Projected Rankings Accuracy
Every projection is scored against the official board that lands after the card, position by position, and compared with simply assuming nothing changes.
On the 2025–26 test window (263 division boards) the projection matched 79.5% of positions exactly against 75.3% for no change, and kept 96.6% of fighters within one place of their official rank.
- Exact positions
- 79.5%
- No-change baseline
- 75.3%
- Within one place
- 96.6%
- Champion right
- 98.1%
Before scoring the test window we set ourselves a bar of 5 points over the no-change baseline. The model came in at 4.2 points, below that bar, while beating the baseline on every measure. We publish it with these numbers as they are.
By year
Boards from 2013 to 2023 fitted the rules and 2024 chose between candidate models, so only 2025 onward is an honest test.
| Year | Boards | Projection | No change | Within ±1 |
|---|---|---|---|---|
| 2013 | 89 | 74.3% | 71.7% | 93.3% |
| 2014 | 132 | 76.3% | 73.4% | 95.4% |
| 2015 | 137 | 76% | 72.1% | 95.5% |
| 2016 | 144 | 73.9% | 70.3% | 94.4% |
| 2017 | 134 | 76.6% | 70.7% | 94.4% |
| 2018 | 141 | 78.8% | 72% | 96.1% |
| 2019 | 153 | 74.2% | 69% | 94.6% |
| 2020 | 122 | 76.4% | 70.1% | 95.8% |
| 2021 | 138 | 82.1% | 77.6% | 96.8% |
| 2022 | 120 | 85.8% | 78% | 97.5% |
| 2023 | 135 | 82.9% | 79.7% | 97.9% |
| 2024 | 133 | 81.6% | 73.7% | 97.4% |
| 2025 | 144 | 79.4% | 75.4% | 96.5% |
| 2026 | 119 | 79.7% | 75.3% | 96.7% |
Method
Each completed card is applied to the last official board. Every bout is classed by the corners' positions (upset, favourite, unranked winner, title change) and moved by a table fitted on 13 years of UFC boards. Fighters who did not fight keep their order, shifting only when someone is inserted above them.