Why BASIS Chose Pools

Somebody has to quote the prices on an order book, and they have to be paid for it. Nobody quotes prices in a pool.

That's the whole argument. Everything below is detail and evidence.

What we measured, and why we refuse to average it

Polymarket and Kalshi are order books. They don't take the other side of your trade. Independent traders post bids and asks, you hit them, and a share pays $1 if you're right and nothing if you're not.

Their explicit fees are low but they aren't zero. Kalshi charges up to about 1.75 cents per contract at a 50c price, tapering as you move towards the edges. Polymarket has historically charged nothing to trade, takes a cut on certain market types, and you still pay gas.

The real cost isn't in either fee schedule. It's in the prices. Measured across 111 settled markets it averages 7.6% of everything traded.

Those 111 markets are where every number in this article comes from. Settled Polymarket events, every trade on every outcome, $422m of real money. Fields run from 3 outcomes to 104, median 13. The favourite won 34 of them and a longshot won 77.

One thing to get straight before any of the numbers make sense.

Almost every result here scales with the number of outcomes in the market, and it scales hard. A three-way market and a hundred-way market are not the same product sold at different sizes. They behave differently enough that averaging them together destroys the finding. So nothing here is averaged. Every figure is reported by field size, and where a blended number appears it's labelled as one.

Most of Polymarket is three to six outcomes. Any blended number is really a report on small fields with a rounding error attached.

Wide fields are where pools do their best work, and they're where BASIS is strongest.

There's something worth knowing about the older research too. Most of the academic work comparing pools to the alternative studied bookmakers, firms that quote you a price and carry the risk themselves, rather than order books. Gabriel and Marsden, in "An Examination of Market Efficiency in British Racetrack Betting" (Journal of Political Economy, 1990), found that pools paid winners consistently more than bookmakers did on identical bets. The detail worth sitting with is that the pools in that study were taking 15 to 20% out of every pot and still paying more.

BASIS takes 1.5%.

Somebody has to carry the risk

Adverse selection is the risk of being on the wrong end of somebody who knows more than you do.

On an order book, someone has to post a price and leave it sitting there. That person is exposed. Whoever takes their price might know something they don't, and they find out too late to do anything about it. So the price has to be wide enough to survive getting picked off on a regular basis.

Hyun Song Shin worked this out formally in "Optimal Betting Odds Against Insider Traders" (The Economic Journal, 1991) and "Measuring the Incidence of Insider Trading in a Market for State-Contingent Claims" (The Economic Journal, 1993). He was writing about bookmakers, but the result carries across to order books without much changing. Same risk. Different person holding it.

In a pool there's no price to defend, because nobody posted one. Informed money changes how the pot gets split and that's the end of it. Nobody has to see it coming, nobody has to insure against it, and nobody has to widen a price to survive it.

And that's where the mark-up goes.

A trader quoting prices can't charge only what they expect to lose to people who know more. They also need a return on the capital they've got locked up, cover for the risk of being stuck holding something they never wanted, and their own running costs on top. All of it goes into the price they show you. They charge it to everybody, because they can't tell who's informed and who isn't. You pay it whether or not a single person in that market knows anything at all.

A pool has nobody to pay. There's no one quoting, so there's nothing to mark up, and that turns the headline into arithmetic rather than a claim.

A pool returns 98.5% of everything put into it, in every market, always.

An order book can't match that at any point in its future, because somebody there is always carrying risk and risk always has to be paid for. It isn't a question of how well their traders happen to be doing. It's structural.

Then there's the part that matters most. The mark-up is worst exactly where wide fields live. Quote a 2c outcome and every time you get picked off you lose fifty times the premium you collected, so you either quote it very wide or you don't quote it at all. That's why longshots on order books have such bad prices. It also gets worse as the number of outcomes climbs, which Shin predicted and Cain, Law and Peel confirmed empirically in "The Favourite-Longshot Bias, Bookmaker Margins and Insider Trading in a Variety of Betting Markets" (Bulletin of Economic Research, 2003).

Here's that prediction turning up in real data.

The fee schedule isn't the cost. The prices are.

Where exactly one outcome can win, fair prices add up to 1. On a real order book the prices you can actually buy at add up to more than 1, and that excess is what you're paying for liquidity. Arbitrage squeezes it. It doesn't remove it.

Measured on the trades themselves, weighted by the money that actually changed hands, the medians run:

Read that as one finding rather than four numbers. Cost rises roughly six-fold going from a small field to a wide one, and it does it monotonically. A maker covering a hundred outcomes has to price a hundred longshots, each of which can take fifty times the premium off him in a single hit. He widens or he leaves.

It's worth being clear about what's being measured here, because it flatters the order book. These are the prices people actually paid, which is to say the prices left over after arbitrageurs have already compressed them. The raw quotes sitting on the book before anyone does that work are wider still, and the gap between the two doesn't vanish. It gets paid to the arbitrageurs.

Blended across all 111 markets the figure is 7.6%. That's the least useful number in this article.

BASIS charges 1.5%, fixed and visible, whether the market has three outcomes or a hundred and four. And it stays in the ecosystem: STASIS staker yield, the market creator, the treasury, the protocol.

An empty order book isn't a market

A bookmaker will price anything, because he wants the business. An order book with nobody quoting on it is just nothing. No bid, no ask, no market.

Fixing that doesn't pay. Covering a 104-runner field means running 104 separate books and splitting your capital across all of them. That only makes sense on markets big enough to justify the effort, which is why liquidity reward programmes exist and why they never reach past the top of the list.

It shows up in the data directly. Of 111 markets, 30 never had a live price on every outcome at the same time. Not thin. Absent.

In a pool the pot is the liquidity. One pot, not 104 books. Every outcome has a price and can be bought from the first dollar, with no liquidity mining and no trading desk deciding whether a market is worth their time.

That's the difference between something that works on next month's headline binary and something that works on the 104-runner Player of the Year.

There's no ceiling, and longshots are where it shows

On an order book your payout is fixed at $1 the moment the contract exists. The cheapest you can normally buy is a cent, so 100x is roughly your ceiling, and that's before you find out whether anybody was quoting at a cent in the first place.

A pool has no ceiling. You get your share of the pot, and that's settled afterwards rather than in advance. Put $10 on an outcome nobody else touches in a market holding a million dollars and you take essentially the whole million. Call it 100,000x.

Extreme case, obviously, and your share drops the moment anyone piles in behind you. But that's what happens when your payout is a slice of the pot rather than a fixed dollar, and it isn't hypothetical. Those 30 markets with dead outcomes are markets where somebody could have been the only money on a live runner.

The averages back it up on exactly the markets you'd want them to. Across the 77 markets where a longshot came in, the winning side collectively took 1.33x out of the pool what the same trades took out of the order book. 61% of winning wallets finished ahead, and 71% of the winning money did.

Longshots are the hardest thing for a market maker to price and the cheapest thing for a pool to carry. That gap is what it's worth.

Bets an order book can't afford to offer

This advantage doesn't fade with scale, which makes it the most durable one pools have.

Every bet type on an order book needs its own book with its own capital sitting in it. An exacta on a 20-runner field is 380 books. A four-leg parlay across live markets runs into the thousands. Nobody is going to quote that.

A pool prices all of it out of one pot.

Tranche markets. Podium, top-N and each-way style bets, up to 10 slices, with every share worth the same so a peer-to-peer order book still works on top of the pool.

Combinatorial markets. Exacta through superfecta, box and wheel bets, up to 360 combinations in a single pass. The market opens priced as though every combination already exists, and prices still add to 100% the whole way through.

Ticket pools. Parlays up to 10 legs, with jackpot rollover.

Survivor pools. Up to 52 rounds, resolved lazily as entrants are eliminated.

On BASIS the legs read live BASIS markets straight off the chain, so there's no oracle anywhere in it. Building these products conventionally takes a data feed, a settlement process and a risk desk.

Note also what these are: field-size machines. A four-horse race becomes a 24-outcome exacta. Wide fields are the expensive end of an order book and the cheap end of a pool, and this is how you manufacture that condition rather than waiting for it to turn up.

Fair warning on this part: the competition isn't Polymarket. It's sportsbooks, tote pools and lotteries. Bigger market, slower moving.

What we changed about the classical pool

A classical pari-mutuel solves the counterparty problem and then creates a few of its own. BASIS kept the solution.

Being early should pay more

In a classical pool everybody gets the closing price. The person who called it a year out and the person who piled in ninety seconds before the close are paid exactly the same. There's no reward for being right early, so nobody bothers being right early, which is why tote prices tell you nothing until the last minute.

BASIS mints your shares at the price when you buy. Getting in early buys a permanent slice of the pot, including a slice of every dollar that turns up later at a worse price. Nothing caps that, and it grows with however much late money arrives.

Here's what it was worth on the markets where the favourite won. Money going in during the first 5% of a market's life, against what the same money would have made on the order book:

On fields of 31 and up it beat the order book in every single market.

Same shape as the cost ladder, for the same reason. The wider the field, the more the order book was overcharging and the more room there is underneath it.

No classical pool study can say anything about this in either direction, because in a classical pool it doesn't happen at all.

You lock your price when you buy

In a classical pool you don't know what you've bought until it's over, because the dividend gets worked out at the end. That uncertainty costs something. A sensible person wants paying to accept an unknown payout over a known one.

BASIS takes half of it away. Your price is fixed the moment you buy and you know your share count immediately. What you don't know yet is how big the pot gets.

So it's a trade, and it should be presented as one. You're swapping a guaranteed $1 for an unknown share of the pot that can be worth a great deal more than $1. The uncertainty isn't a defect. It's where the upside lives.

A price from the first dollar

A classical pool has no price until money shows up, so an empty market is useless. Same dead end as an empty order book, reached by a different route.

BASIS seeds every outcome with virtual liquidity. It's a pricing device, never part of the pot, and it costs nothing, subsidises nothing and pays out nothing. All it does is shape how the price moves as money comes in, which means every outcome is priced and buyable from the moment the market opens.

The second market underneath the first

Every BASIS market has a Predict+ token sitting under it. Buying and selling both nudge its price up, so it only ever moves one way. Neither an order book nor a classical pool has anything like it.

It adds structure along two axes that pull against each other.

Getting in early pays. Buyers who arrive later push the price up for the people already holding.

Claiming late pays. Everyone who cashes out pushes the price up for whoever is left, so the last claimants get more per token than the first ones did.

Put together: a boost for being early, then a choice for everyone. Wait and gain more, or take it now and gain less. Anyone who buys late is paying the people who were early, and anyone who rushes the exit is paying the people who are patient. Come in early and claim late and both axes run your way.

Some things fall out of that.

It's the opposite of a bank run. Rushing for the exit pays everyone else and waiting pays you, so a panic puts itself out instead of feeding on itself.

You can hold more than one position at once. Buy the token, borrow against it at 100% LTV while you're still holding it, then bet with what you borrowed. Both positions stay live.

You can win without betting on the outcome. Leverage the token on its own and you've taken a position on how much action the market gets, separately from who wins it.

The conservative case

Credibility is worth more than any single number, so here's the version with every assumption stripped out of it.

A pool doesn't create money. Winners get the losers' money on either venue. What a pool does is leave more of it on the table for the players, because there's nobody standing in the middle taking a cut for providing liquidity. That's the whole of it, and it's what the research has been saying since 1990.

The clean number is 1.07x.

Assume money arrives in exactly the proportions the order book's own prices imply, so there's no crowd effect left anywhere and nothing is being modelled about how people behave. The pool still pays about 7% more. It came out above 1.00x in 92 of the 107 markets it could be computed on.

That 7% is the order book's built-in cost minus the pool's fee, and nothing else. No invented crowd, no behavioural assumption. It's the floor.

On small fields, pools don't have an edge worth talking about, and we'll say so plainly. At three to six outcomes the embedded cost is 1.3%, which is close to free, and a pool charging 1.5% is charging much the same thing. Neither venue creates money out of nothing.

Pools earn their advantage as the field widens. So that's where BASIS competes: big fields, longshots, structured bets, and rewarding the people who show up early. Plenty of ground.

The whole thing in five lines

1. Somebody has to quote prices on an order book, and they charge you for the risk, their capital and their costs whether or not anyone in that market knows anything. Nobody quotes in a pool, so there's nothing to charge.
2. That charge is a function of how many outcomes there are. Measured across 111 real markets it runs 1.3% on a three-way and 7.9% on a wide field. A pool charges the same on both, so the wider the market the more of the gap belongs to the players.
3. An order-book share pays a fixed $1. A pool share pays a slice of the pot with no ceiling on it, and on the 77 markets where a longshot came in the winning side took 1.33x more out of the pool.
4. Exactas, parlays, tranches and survivor pools all price out of that same single pot with no oracle. An order book would need thousands of separate books to do the same, and every one of them turns a small field into a wide one.
5. Being early with BASIS buys a permanent slice of every dollar that shows up later, which a classical pool never gave you. It was worth 1.40x on a three-way and 4.95x on a field of 31 or more.

---

Figures from 111 settled Polymarket markets: all trades on all outcomes, $422m notional, fields of 3 to 104 outcomes. Every headline figure reported by field size.