GOAT Index - How the rankings work

Fighter information and records come only from the public data of Wikipedia and Wikidata. They are fetched through the APIs those sites provide; pages are never harvested automatically (scraping). Wikipedia is CC BY-SA 4.0 and Wikidata is CC0, and this site publishes an adapted version of them. Every fighter page carries a link to the source article.

The results from the seven promotions are treated as one large network rather than being split up promotion by promotion. If A beat B and B beat C, that is a clue about A and C even though they never met. Three things go into the evaluation.

1. The strength of the opponent - beating a strong opponent moves the number more
2. How the win came - finish or decision; if a decision, unanimous or split; if a finish, in which round
3. Confidence - a fighter with few bouts, or one who has not fought for a long time, carries a wider band around the estimate

Because the whole network is solved as a single answer, the result of one bout moves not only that fighter but the fighters around them. That is why a fighter who did not compete can still see their number shift slightly between last week and this week.

The following, on the other hand, are not considered at all: size, reach and height; age; stylistic match-ups; condition and injury; the weight cut; the venue and the difference between a cage and a ring; betting odds. These obviously affect fights, but they cannot be quantified consistently from public data, so they are left out on purpose.

1. Pure rating - an estimate of ability. So that a fighter with few bouts is not overrated, a conservative figure is shown, with the confidence subtracted. That is why an undefeated fighter with a short record does not immediately pass an established champion
2. Consensus rating - a measure of standing. Because the promotions are merged without breaking the order inside each one (champion, then the official rankings, then rating order), the numbers do not always fall from top to bottom
3. Movement - the difference from the rank at the previous publication. Each view (active P4P, by division, and so on) is compared against its own previous rank

Listed fighters have at least 10 career bouts, belong to one of the seven major promotions, have fought in a major promotion in at least one of their last three bouts, and fought their last major-promotion bout within 2 years. The 2-year window exists because a gap of a year or more after surgery or injury is not unusual; a long absence lowers the confidence, and the rank falls of its own accord. Current champions are exempt from these conditions. Retirement is not decided automatically: what has been confirmed is applied by hand. Retired fighters remain in the all-time rankings.

The site updates automatically once a week, in the early hours of Wednesday. An update runs in the order fetch, recalculate, check, publish; in a week where the checks do not pass, nothing is published and the previous week's data stays in place. Keeping broken data off the site comes before freshness.

The win probability is calculated from both fighters' estimates and the confidence in them. The things listed above as not considered are not considered here either. Only bouts where both fighters are listed on this site are predicted. The hit rate is published rather than hidden. The all-time hit rate is 55.4% (36 right, 29 wrong, out of 65 bouts).The sample is still small and the figure is not statistically meaningful yet. It will be updated as it grows.

In September 2026 the win probability was reduced to a single figure. Before that, the displayed value was the product of two indicators, and the confidence read stronger than it was. Only one estimate is shown now. The fighter picked does not change, so the earlier hit rate remains valid.

Stated plainly:

1. Only the seven major promotions are covered. Regional bouts enter the network as opponents but do not appear in the rankings
2. A fighter with no Wikipedia article cannot be listed
3. Records depend on what the articles say. An error in an article is reproduced here
4. Some weight classes include an inference drawn from the opponents
5. Fighters from older eras such as PRIDE tend to come out a little low
6. Updates happen once a week. A recent bout can take up to a week to appear

Please get in touch through the contact details. A source makes it certain. A correction is not made by editing Wikipedia: it goes into this site's own override table, and the next weekly update recalculates the rating along with it.