LZLZL/Prediction markets/Mechanics
FREESEE IT A · MechanicsCore idea

Price is probability
what 0.62 really says

2026-08-21 · the single most useful sentence on this site

"Price is probability" gets repeated everywhere. It is actually three statements at once: what the crowd thinks, your break-even bar, and your payoff odds. The second is the useful one and the least often said — it converts an unanswerable question into one you can score yourself on.

1One number, three jobs

ReadingMeaningUse
Crowd estimatethe market says 62%locate the consensus
Your baryou need >62% long-run to profitdecide whether to act
Your oddswin 38¢, lose 62¢size the position

The middle row turns "will they win" — which you cannot verify — into "is my number above or below 62", which you can log, score, and review.

2Why price has to equal probability

It is not a convention; it is enforced. Recall the rule from the previous entry: a full set of outcome tokens converts to $1 and back, at any time.

two sides sum to < $1 → buy the set, merge, pocket the difference two sides sum to > $1 → split a set, sell both, pocket the difference both directions push the sum back to $1

So you do not have to take anyone's word for it — arbitrage pins it. That lane is fully mechanised: I censused it and the live opportunity inventory was zero. Which is exactly what makes the reading safe: someone else is maintaining the identity for you.

3Your real bar sits above the sticker

First markup: the fee

Taking liquidity costs. The fee is not a flat percentage — it is heaviest near 50¢ and cheaper toward both tails:

PriceSticker barFee as % of stakeReal bar
0.5050.0%3.50%≈ 51.8%
0.6262.0%2.66%≈ 63.6%
0.9090.0%0.70%≈ 90.6%

Howyou pay price + fee and receive $1, so the real bar is just their sum. Formula and full table in the taker fee. Makers pay nothing, so a maker's bar is the sticker — at the cost of maybe never filling.

Second markup: bid and ask are different numbers

There are always two prices: what you pay to buy and what you receive to sell. "0.62" usually means the midpoint — a price at which nothing trades.

⚠ So you have two bars, not one

Holding to settlement: your bar is your fill price plus fee.

Selling early: you must first earn back the spread, then the fee, and only then start profiting. On a thin book the spread can be several cents — which makes "instantly underwater on entry" structural, not bad luck.

The same 0.62 is therefore two different businesses on a thick book and a thin one. Look at the depth before you look at the price.

4Is the market any good? Calibration

"Price is probability" comes with its own test, called calibration: collect every market that traded at 0.60–0.65 and check how many actually resolved YES. Land near 62% and the prices are trustworthy probabilities.

You can run this yourself — and you should run it on yourself first:

Build your own calibration table

Before each bet, write down the probability you believe — a number, not a lean. After a few dozen, bucket by your estimate and compare to the realised rate.

Two valuable things fall out: whether you systematically overrate the side you like (nearly everyone does), and which price band you are actually sharper than the market in. Almost nobody is sharper everywhere; plenty of people are sharper somewhere.

This beats any P&L record, because P&L is polluted by luck and calibration is not: with enough bets, calibration error cannot be faked.

5Both tails are systematically mispriced

There is a bias that shows up across betting markets everywhere — favourite–longshot bias: very cheap things tend to be too expensive, very expensive things tend to be too cheap. People like small stakes on big payoffs.

I saw its cleanest form in a zero-fee category. With no fee and no rebate, everyone there is a pure directional trader — and the population splits in two:

GroupMedian fillBehaviour
Harvesting certainty0.99589.8% of buys above 80¢
Buying lottery tickets0.00495.5% of buys below 20¢

Two groups running opposite businesses on the same board, and neither of them anywhere near the middle. The 0.995 side looks safe; it is actually picking up pennies — win 200 times to cover one loss.

⚠ A bias existing is not the same as you being able to harvest it

Knowing cheap tails are overpriced does not mean shorting them pays. That needs tail-risk tolerance, size, and costs that do not eat the edge. My conclusion on that category was "no informational advantage, do not enter."

Between finding an edge and collecting it sit three gates: cost, capacity, and risk tolerance. Most of the bodies in my kill file died at one of those, not for lack of a pattern.

6One line to keep

The price is a probability and it is also your bar. So the question is never "who do I like" — it is "how far is my number from this one, and does that cover fee and spread." When you cannot answer the second half, that is the signal not to bet.

EvidenceCheck it yourself

Arbitrage pins the price split/merge are contract calls available at any time; sum the two sides on any board.
Inventory zero full census of that lane; no live executable opportunity.
Bar arithmetic real bar = fill + fee, fee = 0.07 × p × (1−p) × shares; reconciliation in the four rebates (predicted $45.23 / received $46.58).
Split population measured on a zero-fee category: median 0.995 (89.8% of buys >80¢) vs median 0.004 (95.5% <20¢).
Calibration this one needs no trust in me — twenty of your own bets will settle it.
Checked 2026-08. Fees and rules change; verify before betting.

NextWhere to go

A · MECHANICS
You are buying a verdict, not a stock
A · MECHANICS
The taker fee: why 50¢ costs most
F · METHOD
Your backtest won, but the ruler was wrong
A · MECHANICS
Order book vs house: why this is not a casino
This is an educational and research record. It is not investment advice, promises no returns, and offers no personalised trading recommendations. Rules and API behaviour are per the official documentation; this page states when it was checked and both can change without notice. Prediction markets are restricted or unavailable in some jurisdictions — confirm your own before taking part.