Radiance becomes legal in Riftbound on October 23, and nobody has played a sanctioned event with it yet. So any tier list right now is a prediction. I built this one the way I build my college football preseason rankings: start from what carried over from last season, add what the new season has shown so far, and let real results take over the moment they arrive. It gets graded at the bottom on November 13, misses included.
I play in New York, which is a very competitive scene. Most weeks I've been creeping from 0-3 toward 1-2. This set I want to play a properly competitive deck, which is part of why I wanted a list I could trust more than my gut.
At a glance
- S: Kennen, Vendetta's best deck by a distance.
- A: Riven, Kai'Sa, LeBlanc and Irelia. Riven is the only new legend this high.
- Most new legends are a mystery. Evelynn and Mordekaiser sit in C, with ranges wide enough to land almost anywhere.
- Every placement comes from data, not from anyone's opinion, mine included (how).
Jump to: The tier list · What set launches teach · The calls · What I'm playing · How this is made · Scorecard
The tier list
The score compares a legend's top finishes with what an average deck would get with the same number of players. ×1.00 is average; ×1.50 is half again as many. The tiers are fixed cuts on that score, so a tier means the same thing every week: S ×1.40 and up, A ×1.00, B ×0.80, C ×0.65, D below. "Likely rank" is where the legend lands 80% of the time once the uncertainty is simulated, and uncertain marks a score that could plausibly sit anywhere from D to S.
| Tier | # | Legend | Score | Likely rank | Top-10 chance |
|---|---|---|---|---|---|
| S | 1 | Kennen, Heart of the Tempest | ×1.59 | 1–6 | 99% |
| A | 2 | Riven, The Exile · new | ×1.23 | 1–34 · uncertain | 63% |
| 3 | Kai'Sa, Daughter of the Void | ×1.12 | 2–27 · uncertain | 59% | |
| 4 | LeBlanc, Deceiver | ×1.06 | 5–20 | 56% | |
| 5 | Irelia, Blade Dancer (Astral Heron) | ×1.04 | 5–18 | 52% | |
| B | 6 | Rengar, Pridestalker | ×0.96 | 6–25 | 34% |
| 7 | Master Yi, Wuju Bladesman | ×0.92 | 8–26 | 23% | |
| 8 | Jayce, Defender of Tomorrow (Aurora ramp) | ×0.90 | 7–28 | 22% | |
| 9 | Vex, Gloomist | ×0.89 | 7–31 | 25% | |
| 10 | Azir, Emperor of the Sands | ×0.87 | 8–30 | 18% | |
| 11 | Yasuo, Unforgiven (Spiderling swarm) | ×0.85 | 2–53 · uncertain | 37% | |
| 12 | Ezreal, Prodigal Explorer | ×0.84 | 8–35 | 18% | |
| 13 | Kha'Zix, Voidreaver | ×0.81 | 10–35 | 12% | |
| C | 14 | Akali, Rogue Assassin | ×0.78 | 10–37 | 10% |
| 15 | Rek'Sai, Void Burrower | ×0.76 | 13–37 | 5% | |
| 16 | Diana, Scorn of the Moon | ×0.74 | 13–41 | 7% | |
| 17 | Ornn, Fire Below the Mountain | ×0.68 | 14–45 | 5% | |
| 18 | Lux, Lady of Luminosity | ×0.68 | 7–55 · uncertain | 16% | |
| 19 | Evelynn, Agony's Embrace · new | ×0.67 | 3–63 · uncertain | 25% | |
| 20 | Mordekaiser, Iron Revenant · new | ×0.66 | 5–62 · uncertain | 21% | |
| 21 | Viktor, Herald of the Arcane | ×0.66 | 11–52 · uncertain | 9% |
D tier: Lillia, Ziggs (new), Fiora, Ekko (new), Sivir, Lucian, Nasus, Draven, Jhin, Jarvan IV (new), Orianna (new), Seraphine (new), K'Sante (new), Master Yi (Wuju Master), Poppy, Pyke, Jax, Mel, Rumble, Ivern, Ambessa, Zed, Renekton, Renata Glasc, Shen, Vi.
Not enough data: Annie, Volibear, Lee Sin, Teemo, Leona, Darius, Jinx, Ahri, Garen, Sett and Miss Fortune. Almost nobody played them at Vendetta's big events and nobody has in Radiance yet. That's no record either way, which isn't the same as D.
What three set launches teach
Riot publishes a write-up for every Regional Qualifier (RQ): the top 8, each legend's best finisher, and how many players brought each legend. Nine of them cover three formats (Spiritforged, Unleashed and Vendetta), which is enough to check what actually predicts a new format.
About 60% of a deck's edge survives a new set. A legend's results correlate 0.69 from Spiritforged to Unleashed and 0.50 from Unleashed to Vendetta, against 0.84 between events inside a single format. Last season matters, just less than it feels like it should.
New legends aren't weaker at launch. At each set's first RQ they did as well as the older legends or better; at Barcelona, Vendetta's first, they converted ×1.08 to Day 2 against ×0.96 for everyone else.
The crowd picks the winners. Once players register for a big event, the share of the field on a legend predicts its results better than last season does: 0.57 to 0.79 against 0.42 to 0.59 across Vendetta's three RQs. And the most-registered new legend at each set's first RQ became a top deck three times out of three: Draven, LeBlanc, Kennen.
What doesn't predict much is counting new cards. How many new-set cards a legend's best list ran had no relationship to whether it got better (0.00 and 0.08). The cards that matter show up as players choosing the deck.
The calls
Kennen, S. It converted ×1.76 to Day 2 across Vendetta's three RQs, well clear of anything else. It was also the most-played legend at Radiance's biggest preview, and every one of those lists added Trifarian Captain. Its likely range, 1st to 6th, is the tightest on the board.
Riven, A. It finished first at NovaRiftbound's 74-player preview, an online event before the set is legal, with three more of its nine players in the top 16, and their lists already agree on a core. Nine online players is very little evidence, hence the uncertain flag. But Riven is the new legend the field is gathering around, and that pattern has held at every launch so far.
Kai'Sa, A. Twelve players at the same preview, one finishing 2nd, and every one of them picked up Bilgewater Tracker. The model has her 3rd. My own read is that 3rd may be too aggressive for a legend this unproven: she was barely played at Vendetta's RQs, which is why her range runs from 2nd to 27th. One more wrinkle: Origins, Spiritforged and Proving Grounds legends, Kai'Sa included, can no longer win Best-Of prizes at Regional Championships, which may keep her rarer there than her strength deserves.
Irelia and Riven share a lane. Irelia's Astral Heron build is in A and her En Garde build in B, and Riven competes for the same colours and the same players.
The Spiderling meme. One Yasuo list, 39 Spiderlings and a Windrider, was copied by 12 players across several teams at the Nova preview and finished 3rd, 5th, 6th, 9th, 10th and 15th. Yasuo had no players at all at Vendetta's RQs, and Radiance printed several answers to wide boards, so it lands in B with a range from 2nd to 53rd. I don't know if anyone will play it seriously. But maybe!
What I'm playing
Riven. I want to try a genuinely competitive deck this set, and I like how she plays. Her legend makes every unit you play at a battlefield you control enter ready, so new units can attack or use their abilities straight away instead of waiting a turn. Hidden cards flipped there come in ready too, since a hidden card can only be played where you've hidden it. Riven, Broken Blade gets a choice every time she exhausts, each once a turn: buff herself, spend the buff to stun an enemy, or spend it to ready herself. Blade of the Exile, her gear, gives +3 Might and can be equipped by exhausting a unit instead of paying Power; exhausting Broken Blade to do it sets off her trigger. The whole deck rewards holding battlefields.
Sivir. The model puts her in D, and I'm playing her anyway, because I love her. My build leans on Radiance's Hidden cards. Sivir's legend can make a Gold token when you recycle a rune to pay a Power cost, and readies again whenever an enemy unit dies; hiding a card costs Power, so hiding and Gold feed each other. If it works, she'll show up in the results and the model will move. If it doesn't, at least I'll have lost with my favorite.
How this is made
Each legend's score blends three things on one scale:
- Last season. Day 2 conversion at Vendetta's three RQs (about 6,000 players), plus the top quarter at events whose decklists cover most of the field. It's shrunk toward what a legend that popular converts, so a legend with five players can't post ×3, and a legend nobody played isn't a blank slate.
- This season's popularity. Each legend's share of players at the Radiance previews, topped up with its share of posted decklists, translated through the link between popularity and results measured at Vendetta's RQs. Until the first RQ this leans 60/40 toward last season; as real registrations come in it moves to 30/70, which is what the backtest supports. A brand-new legend is scored on popularity alone.
- Results. Radiance results take over as they arrive. The previews count half, because they're small, online and before the set is legal.
A legend is split into separate builds only when its decklists form distinct, stable groups (Yasuo's Spiderlings, Jayce's Aurora ramp, Irelia's two builds).
I also kept a list of card-by-card reasons, like Kai'Sa's Bilgewater Tracker and Kennen's Trifarian Captain, but none of them is in the score: past launches give no evidence that that kind of reasoning predicts. The only Radiance evidence so far is 101 players across two online previews.
Sources: Riot's RQ "Top Decks" articles; decklists and standings from riftbound.gg; the two Radiance previews, NovaRiftbound's $2,000 Online Showdown 5 (74 players) and Haven's Online Challenger Series #5 (50 players). I built the model with Claude as a pair programmer; the choices about what to measure and what to say are mine. It updates daily, and the next edition after the first sanctioned weekend will say what moved.
Scorecard
Graded on November 13 against the RQ-sized events from the first three weeks of sanctioned Radiance play, +1 for a hit and −1 for a miss:
- Kennen converts above average (×1.00) to Day 2 or the top quarter.
- Kennen is the best-converting legend among those with 20 or more players.
- Riven converts above average.
- Kai'Sa converts above average.
- No new legend other than Riven converts better than Kennen.
- Yasuo's Spiderling list converts below ×1.40, the S line.
- At least one A-tier legend converts below average.
- No legend from "not enough data" converts above ×1.40.
Give it time.