Phyrexia: All Will Be One Quick Draft Guide: Drafting with Bots Using Data

Quick Draft for Phyrexia: All Will Be One (ONE) is now available and it is a good idea to give it a try. Why? You may get a slightly different experience of ONE draft and having a fresh approach towards the format is a nice thing to have several weeks in. This article aims at convincing you why Quick Draft is different, and give you some ONE specific tips on how to better navigate it using 17Lands data to identify good strategies and important cards.

Mr. Roboto

Quick Draft is different from Premier Draft because instead of drafting against other humans, you draft against bots – software drafters who developers programmed to imitate humans. But despite best efforts to mimic real drafters as much as possible, bots are markedly different, which opens up several possibility to outsmart them and end up with best possible decks. A task made much easier if you have some trusty data on your side.

There are several key features of bots that you should know to put yourself in a favourable position. Firstly, we do not know exactly what are the principles under which bots operate, but most likely there are several personalities they are given. So the path of your draft will in part depend on which bots are in the pod with you. I can only assume, but there must be things like red bot – a bot with preference to red cards, or a bot that drafts aggro, so one that picks aggressive cards higher. We can only guess about their true nature, but one thing is quite likely – bot’s personality will be consistent during the draft. That is very un-human, as in a human draft your pod neighbour might all of the sudden pivot to a different color, ruining your plan. Bots should be more consistent so reading the signals in a draft bot should be slightly easier and bring higher rewards.

Secondly, even if there are several personalities, bots will have some universal traits. They will all value cards at a similar level, their preferences will be some minor tweaks in their pick lists. If they do have universal properties it is essential to identify cards that are most impacted by those preferences and incorporate this knowledge to your strategy. One thing bots do have universally across all the formats – they shamelessly rare draft. You can see it by comparing Average Last Seen As, a 17Lands metric of how late do the cards go in a draft, between Quick Draft (QD) and Premier Draft (PD).

In the graph above I plotted ALSA of all the rares in the format in both Quick and Premier Draft. All the dots above the diagonal line are cards that are picked earlier in Quick Draft. All the dots below the diagonal line are cards that go later in Quick Draft than in Premier. There is only one problem. Not a single dot is below the line. Every single rare in the format is picked higher by the bots than by the humans. This is of course a careful strategy by WotC – they want to limit rare drafting to boost sales of packs and wild cards. But this also has an impact on draft. Based on those numbers, in an average Quick Draft you will see ~4 rares. In a Premier Draft, ~7 rares. This means that if your strategy in a draft is to get fixing early, pick good cards and hope in pack 3 you will be rewarded with getting passed the rares others con’t play, this strategy is risky in Quick Draft, as the bots pick rares even if they have nothing to do with their deck.

But every cloud has a silver lining. This means some really good uncommons will go later as bots will first pick a bad rare over, say, Annex Sentry. This means you need to plan your strategy around commons and uncommons and that you are less likely to play against busted bombs, a feature some will enjoy quite a lot.

Fixed Pick Order

What bots most likely have is a fixed or semi-fixed pick order. This means that a bot will always pick card X over card Y – or will do is Z% of the time, which allows us to maybe not figure out their pick order, but at least get a good grasp of which cards bots under and over value. And there is plenty of differences. Remember the previous graph, where all the rares were overdrafted by the bots? It is a very different picture, if you look at the commons.

You can see there are points at both sides of the diagonal line, which means that some cards are picked higher by the bots, but there are also some that the bots let go later in the draft. That is good news for two reasons. Firstly, this means that tere might be strategies to gain advantage from the cards bots undervalue, secondly, the draft experience should be different from the Premier Draft.

Openness

But before we dive into individual cards, let us look at a broad picture. ONE is predominantly a two color set so I will focus on looking at the two color pairs. To start, let us look at the color pair openness. If you read my previous article, I looked at how many cards with a GiH WR of 56% or more you will see on average in a draft for each color pair.

This measures the average openness of a color and shows you which colors are safer bets in a draft and which are a bit of a gamble when you decide to draft them. To understand the differences between Quick and Premier Draft we can compare openness of each color pair in both types of events. To avoid comparing apples and oranges, I look at how many cards with Game in Hand Win Rate >56% in Premier Draft you see in both types of competition and look at only commons.

As you can see the differences are minor, but we see 5 color pairs somewhat open in both formats, and 5 color pairs that just do not have many good commons, therefore you never see good cards for those. WB and WR are slightly more open in Quick Draft, you see 1-2 more good commons per draft on average. On the other hand WU and RG are slightly less open, but the difference are ~1 card less seen per draft. This seems small, but can such minimal differences impact color pair play and win rates?

First thing that strikes me in this graph is an almost universal drop in play rate of the 5 lowest win rate color pairs. This means 17Lands users actively avoid them. With 10 color pairs you would expect each of them be played around 10% of the time but Simic, Azorius, Rakdos, Izzet and Dimir are all in the 2-5% range, played 3 times less than expected. And they are played less than in the Premier Draft. This is most likely because Premier Draft contains early format data when we were still discovering decks and were willing to try experimenting with blue decks, while QD data is only from the last few day with a much more stabilised metagame.

On the other hand, 4 of the 5 top win rate archetypes are played more frequently. Only Gruul saw a minimal drop in play frequency, even though it is still the most played archetype at ~25%. Biggest increases in play frequency are in three white archetypes: Boros, Selesnya and Orzhov, but Golgari gained some as well. 17Lands users are usually good in reading the metagame so this is an interesting piece of data, but how does it impact the win rates?

It is striking how the play frequency data mirrors the win rate. Boros, Selesnya and Orzhov with highest win rate increases, Gruul staying the same and in PD – but that means it drops from the highest WR color pair to a mere number 3. Azorius, Rakdos, Izzet and Dimir dropping heavily, even though they are drafted less heavily, so at least in theory they should be more open. Interesting ones are Golgari and Simic with nice increases in win rate, but even with those increases they are firmly in Tier 2-3. This shows 17L users correctly identified slightly more open colors, actively try to be in them and use it to win more.

But what makes some color pairs win more? Is there a method to it? Are there specific cards that look important?

Cards to Think About

To answer this question we will look at the ∆ALSA – difference between ALSA of a card in the QD vs. PD. The positive values mean a card goes later in QD, so if you are drafting with the bots, you will see more copies. Negative values means there will be, on average, fewer copies seen in QD vs. PD. This allows us to observe which card categories are more and less available and if there is any recognisable pattern there.

The first thing that stroke me while looking at this graph is just how many WB cards are on it. Crawling Chorus, Blightbelly Rat, and Duelist of Deep Faith are all going ~2 picks later, making them wheel more frequently. All those cards provide a Toxic/Poison early package. Especially Chorus is an uber important piece for any deck involved with poison. Dropped on T1, it can easily deal 2-3 toxic, and even after it dies, you are left with a Mite. Not an impressive body, but in a deck with poison theme the Toxic 1 and keeping board presence are very welcome traits.

Apart from the mentioned early Toxic creatures, we have several Toxic support cards: Whisper of the Dross helps pushing small creatures through and proliferating poison above the corrupted threshold, and Offer Immortality and Zealot's Conviction do similar things. Especially Zealot's Conviction with higher availability of Choruses (Chori?) can be a much stronger card here than in PD. Charge of the Mites will help you with wider board to push last few poison damage in decks that focus on toxic kills (see my article on toxic strategies from before the release for more details – I think it aged relatively well).

Lastly, Dune Mover is a solid role player in the format also being a toxic 2-drop, and who know, maybe will allow shenenigans like splashing a Necrogen Rotpriest in some deck to make poison kills more consistent? This can be aided by higher availability of Terramorphic Expanse the best common land in the set.

Other important cards on the list include Rustvine Cultivator. And good that it does include it. A 1 mana 1/2 is exactly what other decks need to counteract 1/1 Toxic creatures. And Cultivator will routinely wheel so if you open it you don’t have to worry about considering it as a first pick – you should het all copies opened at the table. And ramp capabilities are nice to have too. Red gets much fewer goodies, but that results from red being severely underdrafted in the PD. Still, Molten Rebuke might be a decent pick in a format where Boros is the highest win rate archetype. And Blazing Crescendo grew to be the best common combat trick in the format.

But being more open doesn’t mean card will do better. Which commons increased their win rate the most compared to the PD?

In this graph we also see a bunch of corrupted and toxic enablers and payoffs. Sinew Dancer and Incisor Glider benefit greatly from the higher availability of the Chorus. Pestilent Syphoner, though still not a great card, does markedly better here. Flensing Raptor is even better than in PD. We also see improvements in Basilica Shepherd, Indoctrination Attendant, Whisper of the Dross, and Compleat Devotion. Things look great for White based Toxic decks in general. Other notable cards? Hexgold Slash, quite logically, does better in a meta where toxic is stronger. Also Chimney Rabble continues being great – make sure you start using this card much more if you still don’t.

But with every winner there have to be losers. Some cards are valued much higher by the bots, leading to their lower availability for decks. If a card in question is key for certain strategy, that strategy will be seriously hampered.

In cards that the bots value more highly, you can instantly see a trend. Quicksilver Fisher, Vivisurgeon's Insight, Gitaxian Raptor, Meldweb Strider, Meldweb Curator, Eye of Malcator, Glistener Seer – all blue spells. Bots value blue commons much higher that humans in PD. This is bad news for blue mages who want to dip into bot drafts. Blue was already underperforming in PD so further limiting availability of some of the key cards, like the Raptor or Seer is definitely not going to help. Especially bots’ eagerness to pick Eye of Malcator hit the WU artifact decks, as you can see from the archetype win rate data.

Apart from that one interesting card on the list is Testament Bearer One of the cards I found better and better as the format progresses, but that in part was because I was able to get those late. With bots drafting them early that appeal is gone. Worth noting – bots do like the Sphere lands. But you most likely should not. Tapped lands that give value late in the game are not exactly what you want in a fast format with loads of impactful one mana spells. Maybe bots try save you from including them in your deck and lowering your win rate?

Blue also dominates the list of cards that lost the most in terms of win rate in comparison with the PD. There just seems to be no good news for blue lovers in ONE. ☹

Conclusions

I hope this outline will help you to switch to QD in an easy manner and you can transition while improving your win rate. Based on the numbers I shared, the winning strategy is to heavily bias yourselves to start the draft either white or red and then, based on what is open select the colors from white, red or green. Of course this is a rough estimate – if bots pass you all the great black cards you can try WB or even BG, but keep in mind that those seats will not come together frequently.

One big take home message from this data is that you should try to avoid blue if possible. Bots draft blue cards higher than humans so you will be very unlikely to make enough playables and because of rare drafting you will also never get that late Blue Sun's Twilight like it could happen in PD, so the risk has little reward. I rarely advise to avoid color, but with bot pick orders being pretty fixed, this is one of those rare situations. By all means play around with blue in PD – there are interesting blue builds that come together routinely, so it is a color worth exploring.

Being able to apply poison or being able to deal with an early poison offensive will win or lose your drafts. So make sure you draft accordingly. Hexgold Slash and Rustvine Cultivator are your best friends against Chorus on turn 1 so keep them in mind during the draft. On the flip side, if you are playing poison make sure you have a clear vision of how you want to end games – don’t fall into a trap of a deck that deals 6 poison damage and then can’t win by neither poison or damage – you want to either commit to 10 poison, get to 3 poison and use Corrupted synergies not caring about further poison or deal zero poison and not care about it at all.

If you want to learn a bit more about this topic, including data from uncommons, make sure to watch my whole seminar on the topic: