17Lands: Where I Agree, Disagree, and What Data You Should Be Taking From It
Hopefully everyone is doing well! Today I’m finally writing about my perspective on the 17Lands database because it’s gotten to the point where it seems like it’s the main source of draft data that Arena Limited players are currently using. It’s a very detailed and in-depth approach to identifying the best cards and strategies of a specific Limited format and another big appeal of it is the large sample size that’s used to collect the data for it. Players who have 17Lands setup can simply draft and have all their information easily added to the database.
However, a huge issue that I come across over and over when I see people discussing 17Lands data is just how bad players are at interpreting the data that is present. I see players often stating how one archetype is better than another because it has a higher win rate according to 17Lands, and that’s an incredibly detrimental mindset to have when it comes to participating in Limited discourse. In my opinion, draft is probably the most complex and dynamic forms of Magic, and it’s impossible to answer all the questions about it by simply going over card/archetype win rates.
17Lands is still a tool to use that could help improve your Limited game, so in this article I want to highlight what data I agree with, which I don’t, and what you should and shouldn’t be using 17Lands data for. There’s an absolute mountain of data on the site, so hopefully I’m identifying it correctly because otherwise that would be really embarrassing! Let’s dive in, starting with the color ratings!
Format Color Ratings
I don’t think this data matters particularly much because you’re not going to often draft mono-colored decks in this set. I’ve done so many drafts and have yet to have it happen for me. Plus if you end up multicolored in a Draft and end up with enough playables, then you probably did a great job reading the table and were the only one in that color. If that’s the case, you’re going to have a solid deck regardless of whatever color it is.
Mono-colored decks are so great in Limited because they get to ignore the biggest issue with Limited decks; their mana base. Having perfect mana every game just increases your win rate by an insane amount. So don’t look to this data to decide what color is the best in draft. In my eyes, this information serves little to no purpose beyond reinforcing being the only drafter of a color at a table is obviously good.
Format Color Combination Ratings
Now this is the juicy data that I was looking forward to breaking down! During every Limited format in recent memory there has been constant debate over what color combination is the best at a given time. In Midnight Hunt Limited, you’re almost always two colors, so I wanted to focus on the win rates for those archetypes. What makes Midnight Hunt a great format is that every two color combination is viable, even the worst one’s like Boros. 17Lands data also seems to support this assumption since the discrepancy between the highest and lowest win rates archetypes is 7.5%, which is a decent, but not insurmountable gap. Here’s how the color combinations are ranked according to 17lands data:
However this data needs to be properly interpreted, so you can’t just assume that this is an accurate ranking of the ten color pairs. Certain archetypes are harder to play, draft, and sometimes they’re being over-drafted or under-drafted depending on the current Limited metagame. Sometimes players prioritize the wrong cards for a certain color combination or maybe they aren’t executing the proper game plan for that color combination. There is also a large discrepancy in games played for each color combination, with Dimir having almost 100,000 games played, Azorius only having 46,000, and Gruul having a surprisingly low 18,000 games in the database. Most importantly, what games is this data being taken from? New and inexperienced players inputting their data does more harm than good when it comes to reaching solid conclusions. In my opinion, there are just too many variables skewing the validity of these rankings and I want to break down these specific issues while I’m talking about some of the color pair rankings.
Let’s start with Izzet, because I’ve actually found Izzet to be one of the best color combinations when it’s drafted correctl,y but it can also be one of the worst when you’re prioritizing the wrong cards. Izzet needs to be fast and based around Festival Crasher, Thermo-Alchemist, Delver of Secrets, and Seize the Storm. If you find yourself relying on Ardent Elementalist, Arcane Infusion, Unblinking Observer, and Mysterious Tome, you’re going to get quickly out-competed on the board by the colors with better creatures. What’s deceiving about this data is that Izzet Aggro and garbage Izzet midrange are grouped together, which means that the power of Izzet Aggro is overshadowed by the mediocrity of Izzet Midrange.
Azorius being labeled as the best archetype over Dimir and Simic is also incredibly confusing to me. Azorius has great creatures because of Disturb, but their removal is so much worse than Dimir and is ultimately what makes the color combination so weak to creatures with powerful abilities like Morbid Opportunist and Skaab Wrangler. A reason why Azorius might be winning so much is players undervaluing the color combination. I’ve noticed during drafts that Mourning Patrol, Beloved Beggar, Lunarch Veteran, and Shipwreck Sifters are consistently going much later than they should be. Every creature with Disturb is good, and it still seems like a lot of players aren’t properly evaluating just how powerful it is to draft an Azorius deck with multiple Disturb creatures and Shipwreck Sifters. I like Azorius, but I feel like it’s win rate is slightly inflated and that there’s a very low chance that Azorius is a more powerful color combination than Dimir and Simic.
It feels criminal for Golgari to have a higher win rate than Gruul! I’m assuming that a ton of players draft Gruul poorly, and that’s why the win rate here is so low. This isn’t like Ravnica Allegiance where you just take cheap creatures and hope to get there. Gruul revolves heavily around having cheap Werewolves and creatures that start the Day/Night cycle early; you’re not interested in bad two drops like Pestilent Wolf and Timberland Guide. I feel like this data is deterring players from drafting Gruul, which is a shame because the color combination is very good because of how good it’s creatures are and how well it makes use of Lunar Frenzy and Duel for Dominance. Again, this data also groups together Gruul Werewolves and Gruul Spells, which are entirely different decks that almost certainly have way different win rates.
Overall, I wouldn’t put too much credence to the color combination data listed here. It’s correct in identifying Dimir, Simic, Orzhov, and Azorius as the conventionally best archetypes because they have the overall highest card quality, but it doesn’t take into account factors like player mistakes, improper deck building, multiple archetypes in one color combination, demand for the archetypes, under and over evaluation of certain cards, etc. The only purpose I could see for this data is it supporting the idea that every color combination is viable and getting a very basic hierarchy, even if it isn’t entirely accurate.
Individual Card Ratings
This was the section that I was most excited to dive into. Having a database that accurately evaluated cards based on their performance at every stage of the game is an incredible tool to use when considering what cards to take during a Draft. It was slightly confusing and a lot to dive into at first, but I’m honestly not the best when it comes to data science. I wanted to break down the relevance of the following in-game data metrics that are being collected on: games in hand win rate, opening hand win rate, improvement when drawn, and games played win rate. I’m not really interested in how late or early cards are being taken because those trends are drastically fluctuating at all times. Measuring the in game metrics is what I believe could end up being the best way to determine just how powerful/weak some of these cards are. This is is what the table looks like on the site:
Let’s start with Lunarch Veteran. Some players love the card, while others hate it. Personally, I’ve found the card to be mediocre in most decks and excellent in only a select few, namely Azorius Disturb with a bunch of Shipwreck Sifters. However, the game drawn win rate of Lunarch Veteran is 59.3% with an opening hand win rate of 63.1%, which is absolutely preposterous to me since it’s higher than the opening hand win rate of Organ Hoarder, which is only 62.2%. The fact that the opening hand win rate of Lunarch Veteran is even close to cards like Adeline, Resplendent Cathar (which has a 66.4% opening hand win rate) makes me incredibly skeptical of players being able to draw the right conclusions from this data. Obviously it’s good to have your one drop in your opening hand, but it’s not anything special or even particularly close to Adeline, so let’s all cool it on Lunarch Veteran!
Let’s go to another couple cards that are making me weary of the data being presented.
Crossroads Candleguide has a similar issue, except it’s a way better card than it’s currently labeled as. I see a lot of players playing it in their streamlined two color decks, which explains why the opening hand win rate of the card is only 47.7 and why the games not drawn win rate is higher than the games drawn win rate. However, Candleguide is actually surprisingly decent in a three or four color deck that isn’t glutted with four and five drops. The same is the case with Mystic Skull because it’s a bad card that does actually have a home where it can shine. Yet again, data being acquired from newer players is incorrectly skewing the value of the card.
Is 17Lands Useful Then?
Absolutely! For example, the data is pretty on point when it comes to a card like Heirloom Mirror. This card is such a trap because the flipside is so powerful and enticing, even though in actuality it’s near impossible to reach. The card is unplayable, and the game not drawn win rate being higher than the game drawn win rate supports that. That means that a new player could avoid pitfall cards like this by looking at the data, which is a great resource to have if you’re newer to Limited and are trying to improve at it.
Overall, I don’t think that 17Lands data is for the very invested/experienced Limited players, but that’s okay! Reviewing the data will help paint a picture of the format in a less experienced player’s head while helping them to not grossly misevaluate a card that they’re unsure about. A decent amount of the data also supports a lot of the same format conclusions that myself and other Limited devotees have come to, so it’s not like I have issues with a lot of the data being presented. In short, it’s going to help you identify the best and worst cards in the format more often than not, even if situational cards like Lunarch Veteran slip through the cracks.
If I’m being honest though, if you’re really unsure about a format, I’d look to the many masters of Limited to get your questions answered. Ben Stark, Sam Black, Mike Sigrist, TheHamTV (Kyle Rose), Martin Juza, LSV, Raphael Levy, PVDDR, Marcio Carvalho, Jean-Emmanuel Depraz, Shuhei Nakamura, Reid Duke, William Jensen, Chrisitan Hauck, Jacob Wilson, Allen Wu, and there’s definitely way more but these are some of masters that came off the top of my head. I trust their opinions and know that the opinions they’re expressing aren’t being skewed by the inputs of less experienced players.
It kinds of reminds me of ancient Greece and why it placed such importance on public education. In order for Greek Democracy to work, all the citizens who were participating in it had to be highly educated. That’s kind of how I feel about this data. I only want input from those who are experienced in the format and general Limited theory because if everyone can give their input then you’re going to have a lot of incorrect data being inputted. That causes the results and data that are being acquired become muddied and significantly less reliable since data input from Bronze players and Mythic players hold the same weight.
Feel free to disagree with my assessment of 17Lands though, and talk to me about it in the comments! I’m always open to being wrong and there’s definitely a chance that I’m missing something here because this data was a little difficult to break down.
Thanks for reading!