Why We Shouldn't Open With Boss Arunka

Boss Arunka is the hero people ask about most, and the question is always some version of the same one: she goes first or second pick in high level RTA constantly, so why does the bot keep putting other heroes above her?

The usual suspicion is that the model is biased against her, or that it is punishing her for a season-long win rate that would look completely different if you drafted around her counters instead of walking into them.

Both are worth taking seriously, and both are testable. So I checked them against every draft she has appeared in since April.

Boss Arunka

24.7%

Pick Rate

47.0%

Win Rate

The Hero Everyone Picks

The Summer 2026 Season ran from April 4 to July 11. Over that stretch Boss Arunka was the most picked hero in the game, on 24.7% of every team drafted, 2.28 million picks in total.

She is also the only hero at the top of that chart sitting under 50%.

EVEN

59.8%

55.3%

54.5%

51.9%

47.0%

40%

Perfumer Byblis

Lady of the Scales

Frieren

Estelle

Boss Arunka

The Autumn 2026 Pre-season repeats it almost exactly: 19.1% pick rate, 48.65% win rate, second most picked behind Estelle.

She is not a weak hero, though. On a board with none of her counters she wins 55.9%.

That is the puzzle. A hero who wins 56% when left alone, and who gets picked more than anyone else in the game, should not be finishing a season at 47%. Something is eating the difference, and it is not happening in the battle. It is happening in the draft.

What the Model Is Actually Answering

Start with the bias question, because it is the easiest to settle. The draft engine has no tier list inside it, which is what a bias toward or against a hero would have to live in.

The predictor was trained on millions of finished RTA matches. It takes a full board, both teams, builds included, and returns a win probability. It never sees how popular a hero is, only how boards resolve. The guider then learned to draft by playing against itself millions of times, rewarded only for the win probability of the board it ends up with. Nobody ever handed it a ranking of heroes by win rate or by pick rate.

So "she is picked constantly, therefore she should be recommended" is not an argument the model can represent. There is no pick-rate ranking in there to be biased by.

What it does compute, at every turn, is one question:

For Boss Arunka on turn 1, that question has an unusually bad answer, and it has nothing to do with her kit.

Many Counters, and Better Openings Than Her

Every strong hero has counters. Hers are unusual in two ways.

There are a lot of them. Fifteen heroes hold her at or under 45%, each measured over at least 20,000 games.

Boss Arunka win rate, worst matchups

Notos

33.4%

93k

Perfumer Byblis

38.9%

723k

Shepherd of the Dark Diene

39.7%

205k

Frieren

41.1%

437k

Monarch of the Sword Iseria

41.2%

101k

42.1%

42.4%

42.6%

42.6%

43.9%

44.0%

44.1%

44.2%

44.5%

44.5%

They are heroes the opponent already wanted. Six of those fifteen were top-15 picks last season. Nobody spends a pick to punish her, they just draft what they were drafting anyway.

Each one that shows up costs her a lot:

• None of the fifteen on the enemy team: she wins 55.9%.
• One of them: 43.2%.
• Two: 35.8%.

And they show up almost every time. When she is the opening pick, only 11% of drafts finish with a clean enemy board. She wins 60.9% in that 11%, and 42.6% in the rest.

So she is not weak. She is a hero who is only strong in a version of the draft you get one time in nine.

And that is only half of it. The model is never asked whether Boss Arunka is playable on turn 1. It is asked which opening is best, and the openings it compares her against are these.

EVEN

59.2%

55.7%

54.4%

52.0%

48.8%

44.6%

40%

Perfumer Byblis

Notos

Lady of the Scales

Rhianna and Luciella

Estelle

Boss Arunka

She was the second most common opening in the game and the weakest of the common ones. A hero can be perfectly fine and still be the wrong opening, because the slot has an opportunity cost. Opening Byblis instead of Arunka is worth about fifteen points of win rate, and the model does not need to dislike Arunka to notice that.

The Turn Decides Everything

The second suspicion was that her 47% is a pre-draft average, and that opening with her leads to a different draft than taking her late. That one is right. Here is her win rate by where she landed in her own team's pick order.

Same hero, seven and a half points apart, and 60% of her picks land in the worst slot.

The earlier she appears, the more draft the opponent has left to answer her with. Open with her and 64% of enemy teams end up holding at least one of her fifteen counters. Take her third and that falls to 48%. Even Legend players open with her 70% of the time, and even there she finishes below her bracket average.

So the 47% is not one fact about her. It is a losing branch and a winning branch blended together, and most players are standing on the losing one.

When the Bot Does Want Her

None of this makes her unpickable. She appeared in 7,029 recommendation sets over the last 30 days of live drafts on the site, and there are plenty of boards she eats.

Boss Arunka win rate, best matchups

Successor Taeyou

62.2%

54k

Zio

59.8%

290k

Spirit Eye Celine

56.3%

90k

Bystander Hwayoung

53.7%

88k

Tori

52.6%

121k

52.2%

51.4%

51.4%

51.2%

50.7%

Look at the best matchups and her role is clear: she is strongest against cleave and against single target DPS.

That reinforces taking her late. By the third or fourth pick you can see what the opponent is building and put her against it. Opening with her only exposes her to the matchups that counter her.

So take her third or fourth. That is where she is a 52 to 59% hero instead of a 45% one.

In Conclusion

If the model ranked heroes by pick rate, Boss Arunka would be number one in the game. It ranks them by the board you are standing on after the opponent replies, and on turn 1 that board is the one where they already have their answer.