MTG Arena Zone Premium
MTG Arena Zone Premium
Zar Ojanen, Scion of Efrava Art by Justine Cruz

Data-Based Heuristics for MTG Arena Drafts

What is the point of using draft data if we don't get actionable advice from them? Sierkovitz looks at the drafting data from Dominaria United (and some from other sets) and propose new drafting heuristics related to drafting mana fixing and multicolored spells.

In recent few years, limited data exploded in Magic universe, mainly thanks to 17Lands and their excellent tracker that shares big datasets with users and allows anyone interested to dig into the win rates, ALSAs and mulligan rates for each format. The depth of the data has made formats easier to understand. We learned new insights on sneakily powerful cards that may have been underestimated if there wasn’t a large body of data to back them up. Data has levelled the playing field, giving people that draft only occasionally a tool to catch up on format wisdom against grinders who churn hundreds of drafts each format. But data insights are used in most cases to look at individual formats. Which cards make them tick, which color combinations are particularly successful or open – these are the questions that people dealing with draft data (me being guilty as charged) try to answer.

However, years of data can be used to produce more universal insights. Ha, I even dare to say they should. That is why today I will propose two heuristics that result from data analyses across multiple formats and confirmed by the draft data from Dominaria United. These heuristics show the universal psychology of drafters and knowing how people at your pod think can aid you in planning your draft and not falling into some traps, which can hinder your win rate.

Procrastinating Fixing

Heuristic number one is very human. Or maybe very Magic-playery? We tend to prioritise flashy spells over boring, but ultimately all important mana fixing. I noticed it first in Kaldheim when I was exploring Average Last Seen As (ALSA) metric from 17Lands. ALSA tells us how late is the card picked in a draft by Arena users. For example card with ALSA of 1.0 will be only seen as pick 1, meaning it will be never passed by people at your pod – cards with ALSA around 1 are the format bombs. Good commons will have ALSA of around 4-6, meaning that on average you will see them around pick 5, but sometimes they will be seen earlier or later.

I looked at ALSA because the generic metric aggregates ALSA from all packs, so if you see an ALSA measure of a card it is its ALSA from pack 1, pack 2 and pack 3. My thought was that surely most cards have very similar ALSA across packs so the average is telling, but there might be some categories of cards that are picked differentially in pack 1 compared to pack 3. And I was not disappointed. When I looked at cards that were picked much earlier in pack 3 than in pack 1 I noticed the top hits were almost exclusively snow dual lands. Almost as if people ignored land fixing in the first pack and later desperately tried to get the fixing they need later in the draft.

I observed the same trend over and over again in other sets: Strixhaven and Streets of New Capenna. And always the same pattern emerged. Land fixing was picked much earlier in later packs. It was a no-brainer to test the same in DMU, a set defined by the presence of tapped dual lands. And I was not disappointed. The same trend is apparent also here. Dual lands go much later in pack 1 of DMU draft than in packs 2 and 3. We can tell by comparing how often do lands wheel in each respective pack when you open them as a potential pick 1.

Fig. 1: Frequency of wheeling dual taplands when opened in the pack by pack number. Source: 17Lands.com

In pack 1 each dual land wheels approximately 40-50% of the time if 17Lands.com user opens it. But it is not the same in later packs. In pack 2 they wheel only 33-40% of the time and in pack 3 when the need for fixing can become dire, only 23-35% of the time.

With several sets of evidence, we can conclude that this fixing procrastination is a universal trend. And this allows the first heuristic: land fixing is more available early in the draft. But a heuristic on its own is worth little without actionable advice. What can you do in a draft with that information? Well first thing would be to prioritise fixing early. I applied that strategy with some success in SNC, when in the first pack I was focusing on the fixing, picking only powerful cards over it. Cards that are a reason to go into a color combination, not a mere reward. And that strategy was successful. With ample fixing I was capable to play fully 3 colored decks and because of that I had no problem getting there on playables later in the draft when the rest of the pod were fighting over the fixing lands to make their mana bases adequate.

Late Opportunities

But that strategy can be only feasible if you will get a solid benefit of having good fixing. Fixing your mana to play mediocre spells is not a recipe for success. Good mana bases make casting spells easy but if you don’t have anything important to cast, you can fix the mana all you won’t but wins won’t come. Land fixing is usually most prevalent in sets that by design want to have decks with more colors. Two color sets usually get their fixing lands in land slots and at a frequency that makes them roughly as frequent as an uncommon, which means you will see an average of 9 dual lands per pod. In sets that support multicolor decks, like SNC or DMU, you get common fixing lands, which usually means 24 lands per pod.

Prioritising fixing early is not a thing for every set. You want to be in the situation where multicolor decks are supported, and in two color formats, fixing is a bit less of a priority. But when you are in the right format, like DMU is, what benefits will you get from focusing on fixing early? Well, here comes the second heuristic. But let’s start from the beginning.

Multicolor cards are arguably more powerful than the cards with the same mana value which are mono-colored. It makes sense for them to be – a card costing 1G will be naturally easier to cast than a card costing RG. You have to put much more care into building a deck that will reliably cast the RG card on turn 2 when you most likely want it to be cast, even in constructed. But in limited the difficulty level is particularly high, because mana bases are, simply put, not good. That is why almost every set has a cycle of signpost uncommon multicolor cards. They often set a tone to the color pairs and are frequently very powerful. And if you draft them well, they can be…

Signpost Rewards

In my original ALSA exploration, dual lands were not the only category of cards that stood out. So were the multicolor signpost uncommons, but in the completely other direction. Signposts are picked very highly in pack 1 but the later in the draft, the later they go. This was true in every format I looked at. Take a look at DMU. There are two types of signpost in this set so that gives us more data to see the trend. First group of signposts are the ones that have two colored pips in their casting cost, like Baird, Zar Ojanen or Lagomos.

These are relatively easy to cast, but still, there is some difficulty added to them. Namely, you need to either want to be in their color combination or be in their color combination to play them. Early in the draft, in pack one, Arena drafters pick the signposts relatively early.

Fig 2: Wheeling rate of signpost uncommons with 2 color pips in the cost.

Most of the cards from that category wheel only 2-11% of the time according to the 17Lands.com data. The exception is Vohar, the Dimir looter who people don’t like drafting, despite the card having a pretty good win rate on 17Lands. 2-11% wheel rate means cards are picked relatively high. But that changes in packs 2 and 3. There are large differences between individual cards, but across the board two-pipped cards wheel more frequently. Excluding the even more underestimated Vohar, signposts wheel 7-25% of the time in pack 3.

The size of this effect will depend partially on the card’s color. For example green signposts show the smallest increase in wheeling as green being the hub of multicolor can effectively splash signposts from multiple color pairs. On the other hand, Baird has a larger increase as it finds home almost exclusively in a focused WR deck. Early in draft a card as powerful as Baird may tempt someone to try and play WR, so it will wheel rarely. But if there won’t be any dedicated WR drafter in the pod, in the later packs nobody will be interested in picking it, making it go much later.

The second cycle of signposts is more difficult to cast. Each spell has 3 colored pips, which already makes it harder to cast. But even more importantly, this means one of the colors requires double pips to cast. Having exactly the right combination of colors is not trivial, so the cheaper the signpost, the less likely you will be casting it on curve. Examples of those signposts are Aron, Rona and my personal favourite, Uurg, Spawn of Turg. And if you are wondering if Uurg is related to Turg – indeed they are.

In early draft if you pick a card like that and want to play it you better make sure that the double-pipped color is one of your main ones. Later in the draft, even if you are splashing the double-pipped color of the signpost, it will not be trivial to cast it in most cases. And you see it in the data.

Fig. 3: Wheeling rate of signpost uncommons with 3 colored pips in the cost.

The gap between wheeling in the first pack (5-12% excluding the generally unwanted cards) and pack three (15-27%) is even deeper than in case of the first batch of signpost. Yet again some of the differences are deeper than others, but the direction of those differences is uniform across the board – signposts go later in later packs.

Flipping Psychology

Knowing what is going on is half of the battle. The other part is figuring out how to gain advantage from that knowledge. In this case we learned from the data of two trends that go in opposite directions. The logical conclusion is to go in the opposite direction. Why not try prioritising land fixing early, and reaping the rewards in good cards going later in pack three when you do not need any more of the fixing others at your pod are by now fighting for?

This strategy is not without its drawbacks. Prioritising fixing means you may not have enough playables later – but with the quality of cards in sets nowadays, you are almost certain to get enough from three or more colours if your fixing is up to scratch. I applied that strategy to draft three color decks in SNC with success: the aim was to get a few good cards early that set a general direction, late focus on fixing, only picking prime cards over it and later in the draft getting great cards late, benefitting from the natural tendencies I described above.

But not every set is fit for such approach. Firstly, this strategy requires a set to support three or more colored decks. Only those sets have enough fixing to prioritise them so do not try it on a regular two-color format where playing multicolor decks is a detriment. You can learn more about the characteristics of two and multicolor sets in my last article. Secondly, it is better if the set has general themes spanning multiple archetypes, not focused and distinct decks. DMU is a great example of such a general synergy set. But with those conditions fulfilled you can try to win with our human weakness, get your fixing first in place and patiently wait for rewards to come.

Conclusions

Some ideas can be only addressed properly when we look across multiple formats. But when we do – you can materialise ideas that work universally rather than falling into traps of extending the format specific truths on to other sets. The procrastination of fixing and prioritisation of identity defining cards in the early packs are such universal trends. But that is only the beginning of discovery in the cross-format comparisons.

If you are interested in these two heuristics and other draft trends from DMU – my seminar is out on YouTube. I talk about the contents of this article and more, including trends in combat tricks, wheeling lands in DMU and why fixing is self-correcting.

Enjoy our content? Wish to support our work? Join our Premium community, get access to exclusive content, remove all advertisements, and more!

MTG Arena Zone Premium
Sierkovitz
Sierkovitz

I am a limited player, who mainly skips playing in order to analyse the limited data using 17Lands.com. I run a podcast: Magic Numbers, where I try to use data to let you improve your limited game play, find out which heuristics work out and which common ideas are not well supported by data.

Articles: 11

Leave a Reply