Searches loaded events. Load history and Load more to include older records. Times match IST or UTC.
Paper executions only. Full dates and milliseconds in IST; CSV timestamps in UTC. History is saved in PostgreSQL.
48-hour data collection
Loading collection status…
Runs on the Swiss VM with your browser closed. BTC and ETH · 576 five-minute intervals each. Downloads include data collected so far; use them after the end time for the full window.
Polymarket: observed trade-event OHLC, counts, shares and price × shares notional, separately for UP/DOWN. Gaps and missing sizes are flagged. These are not unique people or guaranteed exchange-wide volume. Exchange volume is BTC/ETH units. All downloads use UTC.
Cross-platform observations
Waiting for samples…
USD and USDT are different quote currencies. CoinGecko is a slower aggregate. Correlations use changes, require 20 valid pairs, and exclude stale quotes and market rollovers. Association does not establish causation or a profitable trade. These research observations do not place orders.
Five-minute comparison
Loading chart data…
Binance · five-minute spot candles (USDT)
Bybit · five-minute spot candles (USDT)
Inside selected window · price movement
Green: Binance · Gold: Bybit. Basis-point change from each venue's first fresh sample. Both use USDT.
Same window · Polymarket UP price (cents)
Polymarket measures a contract outcome price, not the BTC/ETH spot price. Lines use observed 15-second samples, with gaps for missing quotes. Compare timing and direction; these charts do not establish a profitable edge.
PostgreSQL history
RESEARCH CONTRACT
Fast decisions. Explicit assumptions.
The native engine calculates signals and risk in memory. Redis holds short-lived snapshots, PostgreSQL stores history, and Jev runs outside the execution loop.
Buy at asks, sell at bids, and require sufficient depth within the order limit.
Wait for a post-arrival book after network and venue delay. Stops trigger attempts, not guaranteed prices.
Block entries on missing reference prices, unknown fee curves, feed gaps, or portfolio losses of $5.
Local model probabilities are experimental and uncalibrated. Jev confidence is not a trade win probability.
Jev/API hosting costs are recorded separately and are not included in trading P&L.