How I read these rankings

2026-08-14

Sooner or later every visitor finds a ranking that looks wrong. Right now the model rates Ciryl Gane above Islam Makhachev and Alexander Volkanovski, and it thinks very highly of Alexander Volkov. As a fight fan I raise an eyebrow at that along with the rest of you.

I want to explain why I leave it alone, because the temptation to fix it by hand is real. If I nudge one number, every number on this site becomes my opinion, and there is no way to grade an opinion. The whole point here is that the numbers come from one method, applied the same way to every fighter, with the record published for anyone to check, including the calls it gets wrong. So the rankings stay as the model computes them, even where I personally disagree with it.

There is a subtler temptation too: leave the numbers alone but adjust the model itself until Gane comes out where fans expect. The trouble is that the model is one set of rules for everyone. Bend the rules to move one fighter and every fighter moves with him, and when we test changes like that, the model as a whole usually gets worse at its actual job of calling fights. So a change has to earn its place on the full record, not on one name looking right.

Some of the eyebrow-raisers also make more sense with context on how to read the tables.

Comparisons are strongest inside a division

Fighters in the same division share opponents constantly, so results chain together and the math has plenty to hold on to. Across divisions the chains are thin: Gane and Makhachev have no meaningful opponents in common, so the model's opinion of Gane relative to Makhachev rests on far weaker evidence than its opinion of Gane relative to other heavyweights. The divisional tables on the ratings page are the load-bearing ones. The overall top list is best read as a curiosity.

Heavyweight is a small-evidence division

Careers are shorter, fights are fewer, and one punch ends more of them. The model works with less evidence there and says so: the uncertainty next to a heavyweight's rating tends to be wider than elsewhere. Volkov is also a genuinely awkward case for the eye test. He has quietly beaten a long chain of connected opponents, and the model counts every one of those wins the same whether or not the performance looked dominant. Fans discount ugly wins. The math does not.

The Gane case

Gane divides fans precisely because his wins rarely feel like statements. But look at his stat lines on our event pages. Across his UFC run he has landed roughly two significant strikes for every one he has absorbed, and he has out-landed every opponent the UFC has put in front of him, Ngannou included in a losing effort. Quiet fights can still be one-sided fights, and a record of one-sided wins over well-connected opponents is exactly the kind of record this ledger ends up respecting. Whether that should count as much as it does is a fair argument, and it is one the coming fights will settle.

What would actually change these numbers

More results, or a principled change to the method. If the model is wrong about these fighters, the coming fights will say so and the numbers will move on their own. That correction is the product working, and it needs no hand on the scale. The method will keep evolving too, and when a change reshuffles the rankings, it will be made before the fights it is graded on, and noted here. Until then, when you and I disagree with the ledger, the ledger keeps the receipts either way.