A model that sounds smart and a model that is right are two different things. MMA model validation is how you tell them apart before money rides on a number. You will learn the Brier score, a simple way to grade probabilities against what actually happened, and why one fight proves nothing about calibration. The lesson also covers testing in time order so the model never peeks at information it would not have had, and why fighters with few bouts make results noisy. Finally, you turn a probability into a fair price, compare it with a posted line, and see how much of an apparent edge is real once costs are counted.
Learning objectives
- Explain the relevant technique and distinguish observations from assumptions.
- Reproduce the worked example independently.
- Identify conditions when uncertainty invalidates an apparent edge.
