snipe.exchange / research · essay · 2026-05
Why preset strikes brokeprediction markets.
A structural argument, in five sections, for the market design Snipe sits on top of. Citations are conservative; conclusions are not.
The bucket problem
Prediction markets, as a category, inherited a quirk from sports betting: they sell buckets. The trader picks BTC closes above $100,000, or above $95,000, or above $105,000 — never above $97,400. The strike is rounded into the venue’s pre-issued grid. The view is rounded into the venue’s vocabulary.
For event-prediction markets — “will the Fed cut rates?”, “who wins the election?” — bucketing is reasonable. The state-space is genuinely discrete. For continuous outcomes, bucketing is a UI failure dressed up as a feature. The probability distribution is continuous; the trader’s conviction is continuous; only the venue insists on slicing the curve into ten or twenty squares.
The discretisation cost is invisible in the spread, but it shows up in the trader’s edge. If you think BTC closes above $97,400 and the closest strike is $100,000, you are buying a different bet — one with a different probability, a different payout, and a different break-even — and calling it the same.
The two-leg tax
Worse than the bucket problem is the range problem. Suppose you think BTC closes between $80,000 and $85,000 — a constructive directional-range view that perp DEXes cannot express and option markets only express through verticals (call spreads, put spreads, condors). On a preset-strike prediction venue, the closest you can get is:
- Buy BTC ≥ $80k (a binary “yes”).
- Sell BTC ≥ $85k (a binary “no” — or equivalently, buy the inverse).
1a · preset venue — two legs
1b · snipe — one range contract
Two trades, on two separate markets, each with its own order book, its own depth, its own fee. The payoff approximates the range bet, plus or minus the friction of two-leg execution. The capital sits on both sides until expiry, half of it earning nothing.
We call this the two-leg tax. It is the dominant friction in expressing range views on existing prediction venues, and it is the single most defensible reason a new architecture is worth building.
Fragmentation, compounded
The bucket problem and the two-leg tax compound through liquidity. Every preset strike spins up its own isolated pool, so depth gets sliced thin, slippage compounds, and the bigger your conviction, the worse the execution. Polymarket and Kalshi run a separate book per listed market; depth is per-strike by construction.[1] The books are thin in the first place because the grid that issues them is dense.
Concentration is the standard answer. Shared, pooled liquidity exists across DeFi precisely because it concentrates depth where it would otherwise fragment. The question — for prediction markets specifically — is whether the same concentration trick works when the asset being priced is a contract that doesn’t exist until you ask for it.
DeepBook’s answer is yes. A single vault holds collateral; contracts are priced on demand by an on-chain oracle and the vault’s accounting; the strike grid evaporates. Every trade is its own market.[2]
The shared-vault answer
Snipe’s architecture, end-to-end, is three primitives stacked. DeepBook Predict provides the contract layer — continuous strikes, native ranges, oracle settlement, a shared liquidity vault. DeepBook Margin provides the borrow/lend infrastructure — collateral management, liquidation mechanics, the path from spot leverage to leveraged predictions.[3] Snipe provides the frontend — a pro terminal, mono numerals, hotkeys, the discipline to never disguise risk.
The math is options-like under the hood. A range bet on BTC ∈ [80k, 85k] is — conceptually — a long call at 80 and a short call at 85, settled at expiry against the oracle. The protocol does the legging. The trader sees one contract.
We are not the first to propose this shape. We are — to our knowledge, in May 2026 — the first to put a pro terminal on it on Sui: on testnet today, mainnet when DeepBook Predict mainnet ships.[4]
What it means for traders
The unit economics are unambiguous. Single-contract ranges replace two-leg execution: 1 fee instead of 2, 1 slippage instead of 2, 1 position to manage. Shared vault liquidity concentrates depth: tighter spreads at meaningful size. Leverage on directional or range conviction extends the view without doubling the capital. Liquidation surfaces in the trade ticket, before the sign.
We have a strong prior that pro traders adopt better market structure quickly once a credible terminal exists. That is a belief, not a measurement — the testnet exists to test it. The bottleneck is not demand. The bottleneck is the terminal.
That’s the work. That’s why we’re shipping it.
notes
- Polymarket and Kalshi list one order book per market; no depth is shared across strikes. Venue structure as observed 2026-05. verified↩
- Continuous strikes, native range markets, oracle settlement, and the shared vault are protocol properties of DeepBook Predict — see the DeepBook Predict documentation ↗. verified↩
- Borrow/lend, margin pools, collateral management, and liquidation mechanics are confirmed in the DeepBook launch transcript; not every Margin parameter is in public docs yet. verified · transcript↩
- Our 2026-05 competitive survey found no pro-trader frontend for DeepBook Predict. A first-mover claim is falsifiable by a single counterexample; we hold it loosely. inference · to confirm↩
updates
- 2026-06-11
- Revision pass: added notes [1]–[4] with source tags; drew fig. 1 (the two-leg tax); moved the contents list into the margin on wide screens; rewrote the venue-liquidity claim to what our research supports and relabelled the adoption argument as a belief rather than evidence; dropped the serial “01” until a second essay exists. No conclusions changed.
next: the live board → · the vault →