How scoring works
RoyaleIQ’s score is a structural assessment built from curated card metadata. It is not a probability of winning, and the weights below are starting hypotheses that will be tuned against recorded results.
The formula
score = clamp( Σ component_score × weight − penalties, 0, 100 )
Each component is normalized to 0–100. Critical structural problems — no win condition, no air defense, no utility spell, an extreme elixir cost, no reliable tank killer — subtract a fixed penalty once each, capped at 25 points, so one weakness is never counted twice.
Components and weights
| Component | Weight |
|---|---|
Offensive pressure and win conditions Win-condition reliability, tower pressure, spell finishing damage, support for the main win condition, and how coherent the plan is. | 18% |
Defensive coverage Tank killing, air and ground defense, splash, swarm control, buildings and kiting, resets, and defensive spells — plus a check for single points of failure. | 18% |
Card synergy A weighted interaction graph: curated combos (for example Golem + Night Witch) and structural rules (tank + ranged support). Shared roles alone earn nothing. | 16% |
Matchup resilience Estimated resilience against beatdown, cycle, bait, siege, bridge spam, air, control, and graveyard decks. The weakest matchup counts extra. | 14% |
Air and ground coverage How many cards hit air, how hard, and whether splash covers air swarms. | 10% |
Elixir curve and cycle Average cost against the target for the detected archetype, cost to cycle back to the win condition, playable opening hands, and affordable answers. | 10% |
Spell utility Small and big spell coverage, swarm response, and finishing damage. Three spells suit bait and control; four rarely suits anything. | 8% |
Evolution, Hero, and ability value Each special form's ability value multiplied by how much it adds to this deck. Using a special card earns nothing by itself. | 6% |
Coverage combines independent answers as 1 − Π(1 − v): two separate answers beat one, with diminishing returns, rather than counting cards that share a label.
Confidence is separate
Confidence reflects how much the score can be trusted, not how good the deck is: completeness of card metadata, whether real matchup evidence exists, how many cards are selected, and how fresh the metadata is. Without recorded battles, matchup estimates are heuristic, so confidence usually reads Medium.
Data sources
- Card IDs, names, costs, rarities, artwork, and which cards have Evolution or Hero forms come from the official Clash Royale API.
- Strategic roles, damage bands, and the interaction graph are RoyaleIQ’s curated, versioned metadata. Cards without reviewed metadata lower confidence instead of receiving guessed roles.
- Special-slot rules (Evolution, Hero, Wild; one Champion) were confirmed against official battle logs in October 2026.
- Personal coaching uses only battles the official battle log exposes — roughly the most recent 30 — collected from the time you link your tag.
Known limitations
- No global win rates or matchup statistics are used; the official API doesn’t provide them.
- The game’s share links encode Evolutions but have no documented encoding for Hero forms. Decks with Heroes export as their base cards, with a warning.
- Linked player tags are shown as unverified: the official ownership check isn’t available to standard API keys.
- Personal results show association, not causation. Opponents, game mode, and balance changes all affect outcomes.