Leverage for prediction markets.

Phoebe lends against Polymarket outcome shares. Post USDC margin, borrow from a tranched LP pool, and trade up to 4× on the markets you believe in. The risk engine is built for what breaks every other lender here: prices jump, and every market ends at 0 or 1.

LP PnL, 60 unseen markets
+$523k
Same markets, naive 75% LTV
−$7.98M
Senior tranche losses
$0
Notional replayed
$118M

Backtest results, not live performance: 60 resolved Polymarket markets replayed through the contracts with synthetic traders. Methodology

Φοίβη · “the radiant”

Phoebe was the Titaness of the moon, and the keeper of the oracle at Delphi before it passed to Apollo. Ours keeps an oracle too: every price lands on chain with its volatility, so leverage tightens before the market moves, not after.

The risk engine

How the risk engine sets leverage.

A position's liquidation LTV is the lowest of four limits, re-evaluated at every oracle update. The binding one is shown in the app, so you always know why your leverage is what it is.

Α

Price bands

No leverage within 3¢ of 0 or 1. It ramps to full between 15¢ and 85¢, where a share has room to move both ways.

0.03 · 0.15 — 0.85 · 0.97
Β

Time ramp

Scheduled resolutions are known jumps. Leverage fades to zero over the three days before a cutoff 24h ahead of them.

ramp 3d · cutoff T − 24h
Γ

Volatility + jump cap

The margin cushion must cover a two-sigma day plus a 10¢ gap, so calm markets that suddenly break stay covered.

LTV ≤ 1 − (2σ₁ᴅ + 0.10) / p
Δ

Absorb, never fire-sell

Underwater positions are written off against junior and their shares sold after a 6h cool-off. A Dutch-auction bonus keeps liquidators in thin books.

bonus 3% → +1%/h → 15%

Leverage calculator

Max leverage, by market state.

Set a share price, its volatility and the time to resolution to see the maximum leverage the engine allows, computed with the same code the app runs.

share price 50¢
max leverage 2.4× limited by volatility

For liquidity providers

Senior and junior tranches.

Choose your exposure. Losses run through junior before they can reach senior, and junior is paid for standing in front.

senior · 80%protected

Senior

Paid interest first and impaired last. In the 60-market holdout backtest, senior was never impaired.

  • Borrow interest, first claim
  • Behind the whole junior tranche
junior · 20%first loss

Junior

Takes the first loss on bad debt, and collects the fees that pay for that risk: every opening fee and a cut of every liquidation.

  • 0.3% of every amount borrowed
  • 2% of every liquidation

Backtest

Replayed on real Polymarket history.

60 high-volume resolved markets that were never used for tuning, 10-minute ticks, replayed through the real contracts with a $1M pool each. Synthetic traders open at 90% of the allowed LTV and never exit early.

naive · flat 75% LTV −$7.98M LP PnL, summed over 60 pools
phoebe · full engine +$523k LP PnL, same markets, same traders
Naive, flat 75% LTV−7,980k
Bands + time ramp + volatility−15k
+ 0.3% opening fee+260k
Full: + jump cap, absorb, auction, caps+523k

Backtests are not a forecast. Binary markets can always jump to zero; the aim is a profitable LP book across many markets, not zero bad debt.

Try it on testnet.

Testnet funds only. Mint mock USDC in the app, open a position, or provide liquidity to either tranche.

replay the Biden nomination locallydemo/replay.sh sim/data/biden-nominee.json -- --speed 20 --traders 12