I picked up Lands around January of 2020, so about a year and a half ago, mostly because Oko had made Miracles a deeply boring deck to play. No longer could I even pretend to lock people out with Counterbalance, and I had to face the fact that blue mirrors just weren’t that fun for me. After Oko, I wandered the blue soup world, trying things like Stryfo Pile and my own weird Bant Knight of the Reliquary decks, but I’m a prison player at heart and none of the decks I played scratched that itch.

So one night, half-drunk and mildly depressed (not because of Counterbalance, just early 2020 was a bit of a rough patch for me), I found a Tabernacle on Ebay and made a reckless bid that would change my life.

After I woke up that morning a little hungover and a lot poorer, I had to pick up the rest of cards and become a Lands player. Not long after, I top 8’d a decent-sized local Legacy event with BUG Lands (Oko was too dang good), and I was hooked. I got MTGO and started playing way too much and now here I am, about 20 months later, with just over 1000 matches of Lands under my belt.

I’ve been tracking my matches since just after that tournament and 1000 is a pretty number, so I thought it would be a good to take this chance to step back and analyze the data. While obviously this is all my own matches, I’ll try to make it as relevant as I can to a wider audience.

Overall Winrate

Over the 1000 matches, I had a winrate of 59.8%, which is a number I’m pretty happy with. Since winrate is probably the most important metric for a deck, lets take a look at how it breaks down over various categories.

Winrate by Event Type

Here ‘weekly’ denotes the usual FNM events, ‘tournament’ could be a local event with higher stakes, but is more commonly a weekend Challenge, ‘practice’ is the practice room on MTGO or practice matches with a testing partner, and ‘league’ means, well, leagues on MTGO. Leagues accounted for 690 of the 1000 matches, so they were by far the biggest group.

From the graph we can see that weeklies are in general softer than MTGO events, at least for Lands. This makes some sense since at weeklies there are often less experienced players, and this makes winning a bit easier, especially when you play a deck that is relatively rare and hard to play against. So if you want to grind store credit at your LGS, Lands ain’t a bad choice.

Probably the two most important numbers are the league and tournament winrates. Overall I managed a 56.5% winrate in Leagues and a 58.8% winrate in tournaments. It may seem surprising that tournaments went better than random league play. If you consider how much more combo there is in leagues, and how much more fair blue there is in tournaments, the numbers make a bit more sense.

Winrate Over Time

There are a couple ways to think about winrate as a function of time. First, let’s look at how my winrate was over the 1000 matches. Here I’ll take the average of every 100 matches to see if we improved over time.

Looks like we never got below 50%, and the winrates have been improving more or less steadily. Of course, it’s hard to tell if this reflects personal growth with the deck or shifts in the meta. So it may be more instructive to look at the winrates in different metagames over the last 1000 matches.

Here we see that in the early Oko era when I started playing Lands, the deck was doing just about fine. Then the companions were printed. The Companion Era, however, was obviously just a broken period for Legacy, so it should surprise no one that we didn’t do too well during that time.

What is a bit more interesting is that after the companion ban, Lands started doing a lot better than it had in the pre-companion meta. This could be in part due to my own improving skill with the deck, but it’s also worth noting that it’s during this period that Valakut Exploration is printed. That card gives our deck a powerful new engine that let it fight back against the Astrolabe-powered control decks of the Oko era.

After Oko, Arcanist, and Astrolabe are banned, Lands reaches even greater heights. The meta at that time had a ton of Delver – this was when some of us even started playing Shifting Ceratops and crushing Delver with it.

Of course, then MH2 came along and totally shook up the meta. Still, with just over 200 matches played during the MH2 era and a winrate just under 66%, I think we can confidently say that this meta is pretty good for Lands.

Play vs Draw Breakdown

I won’t waste your space with a big graph since there are only two options for this category. But here are the facts. Overall, my winrate was 59.8%. On the draw, my winrate was 61.9% over 514 matches, and on the play my winrate was 57.6% over 486 matches.

Yes, that does mean that my winrate OTD was more than 4% higher than my winrate OTP. This is kind of surprising. With 1000 matches, one can’t easily just shrug this off entirely as a function of low sample size, though that certainly could be part of it.

Given how Lands really wants to get ahead on mana and use that advantage to take over the game, it may seem crazy that we’d win more on the draw. But it’s also true that Lands has a lot of ways to take back the mana advantage opponents gain by being on the play. A Mox Diamond or an Exploration can easily put us virtually on the play. Given these catchup mechanisms, maybe the extra card is better than being on the play? Something to think about anyway.

Winrate by Lands Archetype

Now lets take a look to see what the top-performing versions of Lands were.

This graph is ordered by descending popularity, meaning I’ve played with Jund Lands more than any other style of Lands (321 matches). That’s because Jund was the most common version of Lands throughout the Oko era, where having access to Abrupt Decay was very important as an answer to opposing Okos.

The blue line across the graph indicates my overall average winrate. Bars above that line are decks that over-performed, while those below it are less excellent builds.

A few observations. First, RG Saga Lands is the best-performing archetype, followed by RUG and BUG. RG Lands is still above-average, while Jund is below average. Jund’s lower winrate is likely a result of the era in which it was popular, and the fact that RG non-Saga Lands has a lower winrate is probably tied to that as well. I’ve played almost only Saga builds since the release of MH2, aside from a couple leagues and maybe a mediocre Challenge result. So the non-Saga build will have its numbers dampened by play during the Oko era, while Saga Lands will be bumped up since it’s been played exclusively post-MH2. That said, in the 22 matches I’ve played with no-Saga RG lands since MH2 came out, the build only got a 41% winrate; hardly a ringing endorsement (though hardly a large sample size either).

Another surprising thing is how well BG Lands did. Given the power of Valakut Exploration (and the fact that a lot of the BG builds were pretty experimental) it’s surprising to see that a build without VE can do as well as it has. Of course, that could be due to small sample size (only 33 matches).

The high winrates of BUG and RUG are also interesting, and each of those has over 100 matches in the dataset. Perhaps these versions may be worth further exploration, but it’s also possible that their time has come and gone and we’re just looking at old successes not suited to the current meta. Still, it’s clear that these variants can do well and Lands doesn’t have to be strictly RG.

Winrates by Matchup

What are the matchups like? Lets take a look.

Here I added the red line at 50% so one can see at a glance which matchups I tend to win more than I lose. As before, I’ve ordered these by descending popularity, from 232 matches in the ‘Brew/Other’ category, to just 4 in the relatively new ‘Jeskai Tempo’ category.

A few notes on the archeytpes. ‘Knight’ refers to Maverick and 4c Loam – there is a separate category for GW Depths specifically. ‘White Creatures’ is DnT and adjacent decks. ‘Big Mana’ is mostly Cloudpost decks. ‘Graveyard’ covers a wide variety of strategies, from Hogaak to Dredge to Reanimator.

Let’s discuss the bad matchups first. The thing I want to point out here is that Storm and Show & Tell are not nearly as bad for us as they’re usually made out to be. Conventional wisdom is that Lands just auto-loses to combo, especially these decks. While they’re certainly not good matchups, neither is much below 40%; hardly a total disaster. In fact, it’s actually combo-control decks like Food Chain and Aluren that are harder to beat as Lands (though I’ve only got 12 matches against those, so perhaps the data is misleading).

Looking at the other matchups, one thing to note is that the good matchups far outnumber the bad. The top 5 most common decks are all good matchups with over 60% winrates. Despite the joke that everyone claims a good Delver matchup while they lose to the deck, I think we can confidently say that Lands actually does have a good Delver matchup with a 63% winrate over 135 matches.

As for surprises, it’s definitely surprising to see such a positive winrate against Doomsday – 58.3% over 24 matches. I’m sure part of that is playing against people who have just picked up the deck, but it’s encouraging nonetheless. A less happy surprise is how badly I’m doing against traditional BG Depths. Lands is usually considered to be favored there, but I’m not doing too well – I’ll have to figure out that matchup if that deck ever comes back in a big way.

Wrapping Up

Overall, I’m happy to see that I’ve been improving over the course of these 1000 matches. Even if a decent amount of that ‘improvement’ can be put down to shifts in the meta, I think it’s alright to take a little of the credit. With some practice and a lot of help from the community, I’ve gone from a total newbie to someone who has a decent clue and can achieve a respectable winrate over a reasonably large sample size. That doesn’t mean it’s time to rest on my laurels, but its encouraging to know that I’m doing alright.

Thanks for reading! If you have any questions going deeper into the data or are interested in how I crunched the numbers, feel free to reach out via the links on the Loam below. If you’re interested in the code and raw data itself, they can be found on Github. Enjoy 🙂