How We Rank Comps

The Question We Actually Answer

Most TFT tier lists answer one question: "what's the best comp this patch?" That's useful, but it isn't the question you're actually asking mid-game. You just crafted an emblem, or an augment handed you one, or it dropped from loot — and now you have a specific item in hand and seconds to decide what to do with it.

EmblemComp.gg starts from that question instead: "I'm holding THIS emblem — what are my best options right now?" Every ranking on this site is conditioned on the emblem you select first, then sorted by how much it actually helps. It isn't a generic comp list with an emblem filter bolted on; the emblem in hand is the starting point of the recommendation, not an afterthought.

That distinction drives everything below: which comps surface for a given emblem, how they're ordered, and why a comp that hits a new breakpoint gets a boost without being allowed to leapfrog a comp that's simply stronger.

Who Builds This, and How Fresh the Data Is

EmblemComp.gg is built and maintained by one person — a long-time TFT player — and updated every patch. There's no team and no editorial staff; there's a builder who plays the game, watches the meta move, and keeps the site current.

Composition statistics come from aggregated ranked-match data — millions of games across regions and skill levels, not a hand-picked sample. The live comp pool is generated from the latest feed each patch; the current count is shown on the tier list. That feed refreshes hourly, with a forced daily refresh at 08:00 UTC. The exact freshness of the data behind the page you're looking at is stamped at the bottom of every page (and on each emblem page) — check there for the live timestamp rather than taking "daily updates" on faith.

The Scoring Model

Every comp carries real match stats — average placement, pick rate, and game count — pulled straight from the ranked-match feed. Game count matters: a comp with a handful of games can post a misleadingly good (or bad) average placement, so low-sample comps are flagged rather than presented with the same confidence as a comp backed by thousands of games.

When you select an emblem, each comp gets a value score:

• ◆+10 if the emblem pushes the comp to a new trait breakpoint, plus a small bonus for higher-tier breakpoints (a bigger jump is worth a bit more than a small one). Thresholds a single unit can activate on its own (like a 1-unit first breakpoint) don't count as unlocks — every comp trivially reaches those, so they earn only the progress credit below.
• ◆+5 extra if the emblem activates a trait the comp wasn't running at all before (a dormant trait coming online, not just leveling one up).
• ◆+2 per unit of progress toward the next breakpoint when the trait is already active but the emblem doesn't hit a new threshold yet.
• ◆×1.15 for non-craftable emblems (drop/augment/Armory-only). They're rarer and tend to be build-defining, so they score a bit higher than an equivalent craftable option.
• ◆+8 per extra emblem when you've selected more than one and the comp gains something from every one of them. Comps that only use a subset are shown under Partial Matches with the unused emblem named, rather than ranked as if they answered the whole question — every full match is ranked above every partial, in the API response and in the list you see. A full match also requires that each emblem can sit on a distinct legal holder and actually switch its trait on: an emblem that is placeable but leaves its trait below the first breakpoint is marked inert and does not make a comp a full match.

That score decides which comps show up and roughly how they're grouped. The sort order is placement-led, not score-led: comps are ranked by average placement, adjusted by a few bounded bonuses, with score as the final tiebreaker. A weaker comp can never leapfrog a clearly stronger one just because it activates a trait — every adjustment below is small enough to reorder similar comps and too small to override the placement data.

• ◆−0.3 placement-equivalent when the emblem unlocks a new breakpoint on that comp. Both modes.
• ◆−0.45 in Emblem First only, when the board is one the emblem is actually for — it unlocks a tier, or already runs the trait past its entry tier so the emblem deepens it. This is what separates the two modes. Without it, "show me boards for my emblem" returned the global meta list with a progress chip on it.
• ◆+0.25 in Emblem First for comps under 2,000 tracked games. A thin sample moves on noise; this is a mild prior, not a filter.
• ◆The top slot only: when two comps land within 0.05 of each other after the adjustments above, the one the emblem is more at home on takes the #1 slot. Nothing else reorders. This is capped twice over — it can only move the lead toward a better fit, and never onto a board more than 0.35 average placement worse than the best-placing result, which is about one tier badge.

Each result card names what the emblem does there — build-around, splash unlock, or progress only — so the order is auditable rather than asserted. Meta First deliberately ignores emblem centrality: there the question is which strong boards can fit the emblem in at all, so placement stays the story.

Sometimes the #1 result is a board your emblem barely changes, and that is the honest answer rather than a bug. On some patches an emblem has no competitive home: every board that would activate it is half a placement worse than the strongest board that merely fits it in. Promoting the weaker board would be telling you to play something worse so the tool could look decisive. Instead the list leads with the strongest board, the card labels it progress only, and a line above the results names the highest-ranked board that is built around your emblem, with the exact placement it costs you. If no board on the current patch is built around it at all, the page says that outright.

That note appears whenever the lead is not itself a build-around board. When the build-around board is further than 3 results down, it is called out more prominently, because at that point you would have to go looking for it.

The same note appears wherever a "best comp" is named — the finder, each emblem's own page, and the Emblem Meta Index — computed once from the same ranking. A deploy check compares all three for every emblem and fails if they name a different comp or explain it differently, because a tool that gives one answer on the homepage and another on an emblem page is not one tool.

One thing the model deliberately does not do: it never adjusts a comp's average placement for how many games it has. A 40-game comp's raw average counts exactly as much as a 4,000-game comp's. Instead of hiding that behind a formula, every comp shows its game count, anything under 100 games is flagged on the card, and you can filter by minimum sample. We would rather tell you the sample is thin than quietly change the number.

The Emblem Meta Index

The homepage's Emblem Meta Index is derived entirely from the scoring model above — no new data source, no separate methodology. For each emblem, every tracked comp is run through the same emblem-value calculation that powers that emblem's own page, and two things are counted: how many comps score above zero (comps that genuinely gain from holding the emblem), and in how many of those the emblem unlocks a new trait breakpoint.

The "best" comp shown next to each emblem is the top comp by the same placement-led ordering described above — by construction it is identical to the #1 comp on that emblem's page, and our deploy checks verify that equality for every emblem before any update goes live. When an emblem has zero viable comps in the current meta, the index says so rather than padding the row — the raw match data has no opinion about emblems in hand; this view only exists because the model computes it.

What We Don't Do

• ◆No invented stats. Every placement, pick rate, and game count shown comes from the ranked-match feed — nothing is estimated or filled in when data is thin.
• ◆No pay-for-placement. Comp ranking is entirely a function of the scoring model above; nothing on this site can be bought into a higher slot.
• ◆Comp pages (/comp/<comp-name>) are indexed exactly when the comp is one of the tracked meta comps listed on the homepage — the same ranked, capped roster every count on this site comes from. Those pages are the ones in the sitemap. A board the ranking does not currently promote keeps a readable page so that links to it never die, but it is not advertised anywhere and stays out of search.

Limitations

• ◆Early in a patch, some comps only have a small number of games. Small samples move around a lot game-to-game — treat a fresh comp's numbers as provisional until the game count climbs.
• ◆Aggregated data reflects average play across every game in the sample. It won't capture every regional or rank-bracket nuance, and it can lag a genuinely new tech for a few days until enough games are played with it.
• ◆The displayed patch label can lag the live patch by a short window right after a new patch drops, before it's manually updated — the data-updated stamp on each page is the more reliable freshness signal.