Let’s Optimize: Rek’Sai Pyke Lurkers

Dr. LoR here with another article from the ‘Let’s Optimize’ series! My goal in this series is to use objective statistical data to figure out the best builds for popular meta decks.

The most popular deck of Patch 2.11 has easily been Rek’Sai Pyke, otherwise known as Lurk. For those who are looking for a comprehensive guide on the deck, check out Asher’s excellent deep-dive (RCCG premium required).

Although many players considered this deck to be ‘built on rails’, it has a surprising amount of deck-building options and depth and still has not been fully solved, especially as the meta evolves around it.

The deck started with close to a 20% play rate in Platinum+ but has recently dropped to a 13% play rate (data from last 3 days). Perhaps it’s because it has lost its novelty – or because it has proved to be deceptively difficult to pilot, which led to a mediocre 51.7% win rate – not exactly befitting of Tier 1.

Let’s see what the best way to build the deck is in this current meta! For the purposes of my analysis in this article, I’ve pulled data from all Lurker decklists with at least 20 matches on Mobalytics in Platinum+ for the period of July 6 – July 12, purposely excluding the first 5 days of the patch as the “experimental period.”

This made for a total of 34.6k matches played with 381 different lists ranging from 20 to 3411 matches per list, with the top five lists accounting for 27% of the matches.

---

Cards Data

• The Playrate super-column shows how often certain cards are included in Lurk versions (‘Overall’ column), and in how many copies (‘0’, ‘1’, ‘2’, ‘3’ columns). The darker blue the shading, the bigger the number. E.g., 17.4% of matches played are with lists that include 3x Feral Prescience.
• The Winrate super-column shows the average win rates for decks that have that many copies of a certain card (‘0’, ‘1’, ‘2’, ‘3’). E.g., decks that run 1 copy of Feral Prescience average a 53.9% win rate, and decks with 3 copies have a 50.7% win rate. You will see that some percentages – for example the numbers in The List column – are written out in grey. It means there isn’t sufficient data behind it, so we can’t fully rely on it to accurately compare with other numbers.
• The Drawn WR column shows the percentage of games won where at least one copy of that card was drawn; the Mulligan WR refers to the percentage of games won if this card is in your opening hand. Note that cards with lower play rates or played in fewer copies have less reliable Drawn WR and Mulligan WR data.
• Keep % is the percentage of time players keep these cards in their mulligan. It’s important to understand when reading this data that when you mulligan a card away but draw a duplicate copy, Mobalytics still counts this case as a ‘keep’. Because of that, I believe that the ‘keep’ number even as high as 15% means that the card in reality is almost never kept.

---

Optimizing The Deck

For the purposes of this section, I narrowed down my initial data sample to 132 lists with at least 40 matches in Platinum+ and calculated their win rates using Bayesian smoothing (it is a technique that allows avoiding statistical flukes. If you want to learn more about the concept, check out this explainer.)

Sadly I have to ignore a portion of the decklists that do not meet my requirements. This is a necessary evil since they would offer unreliable data: the win rates are easily skewed by a single good or bad pilot, and Bayesian Smoothing would also push these numbers toward the average so much that we don’t learn much from them.

Due to the unique synergistic nature of the Lurk archetype, I need to further define a few concepts that we will be using in the analysis below.

Lurk density is how many Lurk cards we play. While we could technically play all 40 Lurk cards (and one list in my sample indeed goes as high as 39 Lurk cards!), most decks have a Lurk density between 60-82.5% (24-33 cards). Lurk density determines how likely you are to hit a lurk trigger naturally, without manipulating the top of your deck via Predict, Bloodbait, Bone Skewer, or Call the Pack. 60% is almost as low as one can go while deck-building because your champions and core followers are almost always played as 3x. Lower Lurk density decks play more Predict cards to make up for the lack of consistency and to increase the chance of Lurking with your champions.

Deck speed refers to how quickly we are aiming to win the game with our Lurkers build. I will refer to decks that run 5x or more big Lurkers as ‘slow’ (61% of decks), and lists that run 4x or less big Lurkers as ‘fast’ (39%). Slow decks have a marginally higher win rate than fast decks overall (1.1% higher). Fast decks tend to have lower Lurk density because they play fewer of the big Lurkers.

---

Champions

Rek’Sai and Pyke are no-brainers at 3x. They are your main win conditions and the reason some decks choose to play Predict cards. The difference between hitting them on your Lurks versus regular Lurkers is huge.

• Rek’Sai offers a quick way to end games when you can turbo-level her. If the game goes a few turns further, her level-up conveniently gives you the fuel to finish the game. However, a big portion of the meta is prepared to deal with Rek’Sai, which often spells disaster if you don’t have a Bone Skewer ready. Concussive Palm, Icevale Archer (and other frostbite effects), as well as Arachnoid Sentry all offer clean answers while developing an additional blocker for your smaller Lurkers. The amount of these decks you see in your local meta should inform you whether you want to take a more or less aggressive approach. On the other hand, turbo-leveling is your best bet against aggro decks.
• Pyke offers a different avenue of attack, giving you some much-needed interaction and long-game potential. Most decks have 0-1 answers to Death from Below. On top of that, Pyke’s champion spell Bone Skewer serves as a synergistic piece of interaction that is perfect for him.

---

Small Lurkers

The cheap Lurkers are all unanimously considered as must-includes.

• Sharkling and Xer’sai Hatchling are the only 1-drops in most versions and they are important to start stacking Lurk. These two units have the highest Drawn WR’s. Easy 3x.
• Snapjaw Swarm lets you Lurk on your opponent’s turn and can even sneak lethals versus some control decks; it has the third- highest Drawn WR in the deck. Easy 3x.
• Redfin Hammersnout offers the deck one of its few tools of interaction but is otherwise underwhelming, with a medium drawn win rate. Nonetheless, almost everyone plays 3x.
• Xer’sai Caller is one of the most important lurkers – it has a beefy body (surviving 2-damage is critical) that helps set up your next Lurk. Its Drawn WR is decent and literally every deck plays 3x.
• Additional 1-drops: A few players have experimented with increasing the number of 1-drops to increase the chance of a turn 1 Lurk when attacking on odds. The list with 6x 1-drops (assuming you go for a hard-mulligan) gives you an 80.8% chance of drawing at least 1-drop on turn 1. 7x 1-drops gives an 85.9% hit rate, 8x is 89.7%, and 9x is 92.6%, so there are diminishing returns. On the other hand, every additional non-Lurk card reduces your Lurk density by 2.5%. Assuming these extra 1-drops come from cutting Lurk cards, 8x 1-drops gives you the highest chance of hitting a natural turn L lurk. The WR data suggests that slow decks benefit from additional 1-drops as long as you have less than 6 big Lurkers. On the other hand, fast decks have too low Lurk density so don’t benefit as much from additional 1-drops.
  • What options do we have? Pool Shark is a crappy body and you will rarely be able to use the Fleeting draw on turn 2, but it offers the upside of a delayed draw in the late game. Treasure Seeker is a decent body that essentially draws a mediocre card, but that package is surprisingly great and it’s become popular across many Shurima decks. In short, they’re 1-drops that aren’t terrible late game draws. Decks that run them actually show significantly higher WRs so I recommend trying 2x of one or the other in both slow and fast decks. Dunekeeper, Baccai Reaper, Bloodthirsty Marauder are other options that some lists have tried but there is far too little data to say much. I like Treasure Seeker the most since it helps with the aggro matchups more.

---

Big Lurkers

The big Lurkers are much less agreed upon and how much are you planning to rely on them is one of the main decision points in building the deck. 6% of decks top out at Pyke whereas others play as many as 9 copies of the big Lurkers! The most common counts are 7 (24%), 5 (23%), 4 (19%), and 6 (11%). By playing more big Lurkers, you end up with higher chance to hit natural Lurks, and so you are less dependent on Predicts. However you will more often end up with less chance of explosive champion-driven draws.

• Xerxa’reth, The Undertitan is one of our big 6-health units that help finish the game with Overwhelm, who is also a reasonable defender. Xerxa’reth is a conditionally-powerful unit – but it the Spellshield matters a lot if you can trigger it. Xerxa’reth is much more commonly played than Xer’Sai Dunebreaker, being a 3x in 80% of decks and often the only big Lurker. Xerxa’reth’s Drawn WR is slightly lower than Dunebreaker’s but I suspect this is because the fast decks don’t play the latter. I recommend at least 1x Xerxa’reth even for fast decks but up to 3x are fine (1-3x all have statistically indistinguishable win rates). For slow decks, 1-3x Xerxa’reth again have statistically indistinguishable WRs but I think 3x should be right for a better curve.
• Xer’sai Dunebreaker is played much less, as a 3x in only 14% of decks (23% of slow decks). It can be better than Xerxa’reth when we’ve hit a reasonable number of Lurks, since a 6/6 or 7/6 overwhelm unit is quite menacing. Dunebreaker also actually has higher Drawn WR than Xerxa’reth, although I think this is driven by the overall slow vs. fast deck WR difference. I recommend 0-1x for fast decks and anywhere from 1-3x for slow decks (statistically indistinguishable).
• Jaull-fish is a significant jump to 8 mana but it offers a unique ability for a deck that is very light on interaction. It actually has the 4th highest Drawn WR in the 61% decks that actually play it. Although the card seems to have highest WR as a 3x, this difference is not significant, and is more an indicator that slow decks are better than fast ones. I recommend 0x for fast decks and 1x even for slow decks unless you face a lot of midrange decks.

---

Other Lurk cards

• Bloodbait is much maligned and perhaps for good reasons. It has the third lowest Drawn WR in the deck and WR for decks that play 0x are higher than those that play 1-3x. It does enable a busted combo with Rek’Sai – I suspect this is why some players are still advocating for this card. Still, the data are pretty clear that this bait stinks. Since you can’t reliably Lurk with champs without Predict, I recommend playing 0x Bloodbait unless you also play at least 3 (preferably 5+) dedicated Predict cards. And even then, Bloodbait is a 2x at most.
• Ripper’s Bay was bugged at the start of the expansion and didn’t work properly to increase Lurk chances, but even after its bugfix, it still maintains a tiny playrate. If you run a skeleton crew of small lurkers and Xerxa’reth, your deck still features 60% Lurk cards, so having a one-of Ripper’s Bay in play increases your lurk success rate to 84%, which is substantial. With a 75% Lurk deck, your success rate increases to 94%. So is a 19-24% increase in Lurk consistency worth a card and a mana? I think not. More importantly, hitting your champs on Lurks is what actually matters the most, and Ripper’s Bay doesn’t help with that.
• Call the Pack is a bit of a do-it-all card for the deck. It doesn’t offer card advantage per se, but does fill your hand with relevant options with the added benefit of putting the card of your choice (usually a champ) on top so you can Lurk with it right away. In decks that don’t run Predict, this is the main way to get enough power to close out games. It also gives the fast decks access to the big Lurkers. Although it has medium Drawn WR, I can’t imagine cutting any. The tiny amount of decks that went to 2x have lower win rates.

---

Predict cards

Running Predict cards is the other massive deckbuilding choice. Aside from Xer’sai Caller, decks in the sample can include as many as 9 total dedicated Predict cards. Since the deck can be built with as high as 80% Lurk cards, the purpose of predicting is mainly to hit Lurk with your champs (which happens about 40% of the time if all of them are still in your deck). The Predict makes a lot of sense with the faster gameplan, and indeed, 75% of fast decks (i.e., fewer than 5 ‘bigs’) play 3 or more Predict cards whereas only 50% of slow decks run 3+ predict cards. Fast decks want 6 total dedicated Predict cards, slower decks want 2, and the slowest want none.

• Feral Prescience does literally nothing but Predict, but that may be worth it to get those champ Lurks for the fast decks. It has the second lowest Drawn WR and Mulligan WR in the deck, which suggests it’s not great, even in your opening hand. For fast decks, 1-2x seems OK in terms os WR, but 3x is too many. Even for fast decks, 0x is the optimal number – although up to 2x is fine. For slow decks, at most 1x can be fine, but I recommend none.
• Ancient Preparations is similar but gives you the added benefit of turning on Shaped Stone and giving you a body in 2 turns. However, it doesn’t help you hit turn 1 Lurks and the 2/2 is less relevant later on. It has mediocre Drawn WR, so I recommend it mainly if you hope to use Shaped Stone to turbo-level Rek’Sai. For fast decks, 3x has the highest WR, significantly higher than 0x or 2x. For slow decks, I recommend 0x.
• Scrying Sands does arguably even less than Ancient Prep with the similar below-average Drawn WR. Its main benefits are doubling as a weak combat trick (in a deck that doesn’t really rely on them) and letting you Predict into an open attack. It also combos with Redfin Hammersnout – but otherwise the -2 attack doesn’t have much impact. I recommend 0x for slow decks and 1x-2x for fast decks, where it seems surprisingly effective, with a big bump in WR over 0x.
• Aspiring Chronomancer is the only commonly played non-lurk unit in the deck and actually has the 5th highest drawn WR in decks that play it. As a unit with Predict, it essentially guarantees a Lurk proc, and sometimes hits a champ. Fast decks play an average of 2 copies but have the highest WR at 3x (2x is similar). Slow decks average 1.5 copies, with 2x being the highest WR though the difference vs. 0x or 3x isn’t significant. I’d like to play some copies because we really only have one true 2-drop in the deck since Snapjaw Swarm can be awkward to play if your opponent has blockers, especially if we didn’t hit a turn 1 Lurk (which happens around 50%-65% of the time depending on our Lurk density).
• Careful Preparation is overcosted and not a good fit for this deck (or maybe any deck really). Not much data on it but it doesn’t look great.

---

Tricks, card draw, and interaction

• Bone Skewer plays a few interesting roles in the deck. It’s one of the only sources of interaction, acting sort of like a Single Combat effect. This has the obvious benefit of turning on Lurk and works particularly well with your champs. It used to be a lock as a 3x but the intrinsic card disadvantage seems to be an issue. For fast decks, 0x is the optimal number and for slow decks, 2x and 3x are similar, with a small edge to 2x (1x may also be fine but there’s very little data there).
• Shaped Stone is the most efficient buff spell in the game—as long as you’ve played a landmark first. However, Ancient Preparation is bad in slower decks and relying on 2-3 Preservarium is not worth it. If we do play Ancient Preparations, 3x is optimal for turbo-leveling Rek’Sai. If we don’t, it’s an easy 0x.
• Ruthless Predator was one of the main innovations to come out of the first few days of the expansion. Since the card is barely played in other Shurima decks that rely on Vulnerable, it isn’t immediately obvious how it fits in this deck until you realize that it can help level Rek’Sai 2 turns early AND it helps to maximize Overwhelm damage. For fast decks, 3x is optimal (assuming you don’t play the Shaped Stone package). For slow decks, 2x is the optimal number.
• Other vulnerable effects: Players have also experimented with other cards that grant vulnerable, including Merciless Hunter, The List, Unraveled Earth (which also helps with Shaped Stone), Exhaust, Ye Be Warned, and Rock Hopper (not shown in table). Vulnerable can turn your small lurkers into removal, aim your big lurkers to deal maximum overwhelm damage, and facilitate a level-2 Pyke board wipe. Unfortunately, there isn’t enough data to comment on these so I won’t. I will say that Merciless Hunter is so strong that I could see it making the list, but I can’t recommend it without more data.
• Card draw is almost a necessity in the fast decks, since they tend to run out of gas quickly. Fast decks tend to play 2-3x total whereas slow decks tend to play 0-2.
  • Preservarium is cheap draw that turns on Shaped Stone; 2-3x copies are better than 0x in fast decks. Slower decks need less card draw but up to 2x is still good is good with up to 6 big Lurkers (no statistical difference between 0-2x); with 7 or more, I recommend 0x.
  • Salvage costs 2 more mana but lets you use spell mana, is burst, and you get both cards immediately. Early data showed it performing well as a 2x, but more recent data show it to have low WR. Regardless, it was never played enough for the data to make strong claims. I can see its merits though if you prefer it over Preservarium.
  • Zap Sprayfin is an interesting option since the slow deck plays few spells and the most consistent two (Bone Skewer and Call the Pack) are important role players. There’s not much data here, but I could see 1x in a deck where it only hits Call the Pack and Bone Skewer.
• Rite of Negation initially wasn’t viewed as necessary when this was thought of as a pure aggro deck, but players have since come around to running 1-2x. 2x seems like too many in a deck that doesn’t want to sacrifice its units OR mana gems. 0x and 1x have similar WRs for the slow decks with a slight edge. 1x is best for the fast decks, with a significant bump in WR.
• Siphoning Strike was popular in early decks as a 1x in part because its interaction with Pyke leads to him getting +2/+2 with every kill. Unfortunately, it’s too narrow, with the lowest Drawn WR in the deck. Decks that include 1-2x, either slow or fast, have significantly lower WRs.
• Make it Rain is anti-aggro tech, but I don’t think it’s worth diluting Lurk density. While there is a list running 3x Ruinous Path provides a bit of reach and also heal against other aggro decks, but it’s again not worth diluting Lurk density.
• Ancient Hourglass, Bayou Brunch, Monster Harpoon, Playful Trickster, Rite of Calling, and Xenotype Researchers all appear in a few lists but don’t have enough playrate to say much about with data.

---

Recommended Builds

Given that there exist two distinct directions this deck can take in terms of speed, I think it’s appropriate to offer two versions of the final build.

In the fast version, we run Xerxa’reth as our only big Lurker, and only six Lurk 1-drops (67.5% Lurk density, 55% chance of turn 1 lurk). IN addition, we have 6 dedicated Predict cards (Ancient Preparation and Aspiring Chronomancer), Shaped Stone over Ruthless Predator, and no Bone Skewer.

In the slow version, we run 6 big Lurkers and 8x 1-drops (75% Lurk density, 67% chance of turn 1 Lurk rate). If you have trouble picking between the two, I recommend this version since slower decks have a higher win rate than faster versions overall.

Comments or feedback are welcome – good luck and have fun! If you liked this article, follow me on Twitter and watch for future data-driven breakdowns of popular archetypes and meta reports.