Inside the Formula Replacing UFC’s Media Rankings

The Meta UFC Rankings are an Elo-style divisional rating model, built with Meta, that the UFC began phasing in to replace its media panel on June 22, 2026. A win raises a fighter’s hidden rating and a loss lowers it; beating a higher-rated opponent moves the number more than beating a weaker one; a finish adds a small bonus; recent results count more than old ones; and ratings decay after 18 months without a fight. A UFC representative told CBS Sports that winning and opponent quality account for roughly 95% of placement.

What nobody outside the UFC and Meta can do is reproduce the numbers. The formula, the raw ratings, the factor weights and the rules for draws, no contests and missed weight have not been published. The rankings can be described in broad terms; they cannot yet be audited.

Beat Strong Opponents and Stay Active: The Factors the Model Counts

The UFC announced the system on June 22, 2026. Its description says the model considers outcome probability, win type, opponent quality, fighter trajectory, recency, inactivity and weight-class factors. Divisional rankings are recalculated on Mondays following UFC events (official UFC and Meta announcement).

CBS Sports dated the rollout to “Monday, June 23, 2026,” but June 23 fell on a Tuesday; the UFC’s own announcement is dated Monday, June 22. The unnamed UFC representative quoted by CBS gave the “roughly 95%” figure for winning and opponent quality. That is an approximate description of their importance, not a published formula assigning them an exact 95% weight (CBS Sports’ explanation of the ranking model).

The logic resembles an Elo rating. Every eligible fighter has an underlying rating, even though the public sees only an ordered list. When two fighters compete, the winner gains rating value, the loser gives some up, an upset over a stronger-rated opponent produces a larger adjustment than an expected win over a weaker one, and the revised ratings determine the divisional positions.

The practical rule: beat strong opponents, do it recently and avoid extended layoffs. A finish helps, but available reporting says result and quality of opposition matter far more.

Pick the winner’s and loser’s current ranks and the result; the projected order updates beside the controls.

What Moves a Rating and What Stays Private

The model evaluates each result relative to the fighters’ existing ratings. It does not treat every win, or every loss, as equally important.

Factor	Reported effect	Not public
Fight outcome	Win raises the winner; loss lowers the loser	Exact adjustment formula
Opponent quality	Beating a stronger-rated opponent is worth more	Raw ratings and pre-fight gaps
Victory method	Finish gets a small bonus over a decision	Bonus size; treatment of finish types
Recency	Recent results carry more influence	The recency curve
Inactivity	Decay reportedly begins after 18 months	Decay rate and exceptions
Historical results	Fights older than five years lose weight; obsolete after ten	Year-by-year reduction
Weight class	Ratings are calculated per division	Catchweights and cross-division bouts

The inactivity and historical thresholds, along with the modest finish bonus, come from accounts of the model rather than a published technical specification (detailed explanation of the reported rating rules).

Opponent quality refers to the underlying rating, not the number beside a fighter’s name. The visible No. 5 and No. 6 could be separated by a fraction of the gap between No. 6 and No. 7. Without raw scores, the numbered list does not reveal those distances.

That distinction also explains movement by fighters who did not compete. Someone may fall because inactivity decay begins, an old win loses influence or nearby contenders gain rating value. A fighter can rise without fighting if the ratings above them decline.

Nothing in the disclosed methodology establishes that the model uses significant strikes, takedowns, control time, submission attempts, scorecard margins or visual dominance. Available explanations identify result-level facts only: the winner, the opponent, the date and the victory method.

Why One Win Can Jump Five Places or One

Rankings show order, not distance. That is why similar-looking results produce very different movement.

Suppose several contenders are tightly clustered in the underlying ratings. One valuable win can carry a fighter past the whole group. The model is not swapping the winner’s numbered position with the loser’s; it recalculates the ratings and sorts the division again.

Two hypothetical results make the point. Fighter A finishes a highly rated top-five opponent. Fighter B wins a narrow decision over a much lower-rated opponent. Fighter A should receive the larger adjustment because opponent strength is the main driver and the finish adds a modest bonus. No point totals can be assigned, because neither the raw ratings nor the finish bonus is public.

Now give two fighters the same rating increase in different divisions. In a tightly packed division, that gain might move someone from No. 10 to No. 6. In a division with larger gaps, it might produce a one-place rise. The visible jump depends on both the value of the result and the density of the surrounding ratings.

CBS reported that the model was designed to move a winner above the opponent they defeated. Early launch-era outputs reportedly included apparent head-to-head exceptions, so this is intended behavior rather than an infallible rule covering every output, eligibility decision or data problem (CBS Sports’ report on the model’s design).

It Is a Rules Engine, Not an AI Watching the Fight

The weekly process is better described as a fixed, Elo-style rules engine than as artificial intelligence judging fights.

Machine learning reportedly helped develop or test the system. Once the assumptions were set, the live model applied predetermined rules to structured fight data: who won, who the opponent was, when the bout occurred and whether it ended by finish or decision. It does not watch footage and decide which fighter looked more dominant (reporting on the data-driven rather than active-AI model).

The distinction matters because “AI rankings” can imply video interpretation of damage, movement or tactics. The available evidence does not support that description.

Automation removes the fresh discretion of weekly ballots, but it does not remove human judgment from the system. People selected the variables, set their importance and wrote the recency and inactivity rules. UFC descriptions of the model as objective, accurate or transparent are organizational claims, not independently proven conclusions. A formula can apply the same rules consistently while still encoding debatable assumptions about competitive merit.

What Changed From the Media Panel

The former process averaged ballots from a media panel, in place since February 2013. The Meta system applies a predetermined calculation after event results are recorded.

Question	Media panel	Meta model
Who orders fighters?	Averaged voter ballots	Automated rating model
Primary inputs	Voters’ assessments	Results, opponent quality, recency, activity, finish adjustment
Update process	New ballots after events	Automatic post-event recalculation
Consistency	Voters applied different standards	Same coded rules every time
Method visibility	Voting structure known; reasoning varied	Broad factors known; ratings and parameters private
Practical authority	Advisory to UFC decision-makers	Still advisory

The wider ranking process has three layers. The model calculates ratings and orders eligible fighters. UFC personnel reportedly manage which fighters belong in each divisional pool, although the intervention and correction rules are unpublished. UFC executives and matchmakers keep discretion over bookings and title opportunities.

The output is therefore not a mandatory title-shot ladder. A position can inform matchmaking without requiring the UFC to book bouts in strict ranking order.

The display is unchanged: each division has a champion above 15 numbered contenders, so No. 1 means the leading non-champion. Reporting also says a fighter can be eligible in two divisions with a separate rating in each; a result in one division does not automatically alter the other rating (reporting on divisional structure, eligibility and separate ratings).

During the transition, the Meta and media lists were displayed together and could disagree. A difference between them did not indicate a processing error; the two systems evaluated divisional order through different methods.

What the Public Still Cannot Verify

Calling the system mathematical is accurate. Calling it fully auditable is not.

Reported mechanics: results and opponent strength drive most rating movement; finishes receive a modest benefit over decisions; recent bouts matter more than older ones; extended inactivity causes decay; older fights gradually lose influence; the division is re-sorted after events.

Undisclosed: the complete formula, raw fighter ratings and starting values, exact factor weights, the size of the finish bonus, the recency and inactivity-decay curves, data-correction procedures, a public history of model revisions, and technical definitions for outcome probability, trajectory and weight-class sensitivity.

Edge cases also remain unresolved. Public explanations do not establish how the model handles draws, no contests, overturned results, disqualifications, missed weight, catchweight contests, replacement opponents or moves between divisions.

Without those details, readers can observe movement after a fight but cannot calculate the expected adjustment beforehand or verify that every published position matches the intended formula. Mathematical consistency is not competitive validity: a model can apply its rules correctly and still produce an order that fans, fighters or matchmakers dispute.

Frequently Asked Questions

How Do the Meta UFC Rankings Handle Missed Weight, Draws or No Contests?

The public methodology does not specify how missed weight, draws or no contests affect ratings, nor the treatment of overturned results, disqualifications or catchweight bouts.

A fighter moving up after winning while overweight would not, by itself, prove that the formula ignores missed weight. Other rating changes, eligibility decisions or movement elsewhere in the division could affect the displayed order. Until an explicit rule is published, neither a penalty nor the absence of one should be assumed.

Are the UFC Pound-for-Pound Rankings Generated by the Meta Model?

Launch reporting said no. The Meta system was built for ratings within individual divisions and did not generate a pound-for-pound list, because comparing separately calculated divisions requires another model (launch reporting on the pound-for-pound limitation).

UFC.com continued to display men’s and women’s pound-for-pound lists. Their presence does not establish that Meta generated them; they may be maintained through a separate or transitional process.

Can a Fighter Hold Meta Rankings in Two Weight Divisions?

Reportedly, yes. A fighter can be eligible in two divisions and carry a separate rating in each. Results in one division do not automatically change the other.

What remains unclear is eligibility. The UFC has not detailed every rule governing when a fighter enters, remains in or leaves a divisional pool. The model can calculate an order only after that pool has been defined.

The rule of thumb: strong recent wins drive the rankings, opponent quality matters most, activity protects a position and finishes add a smaller benefit. The model applies its rules more mechanically than a weekly ballot, but independent auditing stays impossible until the UFC or Meta publishes the ratings, parameters and edge-case rules.