Market structure · How it works

Three prices, one number.

Every event on Foreseez shows a single consensus. Here is how three venues become one number, and why a plain average would mislead you.

FForeseez ResearchOct 13, 20267 min read

Why not just average?

When three venues price the same event at 64%, 58% and 60%, the obvious move is to average them: 60.7%. The trouble is that the three prices do not deserve equal votes. One might sit on twelve million dollars of orders. Another might rest on a handful of trades from last night.

Think of each venue as a weight on a beam. Where it sits is its price. How heavy it is depends on how much money stands behind that price. The consensus is the point where the beam balances. Pick a method below and watch what the money does.

FIGURE 1 · THE BEAMPut the pivot where the money balances

KalshiPrice64%Liquidity$12.6M

PolymarketPrice58%Liquidity$8.2M

PredictItPrice60%Liquidity$1.1M

ALL THREE METHODSEqual60.7%By volume60.2%By liquidity (ours)61.6%The beam tilts when the pivot ignores where the money is.

Circle size is liquidity. The beam always feels the real money; only the pivot moves with the method. Sample prices and liquidity.

Equal weighting hands a thin venue the same vote as a deep one. Weighting by yesterday's volume is better, but volume measures how much traded, not how hard the price would be to move today. A venue can trade a lot and still have almost nothing resting near the price.

Depth near the price

We weight each venue by the orders resting close to its price: the money someone would have to chew through to move it. Too narrow a window and a single order dominates. Too wide and you count orders nobody expects to fill. Try it.

FIGURE 2How wide a window counts as liquidity?

consensus61.5%

Kalshiprice 64% $2,395kweight 56%

Polymarketprice 58% $1,536kweight 36%

PredictItprice 60% $319kweight 8%

Window: ±2 points

Bars are resting orders at each price level, bids in blue below the price and offers in amber above. The shaded window is what counts. Sample order books.

We settled on two points either side. It is wide enough that one large order cannot swing a venue's weight on its own, and narrow enough that it only counts money that is actually defending the price.

When the rules differ

Weighting only works when the venues answer the same question. If one settles on any touch of a price and the others on a daily close, averaging them blends two different events. We have three options, and we always tell you which one we used.

FIGURE 3Brent above $90: one venue plays by different rules

Converting the touch price to its close-rule equivalent turns 31% into about 25%. Now all three answer the same question and the consensus is 24.7%. This is our default when the model is confident.

Venue names hidden; rules paraphrased. The conversion factor comes from a rules model like the simulator in our first explainer. Sample prices.

A price goes stale

A price is only as good as its last trade. If one venue has not traded for an hour while the others have moved, its number is a memory, not a forecast. So its weight fades, halving every half hour without a trade, until it trades again.

FIGURE 4Let a venue go quiet

Quiet for45 min

Weight kept35%Consensus61.6% → 60.1%Kalshi has been quiet for 45 minutes, so its vote shrinks and the consensus leans toward the venues still trading.

Weight halves every 30 minutes without a trade and snaps back on the next trade. Uses the prices and liquidity from Figure 1.

Put it all together and the consensus is a number with a receipt: which venues count, how much each one counts, which rules it was converted to, and how fresh each price is. Every event page shows that receipt one click below the number.

See a consensus, unpacked Next: What a price really costs

Prices, liquidity and order books in this post are sample data. The method is described as designed and may change. Nothing here is investment advice.