How Our AI Scoring Works

What actually happens between typing two names and getting a verdict

Every comparison on APEX Versus follows the same real process, and this page explains it in full — not a marketing summary, the actual mechanism, including where it can go wrong.

Step 1 — Category Detection

When you type two names, we don't ask you to pick a category. A single AI call classifies the pairing into one of 16 categories, based on what the two entities actually are — a real person, a fictional character (and which medium it's from), a company, a country, an animal, a mythological figure, and so on.

This matters because a generic scoring system doesn't work well across wildly different kinds of things. Comparing two cricketers on "batting" and "bowling" makes sense; comparing two countries on the same scale doesn't. So the category the AI detects determines which of 16 dimension sets gets used for that specific comparison.

Step 2 — Category-Specific Scoring

Once the category is set, every entity gets scored 0-100 on each of that category's 7 dimensions. The full list, so you know exactly what's being measured for any comparison you run:

CategoryDimensions
General (real people)Power, Influence, Wealth, Intelligence, Legacy, Popularity, Potential
Anime CharactersPower, Speed, Durability, Technique, Battle IQ, Willpower, Potential
Movie/TV CharactersCombat Skill, Intelligence, Charisma, Resourcefulness, Moral Complexity, Iconic Status, Influence
Novel/Manhwa CharactersPower, Technique, Intelligence, Willpower, Growth Potential, Uniqueness, Popularity
Video Game CharactersPower, Speed, Durability, Arsenal, Combat Skill, Tactical IQ, Legacy
Mythological FiguresPower, Wisdom, Influence, Durability, Magical Ability, Legacy, Fearsomeness
AnimalsStrength, Speed, Size, Durability, Combat Skill, Aggression, Survival Instinct
CricketersBatting, Bowling, Fielding, Consistency, Match Impact, Legacy, Versatility
Athletes (non-cricket)Skill, Athleticism, Achievements, Consistency, Clutch Factor, Legacy, Versatility
CompaniesMarket Share, Innovation, Financial Strength, Brand Value, Growth, Culture, Influence
CountriesEconomy, Military, Influence, Innovation, Quality of Life, Culture, Growth
ElectronicsPerformance, Value, Build Quality, Innovation, Reliability, Popularity, Future-Proofing
CarsPerformance, Comfort, Value, Reliability, Design, Safety, Innovation
Movies/ShowsStorytelling, Acting, Cinematography, Impact, Entertainment, Critical Acclaim, Rewatchability
FoodTaste, Nutrition, Popularity, Affordability, Cultural Impact, Availability, Healthiness
Medical ProductsEfficacy, Safety, Affordability, Accessibility, Innovation, Trustworthiness, Regulatory Approval

Step 3 — The Verdict

01
Each dimension score is generated by the AI reasoning about real, known facts and characteristics of the entity — canonical feats for a fictional character, real statistics and track record for an athlete or company, and so on.
02
The 7 dimension scores average into a single overall Apex Score out of 10, shown for both entities.
03
The higher Apex Score wins — but the AI's own stated winner is what's trusted for the headline verdict and summary, not a raw recompute of the average, so a genuinely close call reads as a close call rather than a forced tie-break.
04
The full result — scores, winner, summary, and (for fictional characters) a power tier — is cached, so the exact same pairing gives the exact same answer every time it's looked up afterward, rather than a new roll each visit.

A Worked Example

Take Naruto Uzumaki vs Son Goku. Both are fictional characters primarily known from anime, so the category is Anime Characters, and both get scored on Power, Speed, Durability, Technique, Battle IQ, Willpower, and Potential — not on dimensions like Wealth or Market Share, which wouldn't mean anything here. The AI reasons from each character's canonical feats within their own series, scores each dimension 0-100, and the resulting averages produce the overall verdict. Compare the same two names on the site right now and you'll see exactly this breakdown.

Where This Has Real Limits

We'd rather say this plainly than pretend the method is more precise than it is:

⚡ Try a Real Comparison