The fight page

One page per bout. The hero shows the event, both fighters and the headline number, the matchup block compares the two fighters, and four signal cards open the deeper panels: the model, the market, the crowd and the news. A written preview with the tale of the tape closes the page.

Consensus outlook

The strip at the top links back to the event and says when it is, whether this is the main event or a title fight, the weight class and the scheduled rounds.

The two big percentages are the Consensus %, one number for each fighter that combines our model, the betting market and fan sentiment. They add up to 100. The pill in the corner is the verdict: CONSENSUS when the three views agree tightly, LEAN when they agree on the favourite but loosely, MIXED when they disagree, PICK'EM when the fight sits within two points of 50/50.

Under the bar, Most likely result is the single most likely winner and method, with its chance and, when a finish is the likeliest ending, the most likely round. How it ends splits the chances that the fight ends by KO/TKO, Submission or Decision, whoever wins; they add up to 100. These are the same numbers Win Probabilities, Fight Flow and the Simulation Lab use, built on the headline %.

When the model has fewer than five UFC fights for either fighter, a warning appears under the names and Confidence is capped at Low. Treat every number on the page as an estimate in that case.

Signal breakdown

The badges are always visible on the bar at the bottom of the hero; expand it to see the three numbers behind the headline, each for the first-named fighter. Model is the machine-learning prediction on its own. Market is what sportsbook odds imply, turned into a win %. Fans is the share of fans backing this fighter, from FIGHT//AI votes and social media. The blend note says which signals carried more of the headline for this particular fight.

Three badges summarise the picture. Consensus (Aligned, Partial, Divergent) says whether the three signals point at the same winner. Confidence (High, Medium, Low) says how sure the headline is, from the gap between the fighters and how much fight history exists. Volatility (Low, Medium, High) says how many believable ways the fight can end; a high-volatility fight can flip on one moment even with a clear favourite.

Matchup breakdown

Both fighters side by side in a card each. Next to the name: a data badge with the number of UFC fights the model has (No Data, Low, Medium, or nothing at eight or more). Below: the professional and UFC record, Form (the last five results as W/L chips; tap one for the opponent, method and date), a style tag such as Striker, Grappler or Pressure, and age, height, reach and stance.

The radar in the middle plots five ratings on a 0 to 100 scale: Striking, Grappling, Cardio, Experience (the overall rating) and Durability (chin). Ratings are percentiles against all UFC fighters, so 90 means better than about 90% of the roster. The Tactical Snapshot line under it names who rates higher in standup, grappling and cardio; even a one-point gap counts.

Scouting brief opens the AI scouting brief for each fighter (strengths and weaknesses when no brief exists).

Signal cards

Four cards, four ways into the analysis. Quantitative Model opens the Fight Map, the Fight Flow Explorer, the model's key factors, simulated win probabilities and detailed skills. Market Signal opens odds, line movement and the market against the crowd. Social Intelligence opens crowd votes, sentiment trends and official news. Fight Intelligence compares the three signals side by side.

Each card shows the headline number for its signal. Click one and its panel opens under the cards; the open card lights up and reads Showing below. On fights that are part of full access the numbers are blurred and the cards say Full access, with an Unlock this fight button above them; the main event of every card is free for everyone. You need to be signed in to open a panel.

Key factors

The model's debrief: the inputs that drove its prediction, ranked. Each factor has a short explanation, says which fighter it favours, and shows a bar for how much of the decision it carried. The first two or three factors are usually most of the story.

Factors that also show up on the Fight Map carry a "Shows up in Fight Map" link with the terrain they belong to. Tap it to switch to the map and see where in the fight that factor matters.

Win probabilities

A thousand simulated versions of this fight, summarised. Most Likely Outcome is the single most common result, winner and method together, with the round it most often happens in; because it counts winner and method as one thing, it can name a different fighter than the headline %.

Method of Victory bars show how often the fight ends by KO/TKO, submission or decision; each bar is split cyan and orange by which fighter gets it, and tapping a bar shows the split. Round Distribution shows in which round finishes land, with DEC for fights that go the distance. Historical Win Methods below shows how each fighter's real career wins came.

Detailed skills

Every rated skill for both fighters on the 0 to 100 scale, grouped by area: striking (power, accuracy, volume, defence), grappling (takedowns, takedown defence, control, submissions) and cardio. Each row shows the gap between the two fighters; cyan and positive numbers favour the first fighter, orange and negative favour the second.

Primary Skill Edge at the top is the single biggest difference and the skill most likely to decide the fight. The Edge score in the header sums all the gaps regardless of direction, so a high number means two very different fighters and a low number means a mirror match. Edge View sorts rows by gap; Raw Scores shows the ratings themselves. TDD is takedown defence; Control is ground control.

Market positioning

What the betting market thinks. Market Implied Probability turns the odds into a win % for each fighter, with the bookmaker's built-in margin removed so the two sides add up to 100; the raw odds are shown in brackets, in American or Decimal format depending on the toggle. American odds read "bet 200 to win 100" for −200 and "bet 100 to win 150" for +150; decimal odds are the total payout per 1 staked.

Signal Comparison puts the market next to the model and sentiment for the chosen fighter. Delta vs Model is the market % minus the model %, in percentage points: plus means bookmakers rate this fighter higher than our model does. Line Movement shows how the market % has moved since odds opened, with a badge for how large the move is; the Advanced view adds the full price history, the open, peak and current values, and a narrative read where "steam" means money pushing the line toward one fighter.

The Overview / Advanced toggle in the header switches between the short version and the full set of charts.

Round pricing

The over/under line, when the market offers one. Over 2.5 pays if the fight lasts past the midpoint of round three; Under pays if it ends before. The bars show the market's implied chance of each side. Compare it with the model's finish chance on the Fight Flow or Win Probabilities panels to see whether the market expects a longer or shorter fight than the data does. The section is hidden when no real over/under odds exist for the bout.

Market vs public sentiment

Where the money and the crowd disagree. Public Backing is the share of fans on each side. Market Implied is the market's win %. Market − Sentiment is the difference in percentage points. A crowd far ahead of the market usually means narrative inflation rather than information; a market far ahead of the crowd usually means informed money the public has not noticed.

The insight line under the numbers names the pattern in plain words, for example narrative inflation or sharp-money divergence.

Your forecast

Save your own pick for this fight: the winner, optionally the method and round, and how confident you are from 0 to 100. Use Model copies the model's pick into the form so you can adjust from there. The pick locks once the fight starts and is graded after it.

Graded picks appear on your profile with an accuracy figure and points. Confidence is what makes the profile useful: over time it shows whether the picks you rated 70% actually land about 70% of the time. Submit Forecast inside the Market panel is a separate, session-only ticket used for the market comparison; the saved pick is this panel.

Crowd positioning

What FIGHT//AI users and social media believe. Social Conviction is the share of fans on each side, from user votes first and, where votes are thin, from the tone of prediction talk on YouTube and Reddit. It measures opinion, not probability: 80% means four in five fans back a fighter, not that they win four in five.

Conviction Level (HIGH, MODERATE, LOW) with its number says how unified the crowd is. A high level with a lopsided split means a strong public narrative; check the Market panel to see whether the money agrees.

Sentiment trends

Part of full access. The chart plots the daily tone of posts about each fighter, where 50 is neutral, above is positive and below is negative. The two lines are independent, so they do not add up to 100 and they are not odds. Narrative Drift says how much the tone moved over the last two readings; Directional Bias names the fighter with the better tone right now.

Source Mix shows which platforms drive the conversation. Trending Topics lists the keywords fans are using. High-Signal Commentary shows the most-engaged comments, each tagged positive, negative or neutral for a fighter, with its reach in likes or upvotes.

Official news

News that could change the fight, gathered daily by AI web search: injuries, camp changes, weight-cut trouble, short-notice replacements. The badge at the top (LOW, MEDIUM or HIGH SIGNAL, or MONITORING when nothing is active) says how much the news could matter; the source count says how much coverage there is. The dominant narrative quote is the storyline most of the coverage shares.

Each report is grouped by category (camp, injury, tactics, travel and weight, other) and carries a status: verified when several sources confirm it, unconfirmed when it rests on a single report or rumour, monitoring when it is still developing.

Fight intelligence overview

All three signals side by side, and what the gaps between them mean. The tiles show the model, market and social % for the first fighter, plus the confidence. A missing signal shows a dash rather than a made-up number.

Consensus State (ALIGNED, PARTIAL, DIVERGENT) says whether the signals agree on the winner, with a sentence explaining the situation. Signal Spread measures the largest gap between any two of them. The Edge Map lists each pair with its gap in points and a HIGH, MODERATE or LOW label; the largest gap is where the analysis is worth reading. The Risk Profile turns the gaps into two reads: Market Efficiency (how close the odds are to the model; a big gap may mean mispricing) and Crowd Alignment (whether fans lean the same way as the money).