How the numbers are made

Results, standings, pairings and decklists come from Limitless TCG, which hosts most online Pokémon TCG Pocket tournaments. Unfinished events are re-checked until they settle. By default the site uses:

• Eventsfinished, 16+ players, with decklists
• Formatstandard (no custom ban list)
• Windowthe last 120 days
• Gamesa win counts 1, a tie ½; byes, forfeits and double losses are not games

Win rates leave out mirror matches. A deck wins exactly half of them, which would pull every popular deck toward 50%; shares still count every entry. The filters pick one of 12 views (window, all games or top cut), each worked out ahead at every data refresh. Every view counts events of 16 or more players, the one event size the site uses. A view with under 10 events or under 2,000 entries says “Small sample”, and its numbers move a lot; in a top-cut view only the entries that made the cut count.

The window can also start at one of the two latest set releases (Since C1, out Oct 29, 2026; Since B4b, out Sep 30, 2026). It then holds only the games from the day that set came out, all games or top cut. In its first 10 days such a window is too short for trends, so they read “Too new for a trend”, and it has no Rising flags (they compare the last 7 days with the last 28).

The meta moves every week, so old results fade gradually. Shares and win rates fade at different speeds. A meta share follows the current field, so an entry counts half as much after 7 days. For 14 days after a set comes out, entries from before the release count 0.3× as much again, because decks change fastest in those days. A card’s play rate and a deck’s core, flex and tech cards count lists the same way, so they follow the same field. Win rates and matchups keep a longer memory and rest on more games. A game counts half as much after 28 days.

The share’s weights were picked by predicting the next week’s field from the weeks before at 62 past dates, four after each of 11 set releases and 18 away from any release. A share counting half after 28 days was the worst predictor in every group; four days after a release only about 15% of its weight came from games since the release. The current rule did better at 61 of the 62 dates.

The trend arrows compare this week with the last 4 weeks (in a 14-day view, with the whole 14 days). A win rate compares its last 7 days with its last 28. A meta share and a card’s play rate already count the last week most, so their trends compare the number shown with the last 4 weeks. Shares trend in percent of themselves (2% to 3% is +50%); play rates and win rates trend in points (2% to 3% is +1pp). A deck’s Share trend shows its 4-week share with its own range. In the 14 days after a set release the meta share and play rates count older games less, and their trends use those same numbers.

A change is shown only when it is bigger than the noise; otherwise the trend reads Steady, with the measured change in its tooltip. A win rate whose last 7 days are worth under 20 games once older ones are discounted reads Too few recent games, because a deck played mostly weeks ago can swing 30 points on two recent games.

The Overview's “Meta over time” stacks the most played decks' shares day by day over the last 8 weeks, each as its deck page showed it on that day. Every other deck forms one hatched band on top, so each day adds up to 100%. The charts on deck and card pages also have a point per day, each the number the page showed on that day. The data tables of charts longer than a month, and the screen-reader table of “Meta over time”, list one day a week, ending with the latest.

A win rate from 40 games can easily be 10 points off; from 4,000 games it is rarely more than 2 off. So each number carries its likely range, which holds the true value 9 times out of 10. Fewer games give a wider range, and a result is colored only when its whole range is on one side of 50% (or of zero, for a card’s lift).

• Decksno tier and faded under 300 games, older games counting less
• Matchupshatched and gray under 20 games, older games counting less
• Cardsfaded when fewer than 50 decklists run them

Faded and hatched results are there for reference. They are never colored, and sorting by a win rate or an effect puts them after the rest.

Meta share

A deck’s share of tournament entries, recent weeks counting most and the last week most of all; right after a set comes out, entries from before it count less (see Recent results count more). A deck played mostly in the last few days can therefore rank above one with more entries over the whole window.

Win rate

The share of games a deck won, a tie counting half and mirror matches left out.

Adjusted for skill

A deck can look strong just because strong players pick it. This version compares each game with what the two players’ ratings predicted, so it takes out how much better or worse the deck’s pilots were than the players they faced. When it differs from the plain win rate, the difference comes from the players.

Pilot rating

The average rating of the deck’s players over the games behind its win rate, beside the average rating of the players they faced in those games. “+8 vs their opponents” is the difference between the two, the gap Adjusted for skill takes out.

Vs the meta

A plain win rate depends on who a deck happened to meet. Vs the meta averages the deck’s matchups instead, each counted as often as you would meet that opponent in today’s field (mirrors count as 50%).

Top cuts

How often a deck makes the top cut compared with what its events’ cut sizes predict. 1.0× is as expected; 2.0× is twice as often. Top-cut share is the deck’s share of the entries that made a cut; in a top-cut view it is the meta share itself.

Tiers use the range of the win rate vs the meta. A deck is S tier only when even the low end of its range is that good, so a great but uncertain number waits for more games, and a deck with a higher win rate from few games can sit below a steadier one. At the bottom, F marks a deck whose whole range sits under 50%, clearly losing to the field, and D a deck whose low end is too low for C while its range still reaches 50%.

STier SLow end ≥ 52%

ATier ALow end ≥ 50.5%

BTier BLow end ≥ 49%

CTier CLow end ≥ 47%

DTier DLow end below 47%

FTier FTop end below 50%

–No tier (too few games)No tier: worth under 300 games once older games are discounted, shown faded

Avatars: the most-played decks in each tier right now.

The precise rule. A deck’s tier reads the 90% range of its win rate vs the meta, in this order: S, A and B on the low end (S ≥ 52%, A ≥ 50.5%, B ≥ 49%), then F on the top end (F top end below 50%), then C on the low end (C ≥ 47%), and D for every other deck (D below 47%). A deck whose top end is below 50% is F before C or D, since it is clearly losing to the field. Both ends are compared as the site shows them, to one decimal (a low end shown as 49.0% is tier B; a top end shown as 50.0% is never F). The tier comes from Vs the meta, so a deck with a higher plain win rate can sit in a lower tier. Only a deck whose games are worth at least 300 once older games are discounted gets a tier. Within a tier, decks are ordered by the low end, highest first, as in the Decks table’s Tier column and on the tier list. Each meta view (the window, all games or top cuts only) has its own tiers, recomputed with every data refresh.

The tier list shows every deck with a tier, S to F, with its win rate vs the meta and its meta share; decks too new or too rarely played to rank are on the Decks page, faded.

Each matchup is one number, the games the two decks played blended with what their usual decklists predict (a model that reads the cards and types of both lists). The prediction also learns from games at past events with ban lists, NOEX and other formats, and gives each deck a small correction from how it has done. With few games the number leans on the prediction; as games add up it follows the games. The same number appears on the deck page, the “how to beat” page and the matrix, with the record behind it as wins–losses(–ties) in the window, and the two sides add up (if A beats B 60% of the time, B beats A 40%). When no prediction model is available, the prediction is each deck's results against other decks instead.

Most of the 1,446 decks in the data are one-off brews, so most pairs have few games. A matchup worth under 20 games once older games are discounted is hatched and gray; its record can show more games than that.

The Overview's “Top Metagame Matchups” draws the same numbers for the 10 most played decks (8 on phones), each circle sized by its meta share. A solid arrow points from a deck to one it beats when that matchup's whole range is above 50% and the two decks have enough recent games; it is brighter the further the range clears 50% and thicker the bigger the margin. Every deck drawn gets at least 2 arrows. A deck with fewer solid ones gets dashed arrows from its other matchups where one side wins at least 55% and is at least 80% sure to be ahead (the 90% range read as a normal curve), few recent games allowed, the surest first, until it has 2. A deck still short of 2 is left out with its arrows, and the rule runs again on the decks left. No arrow means too close to call or too few games.

Decks are the deck types Limitless assigns to each list, with reprints merged, so the same deck built from different printings counts once. Limitless names a deck after two of its Pokémon, so one deck can carry several names; names that cover the same deck are joined. A build around a different version of a Pokémon stays its own deck.

Cards are identified by what they do. Reprints and alternate arts share one page, named after the earliest printing.

Two cards with the same name but different attacks or abilities stay separate, like the two Dustox (B1) and Dustox (B4) cards. Card names anywhere on the site show the card on hover or tap.

Each deck page sorts a typical list into three groups, Core, Flex and Tech, and flags the staples:

Core

Cards in 9 of every 10 lists of the deck or more. They make it this deck.

Flex

One spot in the deck that different cards of the same name can fill, like two printings of a Basic or the four Furfrou. Lists pick one or mix them, and each option shows its share of the lists.

Tech

Optional cards, in from 1 in 20 to 9 in 10 of the deck’s lists.

Staple

A flag on a card most decks play (Professor’s Research, Poké Ball, Copycat, Cyrus). A staple can also be core.

The shares count recent lists most, the last week most of all, like the meta share. A deck with many lists in the window and few in recent weeks gets a note that its groups are a rough guide, and a version with too few recent lists cannot be told apart from the rest of the deck.

Since one deck can carry several Limitless names, Versions of this deck lists each name with its share of the deck, its win rate adjusted for skill and the cards that set it apart.

Play rate

The share of decklists that run at least one copy, weighted like the meta share. The last week counts most, and for 14 days after a set comes out lists from before it count less. Copies is the average count in those lists. The trend compares the play rate shown with the last 4 weeks.

Win rate with it

The win rate of decks running the card. It mostly reflects which decks play it, so it is never colored. Over all decks it is counted in player-games: a game where both players run the card counts twice.

Lift

How many points a deck’s win rate against the meta changes when its lists run the card versus when they don’t, with its other cards and the players’ skill held equal. A 2nd copy lift compares two copies with one. A card in almost every list, in very few lists, or in a deck with too few games has no lift, and the page says which.

Lift across decks

The card’s lift averaged over the decks that play it, each deck counted by its games with the card. It needs three or more decks.

Cards that win more

The Overview’s cards whose lift across decks stays above zero over its whole range, ranked by the low end of that range, so a big number from few games ranks below a surer one.

On the rise

The Overview’s cards whose play rate grew by more than the noise against the last 4 weeks, biggest gain in points first, each with the deck that plays it most. Both lists leave out cards in fewer than 50 lists.

A lift shows how decks with a card did; it does not prove the card caused it. A card that strong players happen to favor can still look a couple of points better than it is.

On the beat-a-deck page, each card’s number says how much running it helps against that deck. Each deck that plays the card into the matchup gives its own number (its lists with the card against its lists without it, other cards and player skill held equal), and the page shows the average over every such deck, counting each deck by its games with the card. Open a card to see each deck’s own number; the head-to-head page shows them for both sides of one matchup.

Cards are ranked by the low end of their range, because a big number from a few games (often a rarely played card) has a wide range and usually shrinks as new games come in. Two tags appear only when the whole range is above zero: Swings a losing matchup (most of its games against the deck come from decks whose matchup is below 50%) and Underused (fewer than one list in four runs it).

A deck carries one tag at a time: Breakout, else Rising, else New. New and Breakout are worked out at every data refresh on the default view’s events and read the same in every view; Rising belongs to the view.

• New: first seen in the last 14 days, however few players run it
• Breakout: first seen on or after the latest set’s release (C1, out Oct 29, 2026), with a card of that set in 50%+ of its lists since then and 6%+ of the entries since the release (the share its “Since” view shows), in 2+ events with 5+ different pilots. It shows from 48 hours after the release until 14 days after it, checked again at every refresh

Breakout is experimental. Its rule was picked on a replay of 11 past releases, where a meta deck is one holding 3%+ of the entries 3 to 6 weeks after the release. Tagged 48 hours in, 12 of 16 decks became meta decks (75%, 90% range 50%–94%); tagged 72 hours in, 15 of 18 (83%, range 64%–95%). The tag finds 15 of the 35 meta decks by 72 hours, and 4.7% of all new decks built on a set became one. While a deck is new to the latest set, its page shows its share since the release where Top cuts usually is, and the Decks page’s top tile goes to a Breakout deck for the first 14 days.

What’s moving puts the decks in four groups, then every other deck: a Breakout deck while the latest set is under 14 days old; a Rising deck; a deck worth 300+ games whose meta share is clearly above its share over the last 4 weeks; a New deck with 0.5%+ of the meta. Breakout and New decks rank by meta share, the other two by how fast their share grows. The Overview shows the first 8, the Decks page’s Trends view opens in this order, and the Decks page’s top tile takes its Breakout or Rising deck from it. For 14 days after a release the Overview names the set and its day, and for 28 days the tier list adds a New this set row: decks first seen since the release with 0.5%+ of the entries since then, too new for a tier.

A deck’s Rising score adds up four signals, each damped when the sample is small so a handful of results cannot carry it:

• Gaining: its share is growing (this week against the month), with 5+ different pilots in the last 14 days
• Strong: its win rate adjusted for skill stays above 50% at the low end of its range once a small sample is damped toward the field, so a deck whose range is just above 50% can still miss it
• Converting: it makes top cut more often than expected, at the low end once damped the same way
• New: first seen in the last 14 days

Strength and top cuts count only once a deck is gaining or new, so a strong deck with a flat share is not flagged. A deck scoring 5.75+ while under 3% share gets a Rising badge on Decks. The badge is experimental, a hint of what to watch.

Cards are scored the same way inside each deck, from how fast the card turns up in more of the deck’s lists and how often those lists make top cut. A flagged card is marked Rising on the deck page’s card usage, and the Cards page says how many decks it is rising in. The Decks page’s Trends view shows how fast every deck’s share is moving.

See the share trends →

The builder compares your cards with every recent tournament list and leans on the lists most like yours, the closest and newest counting most. A deck that mixes two decks borrows from both, and a deck that matches no known deck still gets what the lists nearest to it run.

It suggests the core cards those lists almost always run first, then flex choices (which Riolu, which Furfrou) and techs. A second card for a spot your deck already fills is offered only when the lists run both. Cards few of those lists run are offered as cuts. When your deck reads as a known deck, the suggestions also carry that deck’s lifts; a rogue deck gets no win-rate claims, because the nearest deck’s numbers are about another deck. “Analyze as” a deck compares yours with that deck’s lists only.

The card search’s Recommended order reads the same lists, so what they run comes first, each card with a short reason. Its Lift switch marks each card with its Lift in the deck your list reads as and a small bar for its range. The number is colored only where the range clears zero. The large card view shows the same Lift, and a second copy’s Lift when the deck has one.

On two weeks of later tournament lists with one to three cards hidden, a hidden card was among the first five suggestions about 95 times in 100 for lists of known decks and 67 in 100 for rogue or rare decks. For a deck made of halves of two decks, cards of both halves came back among the first five about one time in four.

The builder’s Insights tab shows each card’s chance to be in your hand at the start, worked out exactly for the deck on screen from one simple rule for the deal. One Basic Pokémon card is picked at random from all the deck’s Basic copies, and 4 more cards come from the other 19. Both players then draw a card at the start of their first turn. Fossils are Items, so they don’t count as Basics here.

Can start with it

A Basic’s chance to be among those 5 cards, either as the Basic picked (as often as its share of your Basic copies) or among the other 4.

Must start with it

A Basic’s chance to be the only Basic in those 5 cards, so it has to be your Active Pokémon. A deck’s only Basic always starts.

Turn-1

Every card’s chance of holding at least one copy at the start of your first turn, after its draw. For a card that isn’t a Basic it is 26.3% with one copy and 46.8% with two, in every deck.

The odds show once the deck is legal: 20 cards and no rule broken. The formulas are in the technical details.

Rating

Each player’s skill, from their results against the players they met, with the strength of the decks they played taken out. 1500 is an average player. Its range holds the true rating 9 times out of 10; it is wide with few games and grows with time off.

The leaderboard

Players with at least 10 events, one of them in the last 30 days, ordered by their rating minus twice its uncertainty, so a high rating from few games ranks below a slightly lower one we are surer of. Every rating shows its range.

Win rate

The share of a player’s games won in Standard events over the last 120 days, recent weeks counting most and a tie counting half. Byes, forfeits and double losses never count, so a player page’s records come from the pairings rather than the event’s standings.

Team events

Every member carries the team’s placement and record, so team events are left out of a player’s event count, win rate and most played decks. Their games still count toward the rating and its number of games.

Players are their Limitless usernames; the name shown is the latest one they used, and search finds every name. A player can ask to be left out. Their page and name are then removed, while their games still count in every deck statistic.

Signed in, you can follow decks, cards and players. “Since your last visit” lists what changed for them since you last opened that list, going back at most 30 days. Players’ new events come first, then the biggest changes: a rating, meta share, win rate against the meta or play rate that moved by more than its noise (the two ranges barely overlap), a new tier, a deck turning Rising, a card’s lift across decks starting or ceasing to clear zero, and a player entering or leaving the top 100. Each change compares the numbers the site showed then with today’s.

A paid plan is on its way for tools that plan and compute for you. Everything on the pages you can read today stays free. Pages marked Premium · , coming soon describe what a tool will do and never show its results. Custom Meta will take a field you describe (starting from this week’s meta, with any deck’s share changed) and banned cards, and give each deck’s expected win rate against it, your saved decks included. Before it goes live we test it on past tournaments that had bans. Signed in, “Tell me when it’s ready” notes that you want a tool, and its launch then shows under Since your last visit; Undo takes it back, and nobody else sees who asked.

• Tournament data: Limitless TCG and the organizers and players who report results there.
• Card data and text: Limitless Pocket card database.
• Card images: TCGdex where available, otherwise the Limitless card CDN; cached and served from our own server. On high-resolution screens, the large art at the top of deck pages is enlarged with an AI upscaler (Real-ESRGAN), which can smooth or invent fine detail.
• Metalodon is an unofficial fan project, not affiliated with or endorsed by The Pokémon Company, Nintendo, Creatures, GAME FREAK or DeNA. Pokémon names and card images are © their respective owners.

Example as of Nov 9, 2026: the most played deck, Mega Charizard Y ex Entei ex, holds 8.5% of the field over 995 entries.