Value notes
When your model disagrees with the fair line
Your model says one thing, the margin-free market says another. What the disagreement is, which side to doubt first, and how the close settles it.
Your model puts the home side at 50%, and the margin-free market price says 60%. That gap is the entire game for a model bettor: it is either your edge or your error, and the two look identical on a spreadsheet. The work is deciding which side of the disagreement to doubt first, and the close decides it in the long run.
What the fair line is
The fair line is the market's estimate with the margin removed: every piece of money and information the market has absorbed, compressed into one probability. The margin removal guide explains why that number is the benchmark, checked 2026-10-16. When your model disagrees with it, you are not disagreeing with a price; you are disagreeing with the market's information processing.
That is not an argument to never disagree. It is the reason disagreement needs evidence, not confidence.
Doubt your model first
Check the model before the market, because the model is yours and its failure modes are known: stale inputs, a feature that broke last week, a fit to a sample that no longer exists. Then check the comparison itself: is your 50% against their 60% measured on the same market, the same moment, the same margin model? A disagreement built on mismatched inputs is a data bug, not a market view.
If everything checks out and the gap survives, you have a hypothesis. A hypothesis is not a bet; it is something you test at small stakes while the market grades it.
The close grades both
The way to know who was right is not to argue, it is to record. Take the positions where your model disagrees with the fair line, log the price you took and the fair close, and let a few hundred of them accumulate. The CLV guide explains why beating the close is the strongest early evidence of an edge, and it applies to your model exactly: if your disagreements with the market keep beating the close, your model is finding something. If they do not, the market was right and the disagreement was noise.
Track the disagreement as its own category, separate from the bets where you agree with the market, because the two sets will answer different questions.
The size of the disagreement sets the stake
A 1% disagreement is indistinguishable from modeling noise; a 10% disagreement is either a big edge or a big error, and the big-edge checklist applies before you stake anything. Disagreement size is not confidence; it is a prompt to verify more, not to bet more.
Your model against the fair line, graded by the close, tracked as its own record. That is how a disagreement stops being a feeling and starts being a result.