Riot Games removes Mel's reflect mechanic: inside the W rework in League of Legends patch 26.22

At second 24 of the teamfight, Seraphine's orb flew into the middle of the formation, touched the ripple of Mel's reflect field, and turned back around. Five members of Seraphine's team were charmed by their own ally's ultimate. The clip was cut within three minutes, spread across forums within three days, and became the defining image of the ability Riot Games itself called the most frustrating in League of Legends.

Then Riot announced the rework.

Riot's balance team issued notice that the reflect mechanic on Mel's W would be replaced entirely, bundled with a tuning pass that runs across her whole kit. Patch 26.22 is expected on live servers roughly three weeks after the announcement. I re-read everything Riot released, and what held me longest was not what they changed but how they justified it.

They chose examples. They did not choose frequencies.

That is the starting point for everything below.

Context: an ability designed to punish, then consumed by its own punishment

Mel entered League of Legends with a fairly clear identity: a mid-lane mage who broke the opponent's rhythm through reaction. Within her kit, W was the load-bearing wall of that philosophy. Triggered at the right moment, W raised a reflective wave that caught every incoming projectile and fired it back at the caster.

The design idea is clean. Opponents must think twice before casting. Mel generates pressure without initiating; she generates pressure by making the opponent tie their own hands. That is psychological pressure, and in a game where every decision is measured in cooldown time, psychological pressure carries the same weight as damage pressure.

But the field had one fatal design flaw: no scale.

A damage ability can go up or down five, ten, fifteen percent and remain the same ability. A slow can move from thirty percent to twenty-five and players still understand what it is. Reflect has only two states: the projectile is caught and sent back, or it passes through. No buffer zone. No middle setting. No way to say the ability reflects a little less.

That turns every balancing effort into a negotiation with no answer. Shorten the window and the ability becomes useless at high skill, because good players will cast into the gap. Lengthen it and it becomes a demolition tool at low skill, where players cast on reflex and get punished for their own mistake. No equilibrium point exists between those two ends.

Riot admits they tried. Per the official notice, the balance team went through multiple rounds of adjustments aiming to preserve the reflect mechanic. They do not say how many rounds, what those rounds changed, or publish any metric proving those adjustments failed. We know the outcome: they gave up.

The two examples they chose to illustrate the surrender are notable. First, Mel reflecting Seraphine's ultimate and charming the entire enemy team with their own cooldown. Second, Mel reflecting Jhin's fourth shot and turning that critical bullet into Jhin's own execution.

Both are beautiful moments. Both are extremely rare moments.

And this is where I have to stop, because across the entire announcement there is not one line stating how often such situations occur in an average game. No win rate, no pick rate, no ban rate, no successful reflects per hundred teamfights. We have a feeling about a frustrating ability, and we have two clips representing that feeling.

In five years working with match data, I learned one uncomfortable thing: human frequency intuition is terrible. An event with strong emotional salience is remembered many times longer than an event that repeats steadily but dully. That is why every report I write starts with a single question: how often does this event happen, and in what percentage of total sequences?

Nobody answered that question in Mel's case.

What actually changes in patch 26.22

Walk through each change, because the details matter more than the summary.

W is the centre of the rework. The reflect mechanic is removed entirely. The new field still destroys incoming projectiles, but instead of firing them back it grants Mel and nearby allies a shield plus movement speed. In other words, W moves from punishment tool to protection tool. Same button, entirely different function.

The passive is pushed in another direction. Early-game damage and duration are reduced, but health and armour scaling per level are increased. This is a very clear signal, and I will return to it, because armour scaling on a mage passive is a rare design choice with specific implications for how Riot imagines Mel's position in a fight.

Q is redistributed. Front-loaded damage drops, damage from subsequent hits rises. This is the classic move to lower short-trade power in lane while raising the value of extended fights where Mel has time to stand and accumulate damage.

E is the subtlest change and the one that says the most. Root duration is reduced at close range and increased at maximum range. This is a mechanic that pays for distance. Stand close and use E to save yourself, and you are taxed. Stand far and use E to open a fight or lock a target, and you are rewarded.

R is cooled down. Cooldown up, per-level scaling down. The ultimate is no longer something you can spend as an engagement tool whenever an opportunity appears.

Read the whole package together and a picture emerges. Mel loses instant punishment, loses short-trade rhythm, loses the cheap engage ultimate. In exchange she gains survival against dives, gains damage over time, gains an ally-protection spell, and gains an explicit reward for standing far back.

Four signals, one direction: Mel is being pushed off mid lane

When armour scaling appears on a mage's passive, that is a statement about surviving dives. When W grants shield and movement speed to nearby allies, that is a statement about support function. When E rewards max range and taxes close range, that is a statement about standing behind the formation. When R is cooled down, that is a statement that her value lies in sustaining fights, not opening them.

These four signals do not contradict. They resonate.

Mel's W rework is an identity migration, not a numerical nerf. Riot is moving her from the mid-lane rhythm-breaking mage class toward the utility mage class with genuine mid-support flex. This is a change of category, not a change of degree.

In professional play, flexibility has standalone value. A champion who can appear in two roles forces the opponent to wait, to hesitate, to hold the final ban for an undetermined possibility. Those delayed bans and picks carry real tactical value, measurable in the ban turns the opponent is forced to spend.

The paradox: Mel may lose mid-lane power while gaining value in the professional pool. This is the kind of change that makes live-server statistics and professional statistics move in opposite directions, and I have seen it many times in previous patches.

But one cohort will not be able to follow.

Mel's old competitive value rested on a reactive read skill: deciding when to hold W, when to release it, which ultimate to wait for, which bullet to predict. That is a hard skill, requiring hundreds of games to accumulate, and the reward for that skill was moments like the Seraphine clip.

Removing the reflect mechanic removes that skill.

This is the harshest class of balance change a player can face. A numerical nerf can be endured, because the skill remains, only the outcome is lower. A skill-obsolescence event is different: the entire investment becomes zero. Dedicated Mel players are not being asked to adapt; they are being asked to relearn a different champion from scratch.

I have written before that a skill-obsolescence event does not create worse players, it creates players who lost the thing they were good at. And in this industry, we rarely count those people in any table.

Contrarian angle: the frustration Riot removes is not the biggest frustration

Riot states the goal of the rework is to limit situations that are too frustrating when playing against Mel. I agree with the goal. But check how much of it is achieved.

The frustration of playing against Mel has two layers.

The first is the spectacular layer: my spell is fired back and kills my own teammate. This layer generates strong emotional response, generates clips, generates posts, generates media pressure. The rework deletes this layer entirely. From patch 26.22, your spell will never return to kill your ally.

The second is the structural layer: my spell disappears and does nothing. The new W still destroys incoming projectiles. So if you are a long-range mage and throw an orb into Mel as she activates W, that orb is still swallowed. No damage. No effect. Nothing happens.

The rework removes the spectacular frustration and preserves the structural frustration, because structural frustration cannot be removed without removing the ability itself. That is a perfectly reasonable design trade. But it means Riot's claim is only half true, and the other half will never appear in the official notice.

One more thing worth noting. Riot says W is Mel's signature trait. They also say the rework will make Mel more balanced while still preserving her unique identity. Those two sentences sit next to each other in the same announcement and they collide. If the reflect W is the identity, and the reflect W is deleted, then which identity is preserved?

The plausible answer: the identity preserved is the projectile destruction. That is the part left after the spectacular shell is peeled off. But a pure projectile-destruction ability is no longer Mel; it is an anti-artillery tool, and this game already has many such tools.

I want to be clear about what I am not claiming. I am not saying Riot made the wrong call. I do not have the data to say that, and by my own rule I do not conclude while the sample is unripe. I am saying the decision was presented with two clips and no frequency, and that is a decision method with systemic risk: it over-corrects phenomena that are rare but striking.

I have seen the same pattern in football. In 2026, after the World Cup group stage, I collected data from 48 matches and found Croatia ran an average of 116.2 kilometres per game, second-highest in the tournament, while their average expected goals was only 1.08. American media wrote that they were old and slow. Both sides were describing the same team with two different data sets. Both sides felt correct.

The lesson I took from that, and apply here: when a decision is justified by beautiful moments rather than by frequency, the people involved are managing emotion, not managing a system. That can be right commercially and wrong competitively.

Data is never in a hurry; it waits until you are sober enough to ask the right question. And the right question here is simple: in an average game, how many reflect plays change the outcome? Nobody answered.

Patch 26.22 and the calendar question

There is one variable this announcement does not address, and it matters more than any damage figure: timing.

Riot ships patches on roughly a two-week cycle. Patch 26.22 falls in the late-season window, most likely after the World Championship has locked its competitive patch and before the following spring split. If so, competitive disruption is deferred and pro teams get a full preseason to adapt.

That is the best-case scenario for everyone involved.

Riot's standard practice is to freeze a competitive patch for the duration of a major international event. A rework-level change announced roughly three weeks before live deployment would, under normal cadence, be excluded from any event whose patch is already locked. This is the correct default assumption absent contrary information, and I hold it.

But there is a structural consequence worth teams thinking about in advance.

A champion whose power migrates to late-game utility becomes more draftable in longer series. In best-of-three and best-of-five, teams with adaptation capacity and deeper draft trees benefit from such champions because they have time to adjust after game one. Conversely, that champion loses value in single-game formats, where surprise outweighs depth.

Mel post-rework, if she genuinely becomes a late-game utility pick, follows that pattern. She becomes more useful in long series and less useful as a surprise pick in an opener.

This is structural inference, not a fact from the article. I say so explicitly because I do not want readers to confuse inference with data.

What to actually watch in the first two weeks of patch 26.22

I have no PBE data. I have no before-and-after win rate. I have no pick-ban figures from any region. Anyone claiming right now that Mel will be stronger or weaker is guessing, even when they speak in a confident voice.

But there are three things I will watch, and none of them require big data.

First, Mel's support pick rate in week one. If it overtakes her mid-lane pick rate, the identity-migration hypothesis is confirmed by player behaviour, and that is stronger evidence than any design statement.

Second, Mel's win rate against projectile-dependent champions. If it does not fall meaningfully despite losing reflect, then the structural frustration I described above is genuinely carrying weight, and Riot will have to return to this problem again.

Third, how many times the new W produces a fight-saving play for an ally in teamfights. This is hard to measure but measurable, and it is the most honest gauge of whether the new ability actually replaces the old one or merely fills the slot.

Every match is a confession; my job is to read between the lines of code. The Mel rework is a fairly candid confession: Riot admits they tried several times and failed, and chose demolition over continued patching. That is a decision with backbone.

But it is also a second, quieter confession: their balance system still runs on frequency intuition, and human frequency intuition is terrible.

Closing

In esports, I hear the echo of football before the data era: big decisions made on the moments most remembered rather than the events most repeated. Mel is the cleanest case study I have encountered. An ability deleted because of two beautiful clips, and nobody asked what percentage of a season's teamfights those two clips represent.

I do not believe in luck, but I believe in the probability of the shots that were forgotten. In this case, the forgotten shot is the frequency question.

If you are a Mel player, relearn the champion from scratch; do not try to convert the old skill. If you are a Mel opponent, do not celebrate: your orb will still be swallowed, it just will not come back to kill your teammate. And if you are a balance designer, publish the frequency before you publish the clip. Skills can be deleted. Trust in the process is far harder to rebuild.