{
  "direction": "short",
  "leverage": 850,
  "take_profit_bp": 12,
  "take_profit_price": 81406.6946,
  "bust_distance_bp": 6.964705882352942,
  "bust_price_estimate": 81561.26548705883,
  "expected_net_return_on_margin_pct": 76.22769094884256,
  "confidence": 0.51,
  "regime": "ranging",
  "key_levels": {
    "support": 81483,
    "resistance": 81575
  },
  "entry_mid": 81504.5,
  "data_time_utc": "2026-10-17T04:47:35.591000Z",
  "reasoning": "Hyperliquid POST https://api.hyperliquid.xyz/info snapshots (candleSnapshot, metaAndAssetCtxs, l2Book and recentTrades; saved under research/) show price below the six-hour candle VWAP of 81549.89 and a -4.28 bp mark/oracle premium; the latest ten prints contain 0.17061 BTC of sells versus 0.00274 BTC of buys, a very small sample. I infer a modest short bias despite rising five-minute and fifteen-minute EMA9s, with support at 81483 and resistance at 81575. At 850x the 6.96 bp bust allows more rebound room than 1000x, while a 12 bp target beyond support retains about 74.7% of the gross gain after the supplied haircut and win fee. Last-15-minute close-return volatility is 1.70 bp per minute, but none of 30 overlapping historical 30-minute windows reached this target in either direction, so reaching it within 30 minutes requires expansion and is uncertain.",
  "risks": "A rebound to the estimated bust, potentially before resistance at 81575, forfeits the entire margin; rising five-minute and fifteen-minute averages oppose the short. Confidence is a weak directional judgment, not a calibrated target-hit probability, and the stated return is conditional on take-profit rather than an expected value across wins and liquidations."
}