BTC-USD perpetual: one PaperTrade decision (2026-10-16, 18:31 UTC)

Decision (JSON)

{
  "direction": "long",
  "leverage": 750,
  "take_profit_bp": 14,
  "take_profit_price": 81418.38,
  "bust_distance_bp": 8.5333,
  "bust_price_estimate": 81235.17,
  "expected_net_return_on_margin_pct": 80.3,
  "confidence": 0.38,
  "regime": "volatile_choppy",
  "key_levels": { "support": 81041, "resistance": 81334 },
  "entry_mid": 81304.55,
  "data_time_utc": "2026-10-16T18:31:50Z",
  "reasoning": "Over the last 30 minutes BTC has built a sequence of higher lows (81041, 81066, 81075, 81126, 81158, 81268) under a flat ceiling at 81330-81334 after a capitulation low at 80694 on the day's largest 15m volume (299 BTC on Coinbase) and a V-shaped rally to 81710; perps trade at a small discount to index (OKX premium -0.026%, Hyperliquid basis -0.041%), which is squeeze fuel for a breakout. 750x gives the widest bust distance allowed (8.53 bp, about 69 USD), which sits below the lows of the last six 1m candles, while realized 1m volatility has contracted from about 10 bp/min (17:30-17:45) to 4.6 bp/min (last 30 min) and 3.2 bp/min (last 5 min). A 14 bp target reaches the bottom of the 81420-81450 supply shelf, clears the haircut with about 76% of the raw move kept (about 80% return on margin), and is of the same order as a single breakout minute (several 1m candles today exceeded 14 bp).",
  "risks": "A rejection at 81330 that rolls back to the 18:23 low region (81212) is a 11 bp drop and would liquidate the long; the 15m trend is still down from 82831, so a resumption of the day's selling or a surprise headline in the US afternoon session is the main liquidation path."
}

Important caveat on entry_mid: see "Data limitations" below. The Hyperliquid mid could not be fetched from this machine. entry_mid is the live Kraken BBO mid (bid 81304.50 / ask 81304.60), corroborated by Coinbase's best bid of 81307.77 at 18:31:47 UTC. Hyperliquid's BTC perp traded 5 to 10 bp above Coinbase spot earlier in the hour, so the user should recompute take_profit_price and bust_price_estimate from the live Hyperliquid mid using the bp values, which are the decision.

Question

Produce one BTC-USD perpetual trade (direction, leverage 750-1000, take-profit in bp of price) for PaperTrade on HyperEVM, which fills and liquidates at the Hyperliquid mid, optimizing gains after PaperTrade's asymmetric win haircut and with liquidation as the only stop.

What was fetched (facts, with sources)

All fetches were made between 18:30 and 18:32 UTC on 2026-10-16 via HTTP GET.

News and calendar (facts as reported by the pages):

• Yahoo Finance, Oct 16 2026: BTC "firm despite two strong inflation reports" (September CPI on Oct 14 and PPI on Oct 15 both came in hot); spot BTC ETFs took in 5 M USD on Thursday after a 261 M USD outflow on Wednesday (SoSoValue).
• Fortune, Oct 16 2026: BTC about 82,645 USD at 7:45 a.m. ET.
• Nexo Markets Today, Oct 16 2026: BTC briefly broke below 80,000 earlier in the week for about two hours; down about 0.9% on the week and over 2% in October.
• KuCoin daily report, Oct 16 2026: links pressure to AI-financing concerns.
• Calendar: a TD Economics weekly table and the Guggenheim 2026 US calendar show the only scheduled US release on Oct 16 as import/export prices at 8:30 a.m. ET, already passed. No Fed speaker schedule for the afternoon of Oct 16 could be found (unanswered, see below).

Metrics (computed from the fetched candles, in bp of price)

Trade geometry at 750x: bust distance 10000/750 - 4.8 = 8.53 bp, about 69 USD at 81,305. At 1000x: 5.2 bp, about 42 USD.

1. Realized 1m volatility (RMS of close-to-close changes, Coinbase 1m):

   • 17:59-18:28 (29 changes): RMS 37.2 USD = 4.6 bp/min.
   • Last 10 minutes: 4.8 bp/min. Last 5 minutes (18:24-18:28): 3.2 bp/min.
   • 17:30-17:45: individual minutes moved 100-300 USD (12-37 bp); the 17:34 minute ran 80962 to 81316 (+44 bp) on 74.9 BTC.
   • Inference: volatility is contracting from an extreme burst into a compression, but 4.6 bp/min is still more than half the 750x bust distance per minute. This is the core risk and why leverage is at the floor.

2. Per-candle adverse excursion vs bust (Coinbase 1m, last 30 candles, 17:59-18:28):

   • Candles whose (high - open) or (open - low) exceeded 69 USD (the 750x bust): 6 of 30 (about 20%).
   • Exceeding 42 USD (the 1000x bust): roughly 15 of 30 (about 50%).
   • Inference: at 1000x a single ordinary minute busts the trade about half the time; at 750x about one minute in five does. 750x is clearly preferable; the bust-first probability still dominates.

3. Driftless barrier probability: with bust 8.53 bp and target 14 bp, P(target before bust) = 8.53/(8.53+14) = 37.9% if price is a martingale. Net payoff on a win is about 80% of margin (see 7), loss is 100%. Inference: the trade is negative expectancy without directional edge; the directional evidence below is what the decision rests on, and confidence 0.38 reflects roughly the driftless odds.

4. Trend and structure:

   • 15m: lower highs 82831 (11:15) -> 82384 (14:15) -> 81879 (15:00) -> 81710 (17:41). Price 81305 is about 25 bp below a rough 5-hour average close (about 81500) and about 68 bp below Kraken's 24h VWAP (81857). Position in the 24h range: (81305 - 80694)/(82831 - 80694) = 29%, lower third. Inference: intraday trend is down.
   • 5m/1m (18:00-18:28): higher lows 81041, 81066, 81075, 81126, 81158, 81212, 81268 against a flat ceiling at 81330-81334 (tested at 18:17, 18:18, 18:19 and again at 18:27-18:28). Inference: ascending triangle, a continuation pattern of the 17:30-17:45 rally; a break of 81334 would likely travel to the 81385-81450 shelf (the 17:44-17:46 candle bodies) within a minute or two.
   • Timeframe agreement: 1m up, 5m up, 15m down. Mixed, hence regime = volatile_choppy.

5. Order flow:

   • Coinbase 15m candle at 17:30 printed 298.8 BTC, the largest of the 24h window, at the day's low, followed immediately by a 190 BTC and a 299 BTC 15m rally. Inference: capitulation and absorption; the sellers who pressed below 80,700 were taken out.
   • Coinbase trades 18:29-18:31: largest prints were taker sells (0.75 BTC at 81267.63 at 18:30:57) and price rose to 81307 within 50 seconds. Inference (weak, three minutes of data): passive bids are absorbing sells.
   • Coinbase top 25 bid levels hold about 1.6 BTC within 20 USD of the touch. Inference: resting liquidity is thin, so the move either way will be fast; asks could not be read (see limitations).

6. Perp premium, funding, OI:

   • OKX premium -0.0258%, funding at the 0.01% baseline, last settled 0.0051%. Hyperliquid basis -0.041% (CoinGecko, 18:21 UTC), funding near zero. Binance BTCUSDT funding 0.0085%, Bybit 0.0017%.
   • Inference: perps sit at a small discount to index after the dump, shorts are mildly crowded, and positioning favors an upside squeeze rather than further downside acceleration.
   • OI: OKX 30,492 BTC, Hyperliquid 3.39 B USD. No history was fetched, so OI change vs price change is unanswered.

7. Haircut and payoff at 14 bp, 750x: move after deadband 13.8 bp; scale = 0.90/(1 + 1/(0.00138 x 15000) + 100000/(1e6 x 0.00138 x 814.598)) = 0.90/(1 + 0.0483 + 0.0890) = 0.791; after the 2% win fee the kept move is 13.8 x 0.791 x 0.98 = 10.70 bp, 76% of the raw 14 bp; return on margin = 10.70 bp x 750 = 80.3%.

8. Time of day: 14:31 ET on a Friday, 90 minutes before the US equity close, no scheduled US releases ahead. Weekly options expiry and funding settlements on Hyperliquid (hourly) are not expected to matter at this size.

Why long, why 750x, why 14 bp

• Direction: the 30-minute structure (higher lows into a flat ceiling, capitulation volume at the low, perps at a discount) points up for the next 10 to 30 minutes, which is the horizon of the trade. The 15m downtrend argues for a short, but the short side already had its move and a short initiated into an ascending triangle would be busted by the first breakout minute (14 to 44 bp candles have occurred in the last hour).
• Leverage: 750 is the lowest allowed and the only choice where the bust distance (8.53 bp) exceeds most single-minute excursions. It costs 25% of the payoff relative to 1000x but roughly halves the per-minute bust frequency.
• Target: 14 bp (81418) sits just below the 81420-81450 supply from 17:44-17:46; 15 bp or more would require trading through that shelf. It keeps 76% of the raw move after the haircut, above the point where the haircut becomes punitive (under 8 bp keeps under 70%).

Data limitations and unanswered questions

• Hyperliquid API not reachable from this environment. The Hyperliquid info endpoint accepts only HTTP POST, and the only network tool available here performs GET. The shell was also unavailable (every command failed in the sandbox with "apply-seccomp: write /proc/self/setgroups ... Permission denied"), so the specified candleSnapshot, metaAndAssetCtxs, l2Book, recentTrades and fundingHistory calls could not be made. entry_mid is therefore a proxy (Kraken BBO mid, 0.4 bp from Coinbase's live best bid), not the Hyperliquid mid. The Hyperliquid last price at 18:21 (81261, CoinGecko) was 5 to 10 bp above Coinbase spot at the same minute, so the live Hyperliquid mid is probably 81,300 to 81,400 at decision time. The bp values are the decision; the prices should be recomputed from the live mid.
• Binance blocked (HTTP 451), so Binance candles, book and premium index were not used.
• Order book asks on Coinbase were not readable (response truncated after the bids), so book imbalance within 10 bp of mid and resting liquidity between entry and target are unanswered.
• CVD could only be sampled over about three minutes of Coinbase trades, not the full 1000-print window.
• OI change vs price change is unanswered (no OI history fetched).
• Fed speakers for the afternoon of 2026-10-16 could not be confirmed; a surprise appearance is an unquantified risk.
• The candle metrics were computed by hand from the fetched rows; the RMS and excursion counts are accurate to roughly 10%.
• Nothing here is a forecast with positive expected value. Under driftless assumptions the trade loses money on average because of the 4.8 bp bust offset and the haircut; the long only pays if the breakout thesis has real edge.