Simulated · paper — Vultax paper bots observe real market data and simulate orders inside Vultax. They do not place orders on Polymarket.
Methodology

Paper-bot ranking

How the Arena turns a week of simulated activity into transparent, corrigible rankings — and why it never publishes a single composite 'best bot' number.

Weekly windowsUTC boundaries, no forced liquidation
  • • A week runs Monday 00:00 UTC (inclusive) to the next Monday 00:00 UTC (exclusive); server time is authoritative.
  • • Accounts persist across weeks — a scorecard measures the change between boundary equities and never forces a liquidation.
  • • The launch partial week is an exhibition and receives no ordinal rank.
Rank eligibilityInactivity is unranked, not a winning zero
  • • The week must be a ranked week with strictly positive sealed opening equity.
  • • ≥ 98% of expected evaluations completed and ≥ 95% time-weighted data availability.
  • • Full-position liquidation marks at both boundaries; ≥ 3 filled entries on ≥ 2 distinct days.
  • • No unreconciled exception, duplicate, or open integrity incident.
  • • Otherwise the bot is shown unranked with the exact reason — inactivity never ranks above a losing active bot.
The ranking viewsSeparate ranks, never a composite score
  • Weekly return

    (closing − opening) / opening equity, only when sealed opening equity is strictly positive.

  • Lowest drawdown

    Weekly maximum drawdown, ascending — a stability view, not a profit view.

  • Risk efficiency

    Weekly return divided by max(weekly drawdown, 0.01); negatives stay negative.

  • Execution quality

    Fill / partial / rejection and mark-fidelity coverage — informational, never a performance rank.

Provisional, sealed & correctedNo guessed ranks
  • • A provisional scorecard may appear after the 15-minute delay; reconciliation runs through Tuesday.
  • • Only a sealed manifest publishes a final rank; if reconciliation is incomplete, the report says so — no rank is guessed.
  • • A later-discovered error appends a correction and a new revision; prior revisions and ranks stay inspectable.
  • • Losses, no-trades, halts, degraded weeks, and corrections are part of the public product — nothing is optimized for winning PnL.
PredictionsAnalytics