63% is of course great, Invasion of New Phyrexia is a clear bomb, making the top 10 of all the cards in the format in that metric. Side note – it is very flavoursome that the only Invasion done by the forces of multiverse is more successful than any of the 35 Phyrexian invasions. But aside from that mythic rare, only six other battles cross the 58% GiHWR, while looking at the whole set, 55 cards reach that threshold, including several commons.
Only one of the uncommon Battles, Invasion of Amonkhet, is in that high win rate group. The remainder of the Battles in top 12 fall in the 56-58% GiHWR range and still here only 2 are at the uncommon level, meaning that in the lower win rate 24 Battles, 17 are uncommon. 56% Win Rate may seem a lot, but we are looking at the data of 17Lands.com users, whose average win rate is 56% – not at a total Arena population, where of course the average would be 50%. This means that even battles in the top 3rd of the whole cohort are at a roughly average power level.
If you look at the worst performing battles, we are at a 46-52% GiHWR range. And that is quite poor. This is the power level of Furtive Analyst or Infected Defector, cards that are solid candidates for a last pick. And yet those battles are still picked up quite highly. Invasion of Belenon is picked higher than Artistic Refusal – one of the top uncommons of the set. Invasion of Lorwyn is picked at the same rate as the actual top uncommon in the set – Marshal of Zhalfir. It seems to me that the novelty of the mechanic is trumping reason in this case. Curiosity and willingness to experiment or maybe thinking about magic Christmasland scenarios – who knows which one – makes players pick the Battles way higher than their performance would suggest.
How to Use Battle Cards
Wow, that was pretty pessimistic. But does it mean I am low on Battles? Definitely not! So let me dedicate the second part of the article to figuring out how and when to use them to maximize the win rate while still having fun with the new mechanic. To start with, yes, I mentioned that people pick some battles too high, but actually if you look at the play data of the 17Lands.com users, they do play the better battles more frequently.
The top played Battle by a very large margin is also the best one, Invasion of Amonkhet. But all the other ones in the top echelons are also reasonable and among the best uncommon ones. Invasion of Eldraine, Invasion of Mercadia and Invasion of Zendikar are all decent cards. There are still some that have mixed results, like Lorwyn and Muraganda though. To check how well does the play rate correlate with the win rate I plotted those two against each other for each uncommon Battle.
You can clearly see that there is a strong link between how frequently the 17Lands.com users play a card and its GiHWR. The R^2 metric shows you how strong that link is – the closer to 1, the stronger and 0.65 is pretty strong in such messy data laden with lots of noise – this means 17Lands.com users are generally capable of identifying good battles and what it means for me – it does increase my trust in the conclusions from the next steps of the analysis, as indicates that any conclusions I drew from my dataset are going to be based on relatively high quality data. Still, there are some Battles that are underplayed and some that are overplayed – this means that they are either looking appealing but not delivering on that promise or look underwhelming but are actually hidden gems. Let’s start with the overplayed ones:
The biggest offender on that list is the Invasion of Kamigawa. A card with a terrible GiHWR but somehow with the great appeal to players. There are two main problems with it. One is the cost. Four mana for the tap effect is way too much. You want to pay 1-2 mana for it at sorcery speed and 4 is very overrated. The second problem is the back side. A 2/3 creature that draws a card on attack is worth probably 4 mana on its own, but flipping a battle is not easy so what you get is a 1-2 mana effect upfront, with maybe 4 mana gain if you dedicate damage to it – this is just too much under rate and the numbers show that.
Invasion of Lorwyn is not far behind. It does look appealing, being an unconditional removal that potentially gives you a large creature to finish the game with. But on closer inspection, you quickly learn that the removal has some conditions when opponent had Glissa, Herald of Predation on board and that most decks can block the massive creature without abilities till the cows come home, and that is if you are able to flip it, in which case you probably were better of going face. The problem with Lorwyn is it is worse than any removal in the format but has to be picked as one. And yes, in some decks you can include it if you are low on interaction but you should definitely aim for better removal instead of taking Invasion of Lorwyn over it.
Invasion of Ulgrotha is better, but still gets played way too much. The front side is below rate, but the back is a powerful evasive creature with some synergies for a sacrifice deck. Still the same applies – other removal is just stronger, so don’t plan on Ulgrotha to be your plan A in terms of interaction, play it when you are short on it or when you do have some Battle interaction, for example if you do have BR deck with Invasion of Regatha.
Invasion of Zendikar is a bit of a weird one on that list. It is a strong card, with PT top 8 credentials in constructed. But it still gets overplayed in draft. The potential problem is that it has a narrow home, a home where you want to use all functions of it: the ramp, the color fixing and the ease of flipping it. Otherwise it is very strong early but becomes a pretty much a dead draw later in the game. If your deck has top end power, and plays a multicolor strategy while having multiple Wary Thespians, Invasion of Zendikar is exactly what you are looking for. Otherwise – you might want to get better cards for your deck.
But some invasions are actually slightly underplayed – they are played less than their win rate suggests.
Invasion of Dominaria is not a card for any deck, but if you are playing a slower and more controlling version of White, maybe a UW filled with interaction and Sunfall as your bomb or some sort of WB attrition deck – Dominaria is something you should consider. The ability to cushion your life total, dig deeper towards your bombs and make sure you don’t miss those key land drops. And when the dust settles, you can try to flip it and get a 5 mana creature on top of your 2 mana worth life and a card. I want to stress strongly that this is not a card for any deck but when you find a shell it can be surprisingly good, as indicated by a decent GiHWR.
Invasion of Kaladesh is another one where you need ot work for it a bit, but the reward is good. 2 mana cost is not much, and the little thopter can slowly chip away towards the vehicle. But even with the thopter, the reward would be a 2/4 flyer that still needs crewing. That is why for Invasion of Kaladesh to be really good you may want some support, and luckily there are some cards that can provide it while still being on plan for a tempo aggro UR deck.
Marauding Dreadship creates 2 artifacts and can surprisingly flip the battle if opponent is not careful. Beamtown Beatstick with the thopter can speed up the flipping of the battle while being an artifact itself and generating additional artifact treasures. Tapping to those micro-synergies can move Invasion of Kaladesh from OK to great. Just keep in mind not to flip two copies, or you are in for a nasty surprise.
Invasion of Azgol is a very good card in a very underrated archetype. It is not an accident Team Handshake was sort of forcing RB in the recent pro tour (and with great success). Rakdos signpost uncommons are all great and synergy driven cards that go relatively late as no other deck wants them so keep a lookout on those and if you see them – speculate on them. Invasion of Azgol has a lot of its value in the front side, so you will rarely be sad about playing it. The sacrifice effects are often mediocre in limited, but this coming on top of a Battle makes a large difference.
Firstly, if you flip it, you get a 5 mana of value for your 2 mana investment, secondly, you sometimes can force opponent to make awkward sacrifices, if they have a low value blocker left over and better tapped creatures they may have to sacrifice better ones in order to prevent you from flipping the battle. There is also a hidden mode on the card – you can target ant player so on rare occasions Invasion of Azgol can become a sacrifice outlet for you, especially useful if you have Furnace Reins to steal opponent’s creature but on sacrifice outlet.
Invasion of Pyrulea is another undervalued invasion. The front side will net you a card a lot of time, with 17 lands in your deck and most likely 3-4 other double-faced cards in a UG deck. On top of that, you will also arrange top 3 cards of your library, which is basically as good as drawing a card in the late game, so the card doesn’t lose its utility in the late game. And the body you get if you manage to flip it is easily worth 4 mana, so you get your investment back and some. And the best part – you don’t have to play it on curve. Why not play it on t5 when you are missing a land drop and follow it up with a 3 mv spell, while fixing your future draws?
Number of Battles in Decks
But focusing on individual battles is not all that can be done using data. I was always more interested on larger scale trends and using those trends to go more in detail. So let’s look at how many battles were people playing in a deck?
It did come to me as a bit of a surprise, but in almost half of the games I looked at, there were 0-1 Battles in the deck. Given that there are 24 Battles per draft pod, it means they see surprisingly little play. In roughly 25% of games decks have 2 Battles, 15% 3 battles and only 10% of the games are with the decks that have 4 or more battles. My congratulations to the person who played as much as 11 battles in their deck, and actually with a decent 5-3 result with a Jund deck. But the general trend is to play few, if any Battles at all. Is it reasonable? I also looked at the link between the number of Battles in the deck and the win rates.
And here we see that with addition of Battles your win rate is actually suffering. At each increase of Battles in your deck (except the last, but that is based on a relatively small sample so will be subject to variance, so lower trust in the result) causes a decrease in the win rate. So. That’s it. Battles are rubbish and you should never play them. The end.
Or is it? This is where subtlety in data analysis comes in handy. It is easy to stop here and draw a conclusion – Battles are bad, you should not play them. Hey, I am pretty sure it would make for a catchy and clickbaity headline. But would it be right? Not for me at least. We are talking about a new mechanic still in development. And aside from data, I do have play experience with Battles and see many merits of including them in my deck with success. A result like this is a good starting point to a broader analysis, and the first question that was rattling around my brain was: what is the link between Battles and game duration?
The reason this interested me is the nature of Battles. In order to flip them you have to not attack opponent’s face and direct damage towards the Battle instead. And this will mean you may – depending on the impact of the flip – take longer time to win the game. Battles also often do not impact the board, or do so under the normal rate, which may mean that decks containing too many Battles can get easily overrun in some games. So I decided to look at how long do the games last depending on how many Battles are in the deck. And because I anticipated different outcomes, I wanted to see this impact on games 17Lands.com user won and lost.
Turns out one of my hypotheses was completely correct and the other at least partially so. The more Battles in a deck, the faster your losses are. Between having no battles and having 7 of them in your deck, you are going to lose your lost games over half a turn faster. And since, as I showed earlier, you lose much more as you include more Battles in your deck, this is a pretty impactful effect. Adding more Battles to your deck makes it also slower for you to win the games you win, but that effect stabilises and between 3 and 7 battles the differences are not that large.
Another set of bad news for Battles? Not necessarily. This is actually a key piece of information you can use to play Battles better. Remember, we are looking at the first time Battles are in the game so lots of that data, even though I curated out early format, are going to be from badly built decks. What this graph suggests me is: you probably should avoid Battles if you plan to beatdown – you can’t slow down your clock. There are possibly some Battles that are tempo positive and because of that will play OK in aggro, but you should be careful there. The other part is that Battles can be actively good in decks that don’t mind – or even want to play a longer game, but if you want to go this way you have to make sure you have means to survive any early aggression despite playing Battles that potentially don’t contribute to the board in a meaningful way.
But does it work? To test it I though – why not look at the win rate data of different color pairs. The problem with such general graphs as the ones above is that they miss out on detail. If 8 of the 10 color pairs suffer from putting too many Battles in the deck, and 2 benefit, a generic graph encompassing all data will miss on that and lead you to easy to reach, but ultimately wrong conclusions. Driven by that I looked at how does the win rate change in each color pair depending on the number of Battles in deck.
It is best to look at this graph top to bottom for each column and see how the numbers change. Color pairs where values decrease don’t do well with Battles, color pairs where the numbers stay roughly the same are OK with Battles and where it increases, they actively want more Battles. One thing that will strike you is how bad are the Battles in white decks. In WU, WR and WG addition of Battles hits the win rate by a lot. WU has almost 60% WR with no battles and it drops right down to 52% with 4. WG with no Battles is actually one of the better color combinations whatsoever, but the whole color pair’s win rate gets dragged down by decks containing Battles. WB does see some decrease in WR with addition of Battles, but the problem with this color combination is, it doesn’t win much even if you don’t include them.
But other color combinations don’t do bad. UB has a solid WR across the board, but in this case a question remains how much of this effect is linked to multiple copies of Invasion of Amonkhet. But also UG and BG show stable levels independently of playing zero or 4 Battles in the deck. UR also shows somewhat stable WR across different numbers of Battles in deck. Finally a good result for Battles as a limited mechanic. And there is better news to come – BR and RG show even some increase with more Battles in the deck. Yes, there can be Battle-independent explanations of BR (you will get more Battles when the color is more open), but that is still an encouraging result. And in case of RG the results are somewhat wonky, but it is clear that it is better to have Battles there than not to have them, minding that this is one of the less played archetypes, thus sample sizes are not gigantic.
Looking at this data, you can conclude that the drop in Win Rate seen as you put more Battles in the deck is mainly driven by white decks, but you wouldn’t be able to conclude it from the generic graph. This is a major problem by drawing sweeping conclusions from diverse datasets. Even looking at the level of detail I used here – you will miss out on the corner cases. I am sure there are white decks where Battles are good, but signal from those will be lost in the noise from the more popular white aggressive decks that don’t want Battles, or want maximum one of them. This should be a warning from putting too much weight in such sweepingly generalising analyses, which have easy and catchy slogans but miss out on subtlety of Magic.
Game Length and Battles
That is about all I can say about Battles themselves, but there was one aspect I wanted to look at before the end and that is the speed of different color pairs and how does that link to the data on Battles. The idea is that color pairs that don’t want Battles – want to win games early, and the ones that want to play longer games – don’t mind Battles.
And it turned out to be true. Here you have a graph? table? grable? with Win Rates for each color pair depending on when did the game finish. And you can see clearly that four color pairs: WU, WG, WR and RG win a lot of games if they finish early, but struggle if the game goes longer. Particularly in case of Azorius, that is a spectacular trend, going from 66-68% WR in games, which end on turns 5-7 to 45% WR in games that end on turns 12-15. Dimir is a polar opposite. Not a spectacular deck early with low capacity to win in the first 7 turns, but once it passes turn 9 – win rates skyrocket to 60%+. UG and BG both look much more mid-range – winning less early and also not being able to keep up their win rates very late in the game. And BR and WR are not having great numbers in any zone, with BR showing some slight potential in the early game.
And the results of that analysis feed in to the Battle data. All the White decks not only don’t want Battles but also need to kill opponents early or risk running out of steam – quite literally. I had several games in MOM where I dropped the opponent to 1 life and then hit a brick wall, watching helplessly as they swing the game state and recover for the win, and this data confirms I was not the only one. There is a strong link between wanting to win early and avoiding Battles and while deckbuilding this should be the question you should be asking yourselves: do I want to go to mid game or finish the games before that. It you want to finish before that – don’t exceed one Battle in your decks. If you are interested in the late game – pick them and play them but don’t use the low impact ones.
If you want to hear a more convoluted version of this article that for some reason includes a cat and a dog fighting, a cat begging relentlessly for food and some more data on the topic of Battles and not only – watch this YouTube version of my seminars!
Enjoy our content? Wish to support our work? Join our Premium community, get access to exclusive content, remove all advertisements, and more!
- No ads: Browse the entire website ad-free, both display and video.
- Exclusive Content: Instant access to all exclusive articles only for Premium members, at your fingertips.
- Support: All your contributions get directly reinvested into the website to increase your viewing experience!
- Discord: Join our Discord server, claim your Premium role and gain access to exclusive channels where you can learn in real time!
- Special offer: For a limited time, use coupon code L95WR9JOWV to get 50% off the Annual plan!
Great article Sierkovitz, as always, much appreciated. I especially liked the ‘grable’ representing the WR/turn no. correlation per colour pair. Technically it seems intuitive however seeing it in such a clear form is very helpful for me. Please consider publishing it regularly from now on for every format 🙂 keep up the amazing work.