MayhemLab
PATCH 16.22 · Nov 18, 2026
Trust

Methodology & Data Sources

What each number means, where it comes from, and what it cannot tell you.

What this site does and does not publish

This site publishes pick rates, augment adoption, curated combos, and item metadata for ARAM: Mayhem. It does not publish win rates, tier lists derived from win rates, or per-champion augment win rates.

Why win rate is missing: two separate walls

This is the question the site gets most, so it is worth answering properly rather than with a one-liner.

The API has the fields, but not the matches

Riot's public match API does carry augment data: a participant record has playerAugment1 through playerAugment6. In principle, one match record contains both who took which augment and which team won, which is everything needed to compute a win rate.

In practice, ARAM: Mayhem matches are not served. Requesting one returns 403 Forbidden, and Riot has confirmed this is intended behaviour rather than a bug. The fields exist; the match records that would fill them do not reach third-party developers. So the compute-it-yourself route fails at step one — there is no raw material to aggregate.

And even with the data, publishing is prohibited

Riot's developer policy states that products cannot display win rates for augments or Arena mode items, and that this applies to all websites, applications and overlays. That rule governs the display, not the provenance — so collecting the same numbers by other means and publishing them would still breach it.

Both walls have to be cleared, and neither is. Rather than estimate around them, the site leaves the metric out and shows what can be published honestly.

A note on scope: the policy text names augment and Arena item win rates specifically. It does not name champion win rates for this mode. For champion win rate the barrier is availability (the same 403), not an explicit prohibition. The site omits it for the same practical reason: no usable source.

How other sites get win rates, and why each route is closed here

Sites that do show Mayhem augment win rates are not doing magic. There are five routes, and it is worth naming them so this site's choice does not look like a limitation.

1. Tencent's China server statistics

This is the only route that reaches millions of matches. The China server publishes far more aggregate data than Riot's international servers, and several sites publish 17–21 million-match samples built from it.

It is not usable here, for two reasons. First, it is China-region data: presenting it to a North American or European audience would repeat exactly the mistake this site is built to avoid. Second, batch republication has never been explicitly licensed, and its relationship to Riot's augment rule is unclear.

2. Client-side upload from players

A companion app reads finished matches from the player's own League client and uploads them. This is how the English-language trackers work. Samples are in the tens of thousands rather than millions, growth depends on installs, and contributing is opt-in.

It is closed here because publishing augment win rates is the prohibited act, regardless of how the data was collected — and because this project has no client by design.

3. Reselling China-server data as an API

A few services wrap route 1 and offer it as a paid or keyed API. Same origin, one more layer of terms risk.

4. Substituting a different mode's numbers

Regular ARAM has no augments, and Arena is a round-based 2v2v2v2 with a different augment pool. Sites that borrow either one are publishing numbers that do not describe this mode. This is the most common shortcut and the one that makes tier lists quietly wrong.

5. A one-off arrangement with Riot

At least one operator has a private agreement with the Game Modes team granting access to all Mayhem matches, on the condition that nothing Riot does not want public gets shown. Not available to anyone else, and not a product path.

Where that leaves this site

What "client-side upload" actually means

Because route 2 is the one people ask about, here is the mechanism in plain terms. It is not a leak and it is not scraping — it is a small program sitting on the player's own PC.

  1. You install a companion app — a normal Windows executable, a few tens of megabytes. It is usually open source, and it asks for no Riot password and no account login.
  2. It finds your local League client. The client exposes a local-only interface on your own machine (the same one the client itself uses for its own screens). The app connects to 127.0.0.1 on your computer — traffic never leaves the machine to reach Riot.
  3. When a match ends, it reads that match's result file — the ten players, champions, items, augments, and the win/loss result. All of it already exists locally, which is why your own client can show you the post-game screen.
  4. If you opt in, it uploads a stripped payload to the site's server: the match, no names, no Riot IDs, no account ids. The destination database often has no columns to store them at all.
  5. The site aggregates thousands of such uploads into the sample behind its percentages.

So the pipeline is: your own client already knows the result → the app copies it locally → you choose to send it → the site aggregates. Nothing here requires Riot to publish anything.

Two honest caveats. First, the sample is only as big as the number of people running the app, which is why these sites report tens of thousands of matches rather than millions, and why their own guides warn about small-sample noise. Second, and more importantly for this project, Riot's rule governs the display of augment win rates, not the collection method — so a site can collect this way and still be publishing something the policy prohibits.

Why this project does not do it

Two reasons, and the second is decisive. It would mean shipping a desktop client, which this project explicitly excludes. And even with the data, publishing augment win rates is the prohibited act — collecting differently does not change that.

Where that leaves this site

Route 1 is another region's data. Route 2 is the prohibited act plus a client this project will not ship. Route 3 inherits route 1. Route 4 is fabrication. Route 5 is not obtainable. So the site publishes what can be published honestly: pick rates, adoption, combos, and item metadata, each labelled with its patch, date and sample.

If you need a win-rate tier list, this site is not the right source. If you want to know what players actually choose, that is exactly what the tables here measure.

Metric definitions

Champion pick rate

The share of all player slots a champion occupies in ARAM: Mayhem games. Because ten players enter each game, the pick rates across all champions sum to roughly 1,000%. A champion at 10.00% appears in about one of every ten player slots.

Augment pick rate

The share of times an augment is taken when it is offered. This is not a per-game appearance rate. An augment at 60% pick rate does not appear in 60% of games — it means that when the game showed it, players took it 60% of the time. Rarity gates how often an augment is offered at all, so cross-rarity comparisons are weak.

Tier (T0–T4)

A percentile of pick rate, not of win rate. T0 is the top 15% by pick rate, T1 the next 20%, T2 the middle 30%, T3 the next 20%, T4 the bottom 15%. A high tier means popular, not strong. The label is deliberately kept as "tier" to match how players talk, but every table that shows it carries the same caveat.

Patch change

The difference in pick rate between the current snapshot and the previous one, in percentage points. A value of +0.50pp means the pick rate rose half a point.

Combos

Champion and augment pairings that appear in the source dataset, sorted into editorial types. These types are the source's judgement, not a measurement this site performs.

Data sources

  • Pick rates, augment adoption, combos, item metadata: arammayhem.com, published under CC BY 4.0. Attribution appears in the footer of every page.
  • Champion and item names, titles, stats, icons: Riot Data Dragon, version 16.23.1.
  • Augment names, rarities, icons: CommunityDragon.
  • Augment effect summaries: written for this site, based on in-game behaviour. Not copied from a data provider.

Sample and limitations

  • It is a limited sample. The source describes its data as a limited match sample and explicitly says it is not worldwide. Small gaps between champions are usually noise.
  • No regional split. There is no reliable region dimension in the data, so the site does not offer a region selector and does not claim global coverage.
  • No item tiers. Item performance data is not published by the source, so the item pages show gold, category and stats only.
  • No Arena module. Arena augments and items fall under the same policy restriction, and no compliant data source exists for them.
  • One patch at a time. The site builds from the current snapshot. Historical pages are not backfilled, which is why only the current patch has a page.
  • Augment summaries are editorial. They describe what an augment does in plain English. Numbers in tooltips shift between patches, so exact values are omitted on purpose.

Update and correction process

The pipeline fetches sources, validates them, rebuilds the dataset and prerenders every page. If a source fails validation the build stops and the previous build stays live — a partial dataset is never published. Each build records its patch, data date and build id in /revision.json.

If a figure looks wrong, send a note through the feedback page with the page URL and the value you expect. Corrections are applied at the source layer so they survive the next rebuild.

How to read this site honestly

  1. Check the patch and data date in the header before quoting anything.
  2. Treat pick rate as popularity. Ask why something is popular before assuming it is good.
  3. Compare champions within a role and against their own previous patch, not across the whole table.
  4. Ignore differences under about one percentage point on a limited sample.
  5. Use combos as hypotheses to test in your own games, not as instructions.