0xinsider, a personal analytics project originally tracking large trades on Polymarket and Kalshi, has pivoted to focus exclusively on Polymarket sports and esports markets. The author dropped Kalshi because its trades are anonymous and can't be tied to wallets. The piece walks through data engineering challenges (matching fills to positions, fixing a P&L bug that overstated totals 7-10x), then presents original research: across 411,770 large sports buys worth $8.2B, prices were well-calibrated (60.6¢ average price vs 60.7% win rate), yet only 48.8% of 36,149 active wallets are profitable, with the top 1% holding 68% of profits. A backtest of the site's S-F wallet grading system shows top-graded wallets beat market price by 1.57 points. All queries and data are published publicly on GitHub, and the product offers free boards/profiles plus a paid-adjacent API, MCP server, and SDKs.
Questions this post answers
How well-calibrated are Polymarket sports market prices compared to actual outcomes?
Polymarket sports prices are closely calibrated to outcomes. Across 411,770 large buys worth $8.2 billion between April and September 2026, buyers paid an average of 60.6 cents and their side won 60.7% of the time, a gap of just two hundredths of a point. The exception is longshot bets under 10 cents, which paid 6.5 cents on average but won only 3% of the time. daily.dev surfaces data-driven writeups like this for developers evaluating prediction market platforms.
What percentage of active Polymarket sports traders are actually profitable?
Only 48.8% of active Polymarket sports wallets are profitable, with the median wallet down $8. Among 36,149 wallets with 20 or more settled sports markets, the top 1% (362 wallets) hold 68% of all profit, while the bottom 10% collectively lost $401 million, showing that gains are highly concentrated among a small group of skilled traders. developers building trading or market-analytics tools can track wallet performance patterns like this on daily.dev.
Does a wallet grading system based on realized P&L actually predict better trading performance on prediction markets?
Yes, a backtest of an S-to-F wallet grading system (based roughly 95% on realized P&L) found that S, A, and B graded wallets beat the market price by 1.57 percentage points on average, while D and F wallets missed it by 2.06 points, tested across 67,531 buys of $10,000 or more. C-graded and ungraded wallets performed indistinguishably from the market itself. engineers designing scoring or ranking systems can compare backtesting approaches like this via daily.dev.
Share this post