How HEXORA calculates what it publishes

1. What HEXORA is allowed to claim

Three principles govern everything below, and they are the reason most of the decisions on this page look conservative.

• Fact, derivation, presentation. What happened is stored as it happened. Everything published is a derivation of those facts, and the derivation is written down. Nothing is invented at the presentation layer to fill a gap: when the evidence does not support a number, HEXORA shows nothing rather than a plausible one.
• The same engine for every game and every market. Scoring, aggregation and territory do not branch per game. The only thing each game decides is which of its events are competitively eligible, and it declares that in one place. A new game does not need a new formula.
• Versioned and frozen. Every published generation carries the methodology version and the exact parameters that produced it. Changing a coefficient raises the version: a historical score has to be able to say which formula produced it. Nothing is ever rewritten retroactively.

2. Season ranking

The Season ranking answers one question: who has performed best during this Season. It is a sum of the best results of the period, so it rewards competing well and competing often — that is what a seasonal circuit table is for.

Which events enter

An event has to pass every gate, not just one. Failing any of them removes the event from the ranking, never the player from the history.

• Competitive eligibility. Decided by each game's canonical methodology from the event's own format and category, never from its title alone or from an editorial badge. Casual, limited or non-competitive formats are played and recorded, but they do not score.
• Minimum field. Below the published threshold an event does not score. A four-player table is a real afternoon of play and a poor measurement.
• Complete and final results. A partial standing is not a result. Provisional versions score only when the ruleset in force says so.
• A proven finishing position. Scoring is articulated around the finishing position, so a participation without a demonstrable position does not score. It is not treated as a last place: it is treated as unknown.

How an event scores

event = ( points(finishing position)
          + tournament points x weight
          + wins x W + draws x D + losses x L ) x size multiplier

season = sum of your best N scoring events

Ties are broken by score, then by the best finishing position achieved — never by how recently someone played or by any hidden criterion.

The actual numbers — points per position, weights, multipliers, minimum field and how many events count — belong to the ruleset published for each game and Season, and are shown on the ranking itself. They are not written here because a methodology page that copies live values is a page that goes stale without anyone noticing.

3. Career Mode

What it measures: demonstrated competitive quality over time. What it does not measure: how often someone plays.

Career is not a sum. A sum rewards availability: whoever plays twice as much accumulates twice the chances to score at the same level. Career is a bounded average over your best performances, which produces the property that defines it — adding an event worse than the ones already counted never raises the score. Playing more gives you more chances to improve; it never gives you points by itself.

Methodology in force: career-v1.1.0

AWR = (W - byes + 0,5xD + 6x0,5) / (W - byes + L + D + 6)
S   = 1 + (strength - 0,5)/0,5 x span              S: 0.85 .. 1.2
q   = 0,5 + (AWR - 0,5) x T x S                    T = 1 | 2
q'  = 0,5 + (q - 0,5) x 0.5^(seasons ago)
k   = min( n , max( 10 , ceil(n x 0,50) ) )
score = 100 x ( 0,5 + sum(q' - 0,5) / max(weights, 10) x 1.4 )

Adjusted win rate

Corrected for sample size: every career starts with 6 imaginary neutral matches, so 3-0 does not claim what 30-10 claims. A draw is half a point.

Event tier

A featured competitive event weighs about double an ordinary one. The tier comes from each game's canonical methodology, never from the event title or an editorial badge.

Strength of opposition

The same result is worth more against stronger opposition, bounded between x0.85 and x1.2. Strength is always measured with information from BEFORE the event: an event cannot retroactively prove that its own field was strong, and a rival climbing the ranking today does not rewrite a past career.

Maturity

A career starts with 10 imaginary average performances; each real counted event displaces one. Until they are all displaced the score stays close to neutral, so two brilliant afternoons cannot outrank a long, excellent career.

Your best 50%

Below maturity every event counts, so nothing can be cherry-picked. Above it, Career keeps the top 50% of your performances — the same proportion for everyone. A fixed cap ("your best 12") would not be neutral: picking 12 out of 40 is worth more than picking 12 out of 12 at exactly the same real level.

Byes

A bye is not a win: there was no opponent. It is discounted from performance and it never raises the strength of opposition.

Final position

Not part of the Career Score. Matches reconstruct wins and losses but not official tiebreaks, and depending on position would leave recovered events incomparable with the ones published with a full standing.

Consistency

Shown as a statistic, with weight 0 in the score. Measured one way it simply repeated performance; measured the other it rewarded being uniformly mediocre.

Scale

0-100 with a fixed anchor: 50 is neutral performance and around 80 a very strong career. It is not normalised against the current best player, so nobody's history is rewritten when a phenomenon appears.

4. Age of the result

Each performance is weighted by the distance, measured in Seasons, between the Season it was played in and the current one. Every Season that passes, it is worth half as much.

Why

Without decay, a player whose best results were four Seasons ago lives off an annuity: their score never falls and they can barely drop out of the top, whether they compete or not. Career would stop measuring a living trajectory and start measuring a historical peak.

The precision that avoids the perverse effect

The multiplier is applied BEFORE choosing your best performances, never after: the selection ranks the already-weighted value. The consequence is deliberate and explicit — if an event from an earlier Season is still worth more than one from the current Season after its multiplier, the older one is the one that counts.

Without that precision the rule would punish twice — a worse multiplier and, on top of it, exclusion — and would throw out performances that are still better than anything recent. A great past performance is worth less over time, but it does not stop being worth something. The decay is symmetric for the same reason: an old bad result also hurts less.

Nothing else changes: the best-half rule, the maturity padding, the tiebreaks and the scale are untouched. And the quality of each event is not rewritten — what ages is how much it weighs today, not what happened that day.

5. How this was validated

The rules above were not chosen because they sounded fair. Each one was measured against the circuit's real careers, and several plausible alternatives were discarded because the measurement said so.

• The previous Career added up each Season's score, and that sum rewarded volume: its correlation with the number of events was around 0.9, far above its correlation with performance. Under the current rule that inverts — performance leads and volume follows far behind.
• A fixed cap such as "your best 12 events" still rewarded volume invisibly: picking 12 out of many versus 12 out of 12 was worth 9.24 points out of 100 at equal real level — more than the distance between the median and the 90th percentile. The proportional rule reduced that to 0.18.
• A confidence correction applied as a multiplier produced an absurd result: adding a defeat could RAISE the score, because the bad result did not enter the average but did increase confidence. It was replaced by additive neutral padding, where an event can only move the score through the numerator.
• With 6 starting performances, careers that were too short reached the top of the table. With 10 the maturity proved defensible and no short sample survived in the top 20.
• Each statement is tested by trying to break it with tens of thousands of randomly generated careers, not by checking one example. Three real defects survived a suite that was entirely green and fell at the first search for counterexamples.
• No gaming strategy pays: farming small events, avoiding strong opponents, playing only featured events, skipping tournaments to protect an average or accumulating mediocre samples all lose points.

6. Product decisions

Not everything that could be computed should be published. These are the decisions that explain what you will not find on HEXORA, and why.

• No public rating. There is no canonical, universal rating across trading card games, and pairing coverage is neither sufficient nor homogeneous across markets. Publishing a rating on top of that would be an opinion dressed as a measurement. If one ever exists, it will require its own written methodology, its own backtest and its own approval.
• Participation comes only from a published standing. If an organiser never publishes the standing, HEXORA has no proof of who played, and the event does not become participation. Individual matches verify a record — they never create one. It is a limitation we prefer to state rather than fill in with estimates.
• Played is not the same as scorable. A profile separates what you played, what was eligible and what was counted, and it shows the three numbers. An event that does not score is still part of your history and is not deleted — hiding it would make a profile look smaller than the player's real activity.
• A position always states its population. A bare "5th" means nothing. The denominator is the number of classified players in the standing in force, computed from the result itself, and the positions shown in a table are numbered over the published population, so a list never has gaps.
• A territory is published only with enough activity. A national scope needs more than 50 computable events in the Season to be shown. Below that the sample describes the calendar more than the players. Nothing is deleted or left uncalculated: it simply is not published, and it appears on its own when the activity arrives.
• Three territorial levels, chosen by scale. Country, first subdivision and second subdivision — no fourth level. For each country the two official layers whose scale matches the reference pair are chosen, not the ones that happen to share a name, because picking by name is how you end up with one level of 50,000 inhabitants next to another of 900,000. The interface always says Territory · Region · Province, in every country.
• One profile per competitor and game. Identity is resolved per game: the same person competing in two games has one competitive history in each, because their results are not comparable across games. A player can claim their own profile and ask for corrections, and a claim is reviewed before it takes effect.
• Nothing is rewritten backwards. The signals of an event are frozen at the moment it is played: its result, its tier, the strength of its field and its eligibility. A later change in a rival's ranking, in a territory or in the calendar does not alter what an event was worth the day it was played — only how much that day weighs today.

7. Change log

• career-v1.1.0 — Age of the result: each performance is weighted by its distance in Seasons to the current one (x0.5 per Season), applied before selecting your best performances. Everything else in v1.0.0 is untouched.
• career-v1.0.0 — First frozen Career methodology: adjusted win rate, event tier, strength of opposition, maturity padding and the best-half rule, replacing a previous Career that was a sum of Season scores and therefore rewarded volume twice.

Every published generation states the version that produced it, so a score and its formula always travel together.

Who we are →Report an error in a result →