CryptoPreview
Will BNB close above $1,000 on Dec 31, 2026?
Resolves Dec 31, 2026
31%implied YES
YES 31¢69¢ NO
Coming soon — not live yetThe agent asks where a price goes.
The agent asks where a price goes.
Event contracts ask whether it happens.
Helion trades continuous prices: momentum and mean reversion over a live oracle, sized by a hard risk engine, filled on-chain through the aggregator. A prediction market swaps the price for a probability — one question, one resolution date, a contract that settles to exactly 0 or 1. This page is the design work and an honest account of the gap. None of it runs today.
- Status
- Not built
- Venue
- Undecided
- Reused from today
- Engine, risk, wallet, logging
01
Why this agent, specifically
Nothing in the running engine is price-specific except the strategies. The loop — sense, decide, size, execute, log — is indifferent to what it is deciding about. That is the whole argument for pointing it at a different kind of question, and it is the only part of this page that is already real.
Sense
It already ingests live data
The tick loop pulls fresh prices every 20 seconds and keeps per-token history in memory. An odds feed has the same shape — one number per market, refreshed on a clock. What changes is what the number means, not how it arrives.
Decide
It already forms a view it can defend
Every signal carries an action, a requested size, a confidence and a written reason before anything executes. An event contract needs exactly that record, with the confidence pointed at a probability instead of a trend.
Check
It already refuses its own trades
Position caps, per-trade caps, a daily budget, per-token limits and cooldowns sit between every signal and every fill, and can only shrink an order. A contract that can settle to zero needs a gate at least this strict.
Prove
It already signs and shows its work
The private key stays local, live fills land on-chain, and every signal, veto and fill is timestamped in a log you can query. A BNB-native venue could be driven by that same wallet, with the same audit trail.
02
What it would look like
Illustrative — not live market dataIllustrative — not live market data
MacroPreview
Will the Fed cut rates at the December 2026 meeting?
Resolves Dec 10, 2026
58%implied YES
YES 58¢42¢ NO
NetworkPreview
Will BNB Chain settle over 20M transactions a day by mid-2027?
Resolves Jun 30, 2027
24%implied YES
YES 24¢76¢ NO
CryptoPreview
Will BTC set a new all-time high before Apr 1, 2027?
Resolves Mar 31, 2027
67%implied YES
YES 67¢33¢ NO
EquitiesPreview
Will NVDA close above $250 on the last trading day of 2026?
Resolves Dec 31, 2026
43%implied YES
YES 43¢57¢ NO
MacroPreview
Will US CPI print below 2.0% year-over-year in any month of 2027?
Resolves Dec 31, 2027
22%implied YES
YES 22¢78¢ NO
Questions, probabilities and dates above were written by hand to show the format: a question, a fixed resolution date, an implied probability and the YES/NO prices that sum to a dollar. There is no odds feed behind them, nothing refreshes, and none of these is tradable in Helion.
03
How it would actually work
A market is a question with two sides. YES and NO trade as a pair, priced in cents that add up to a dollar, and the price of YES isthe market's implied probability. Buy YES at 31¢ and you are paying 31 cents for a contract that pays exactly one dollar if the event happens and exactly nothing if it does not. There is no long tail either way — on the resolution date the position is worth 1 or 0 and the trade is over.
| Dimension | Token trade — live today | Event contract — not built |
|---|---|---|
| Instrument | A BEP-20 token with an unbounded, continuous price | A YES/NO pair priced between 0.00 and 1.00 — the price is the implied probability |
| Terminal value | None. The price keeps moving for as long as the token exists | Exactly 0 or 1, on a resolution date that is known when you enter |
| Exit | Sell into a DEX route at any time, at whatever the book pays | Sell to a counterparty in the venue book, or hold the contract to settlement |
| Downside | Marked to market and cut by the −5% stop before it compounds | The entire stake, if it resolves against you. A percentage stop is meaningless here |
| Sizing input | Share of portfolio value, capped notional, per-token budget | Edge against the implied probability, and how much time is left to resolution |
| Signal source | SMA(10/40) crossover, RSI(14), scheduled DCA — all shipped | Base rates, news and data-driven mispricing — none of it written |
The strategies do not transfer
RSI(14) and SMA(10/40) read a price series with no fixed endpoint — that is the entire premise of a moving average. A probability has an endpoint, and it is converging toward it by construction, so momentum on it is close to meaningless: a contract drifting from 60¢ to 70¢ as its date approaches is doing exactly what it should. Event markets need their own strategy work — mispricing against base rates, news and data-driven edge, and decay into the resolution date. That work has not started, and reusing the current strategies would be worse than shipping nothing.
04
What it takes to ship
Tokenized equities were easy. xStocks are ordinary BEP-20 contracts, so supporting them meant adding the address and letting the aggregator route it — the execution path never changed. Prediction markets are not that. Checked empirically against the aggregator's route API: no prediction-market outcome token is routable. A contract is a position in a venue's own orderbook, not a token you can swap, which means a second execution adapter alongside the swap path — quoting, order placement, fill handling and settlement, none of which exists in the current code.
| Route | Where it lives | What it would need | Lift |
|---|---|---|---|
| An on-chain venue (BNB-native) | BNB Chain mainnet, TypeScript SDK | The keypair the agent already holds, plus an adapter for the venue orderbook. No bridge, no second chain, no KYC, no new custody model. | Lightest |
| Polymarket | Polygon | A second chain end to end — a bridge for collateral, a second wallet, a second RPC path and a second set of signing rules alongside the BNB Chain one. | Heavier |
| Kalshi | CFTC-regulated US exchange | An exchange account, KYC and bank rails. Off-chain end to end, so the wallet plays no part and none of the on-chain proof surface carries over. | Heaviest |
The BNB-native route is the one that fits the current architecture: same chain, same keypair, same signing code, nothing new to custody. Liquidity depth on all three has not been measured, so this page carries no volume, market-count or depth figures — including here.
05
Build status
Shippedrunning today
- Execution engine
20-second tick loop over a self-scanned universe, live prices from market-wide pool data, real swaps routed and confirmed on mainnet.
- Risk manager
Position caps, per-trade caps, daily and per-token budgets, cooldowns, stop-loss and take-profit sweeps on every tick.
- Wallet and local signing
Versioned transactions signed with a keypair that never leaves the machine. No custodian in the path.
- Decision logging
Every signal, veto and fill timestamped, persisted and queryable over the local API.
- On-chain proof
Every live fill carries its transaction hash and is verifiable on BscScan, independently of this app.
Not built yetno code, no date
- Venue adapter
A second execution path that quotes, places, fills and settles against an orderbook. The only execution path today is an aggregator swap.
- Event market discovery
The universe scanner understands BEP-20 contracts and liquidity gates. It has no concept of a question, an outcome pair or a resolution date.
- Odds feed
No probability data enters the tick loop. Nothing in the engine reads or stores an implied probability.
- Settlement accounting
The ledger models a continuous quantity marked to a live price. Nothing models a position that expires to exactly 0 or 1 on a date.
- Event strategies
No strategy reasons about probabilities, base rates or time to resolution. The three that ship are price-trend tools.
The right column is the honest reason there is no launch date on this page. Prediction markets are on the roadmap; they are not scheduled, not in progress and not tradable.
The prediction-market work above is unbuilt. The trading engine is not.
The console is where that loop runs — a self-scanned universe spanning majors, memecoins, DeFi, liquid staking and tokenized equities, a decision every 20 seconds, every signal and veto on the record, and every live fill verifiable on BscScan.