TOOL NOTES

Implied probability is a price, not a forecast

One divided by the price is a translation with the margin inside, not the market's prediction. Remove the margin before treating it as a probability.

One divided by the decimal price: that is the implied probability, and the number is so clean it invites a reading it does not deserve. A 2.00 price implies 50%, and it is tempting to read that as "the market says this is a coin flip". The market said no such thing. It posted a price, and the probability is the shape of that price, not a forecast of the match.

What the number actually is

Implied probability is a translation. The odds converter computes it as 1 divided by the decimal price, and its own framing is the careful one: it is the price restated in probability units, margin included, checked 2026-11-01.

That last clause carries the weight. A two-way market at 1.91 on both sides implies 52.4% for each, and the two sides sum to 104.7%. Two forecasts cannot sum past 100%. Two prices can and do, because prices are terms offered for trade, and the excess over 100% is the margin, the cost of the trade built into the numbers.

A price answers a different question

A forecast asks: how likely is this outcome. A price asks: at what terms does a book stand ready to trade. The two are linked but not equal, because the price must leave room for the margin, and it may also lean toward where the book expects money to arrive.

The fair version of the number, the margin removed across the complete market, is closer to a forecast, and even that is the market's view at a moment, not a guarantee. The no-vig calculator says it in one sentence on its own page: the model does not establish true probabilities, checked 2026-11-01.

Why the difference matters in code

Pipelines that treat implied probability as a forecast inherit the margin as bias. Sum a thousand matches of raw implied probabilities and your expected totals run hot by exactly the average margin. Models trained or calibrated against raw implied numbers learn the book's take as if it were signal.

The clean habit is one extra step: remove the margin across the whole market first, then treat the result as the probability-like quantity. Raw implied for prices, fair for thinking, and never the first dressed as the second.

The reading rule that keeps you honest

Ask of every probability-shaped number where it came from. One divided by a listed price: a translation, margin included. One divided by a fair price after removal: the market's view, margin excluded. A model's own output: an estimate, yours to defend.

Implied probability is the right number for comparing prices and the wrong number for predicting matches. Keep those two jobs apart and the number stays exactly as useful as it looks.