@dimes_fi

With the World Cup group stage behind us, some numbers on @Polymarket's pricing. TLDR: the market was near-perfectly calibrated AND discriminating. It got both the probabilities and the rankings right. ✅ AUC = 0.81 (CI [0.72, 0.88]). For any outcome that happened and any that didn't, the market gave the real one the higher price 81% of the time. ✅ Pre-game favorites won 66% of the time (46 of 70) with the market pricing them at 64%. ✅ Scoring each market's accuracy (Brier) at every point before kickoff: three weeks out, a week, a day, close, we see that the line 3+ weeks early was as sharp as the closing price. No real edge from waiting. ✅ Dimes-powered levered positions won ~78% vs ~62% on Polymarket at large, driven by stronger selection toward favorites. ❌ The one big crack was draws which landed at 27% vs ~21% priced, and 29% vs just 9% priced for lopsided games (80%+ favorites). In a 48-team field full of debutants, the giants got held back. Spain, 90% to beat Cape Verde: 0-0. England, 84% over Ghana: 0-0. Ecuador, Switzerland and Portugal all dropped points to rank outsiders. Now the real fun begins.

Since launching: 48% of positions powered by Dimes finished in the green, net of fees +45% cumulative return on all capital deployed, winners and losers, fully netted 82% of winning positions beat their unlevered counterfactual Leverage made the winners worth 3.3x more on average

An early slide from Dimes' institutional raise for levered-position margin capital. The 5 core risks for leverage providers in prediction markets: 1. Jump and resolution risk 2. Execution and unwind cost risk 3. Liquidity and market microstructure risk 4. Netting and concentration risk 5. Venue dependency risk

On Polymarket, ~60% of non-MM positions are held to resolution. In US options, only ~10% are ever exercised. One is passively held. The other, actively traded. With deeper liquidity and capital efficiency tools like leverage, PM turnover distribution will increasingly look like the option market, where a position is something you dynamically trade, not something you wait on.

Dimes now processes 350-500k datapoints per minute from @polymarket to power its risk and execution infrastructure. Running more smoothly by the day thanks to the work of @devjoshstevens, @SuhailKakar, and team. Prediction Capital Markets https://t.co/Ef1iFOTb2G