Celeste’s nerf has a tipping point, and it is 167 spirit power
Three of the patch’s twelve lines are not nerfs but redistributions: weaker below 167 spirit power, stronger above. In ranked, Celeste only crosses that threshold at minute 43 — in one game out of three.
On Friday evening, Deadlock received twelve lines of patch notes. Ten of them concern Celeste, and six are nerfs. On paper, the verdict is simple: the kite-flying witch got hit. The numbers — and the game files — tell a subtler story.
Why her
Celeste has come a long way. In late April she sat at 46.4% wins across all queues, one of the worst rates in the game. Valve lifted her twice this summer: five buffs on June 30, three more on July 28. The curve followed, steadily: 46.9% at the end of June, 48.4% at the end of July, 50.9% on patch day.
That global figure hides what matters. In ranked, from August 8 to 22, Celeste won 52.0% of her 73,210 games. But among Initiates, the first tier, she caps at 49.8%. Among Oracles, 53.8%. Among Phantoms, 55.1%. Among Ascendants, 58.3% — on just 405 games, but the direction leaves no room for doubt.
Celeste was not too strong. She was too strong at the top of the ladder, in hands that know what to do with an ultimate that bounces eight times.
What the patch says, and what the files say
Three of the twelve lines are not nerfs. They are redistributions, and they all cross at the same spot.
Shining Wonder, the ultimate. Base damage drops from 165 to 140, and its scaling with spirit power falls from 0.9 to 0.6. A clear nerf — except that the tier-2 upgrade, which already added +90 damage, now also grants +0.45 scaling. With tier 2, the ultimate used to hit for 255 plus 0.9 times your spirit; it now hits for 230 plus 1.05 times your spirit. The two formulas meet at exactly 167 spirit power. Below it, the new version hurts less. Above it, it hurts more than before.
Dazzling Trick, the barrier. Its tier 2 goes from "+80, 0.7 scaling" to "+70, 0.76 scaling". Do the maths: both versions are worth the same at 10 divided by 0.06, or 166.7 spirit. The same threshold, to the decimal.
Radiant Daggers. Spirit amp per stack drops from 8% to 7%, but tier 3 rises from 3% to 4%. With tier 3: 11% per stack before, 11% per stack after. Nothing moved for anyone who buys it.
Valve did not draw that line by accident. It is the signature of a calibrated nerf: weaken the early game, spare — even reward — the late game.
When the 167th point arrives
The question is when an ordinary Celeste reaches that threshold. The match table records every player's spirit power every five minutes; we pulled it from 73,208 ranked games played before the patch.
At minute 20, the median Celeste has 91 spirit. At minute 30, 123. At minute 40, 154. She only crosses 167 at minute 43 — and while a quarter of players get there by minute 36, another quarter still have not by minute 60.
A ranked Celeste game lasts 37 minutes on average. The result: one game in three (32.3%) ends with Celeste above the threshold. The other two are played entirely in the zone where the patch bites.
One detail deserves a careful reading: games that cross the threshold, Celeste wins at 62.3%, against 47.1% for the rest. Spirit power is not what wins those games: long, well-engaged games are what make 167 reachable. The threshold does not create the win, it signals it.
In practice, in game
It all comes down to two ability points. With tier 2 on the ultimate, a Celeste at minute 20 loses 3% damage per bounce, 2% at minute 30, next to nothing at minute 40, and gains half a percent back past minute 45. Without that tier 2, the same ultimate at minute 20 drops from 247 to 195 damage: minus 21%.
The message to Celeste players is clear: tier 2 on Shining Wonder is no longer a luxury, it is what cancels the nerf. And tier 3 on the daggers erases the other one.
The rest of the patch is not up for negotiation: Dazzling Trick's cooldown goes from 32 to 34 seconds, its tier-3 silence from 1.5 to 1.25 seconds, the ultimate's bounce range loses a metre — back to its June 30 value — and its base damage falls almost back to where it stood before the summer's two +10% buffs.
Three days in
It is too early to conclude, and we will not. But the first numbers exist: since Friday 11:40 pm, Celeste has played 9,059 ranked games and won 53.3% of them, against 52.0% over the previous two weeks. The gap sits within the margin of error. The direction, however, is not that of a hero collapsing.
We will be back in two weeks, as we were for patch 44, with enough games to settle it tier by tier. If the nerf bit where it aimed — Phantom and above — that is where it will show.
What we cannot claim
The formulas above come from the game files as the public catalogue exposes them: we checked that the live values (140 damage, +0.45 at tier 2, +70 and 0.76 on the barrier, 4% at tier 3 of the daggers) match the patch notes. Match data, however, does not tell us who buys tier 2 on the ultimate, or when. The "with or without tier 2" comparison is therefore arithmetic, not observation. If most players take it late, the nerf is heavier than our averages suggest.
Methodology. Ability values read on August 25, 2026 from the deadlock-api.com catalogue, which mirrors the live game files, and checked against the August 22 patch notes. Spirit power taken from the `match_player` table of the public SQL endpoint, field `stats.tech_power`, at the table's fixed timestamps (every five minutes after the fifteenth); 73,208 ranked Celeste games from August 8 to August 22, 21:40 UTC. Win rates per tier computed over the same window from the match's average badge; Ascendant rests on 405 games and Eternus falls short of the minimum sample. Post-patch period: 9,059 ranked games from August 22, 21:40 to August 25, 15:00 UTC, margin of error around one point. Global trajectory from our own per-patch snapshots, all queues combined.