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

Paper-bot simulation

Vultax paper bots observe real market data and simulate orders against a captured order book. Here is exactly how the simulation works — and what it refuses to invent.

The invariantTrue on every surface, share, and export

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

Market eligibilityRecomputed every 5 minutes from a versioned catalog
  • • Standard binary Polymarket markets with two canonical outcome tokens and a resolvable identity.
  • • Order book enabled and accepting orders, book no older than 15 seconds, two-sided and not crossed/locked.
  • • Resolution between 24 hours and 180 days out; intended price between 0.05 and 0.95.
  • • Provider liquidity ≥ 5,000 USD-equivalent and executable depth ≥ 500 PAPER_USD through one point of price.
  • • A contiguous, gap-free, quality-approved input watermark for every strategy lookback.
Execution & fillsMarketable limits, exact depth, no synthetic fills
  • • Entries target 100 PAPER_USD as a marketable limit, capped one point above best executable price, with a modeled impact ceiling.
  • • Fills consume exact captured multi-level depth and record partial fills at true VWAP. If under 25 PAPER_USD can fill inside the cap, the entry is rejected — a smaller position is not opened.
  • • Exits are capacity-bounded: when liquidity is absent the position becomes exit-pending and retries; there is never a synthetic or forced fill.
  • • Reward Radar rests conservative post-only quotes with a queue-ahead model; a touch alone never fills, and gaps freeze matching.
Marking, PnL & feesConservative liquidation, FIFO cost basis
  • • Marks use a quality hierarchy: full-position liquidation VWAP first, then progressively degraded states that are analytics-only.
  • • Realized PnL uses deterministic FIFO lot disposition net of event-time fees; booked fees hit the cash journal exactly once.
  • • Estimated liquidity rewards are labeled estimated-not-booked and are excluded from cash, equity, PnL, drawdown, and rank.
What is never fakedFail-closed by design
  • • No last-known-good data creates a new simulated fill; a failed input gate records a no-trade or data halt.
  • • A healthy zero-opportunity interval is distinct from missing data.
  • • Positions with unavailable marks are never substituted with cost basis.
  • • Ambiguous outcomes stay unresolved; no payout is invented.
PredictionsAnalytics