{"direction":"long","leverage":750,"take_profit_bp":12,"take_profit_price":81640.35,"bust_distance_bp":8.53,"bust_price_estimate":81472.92,"expected_net_return_on_margin_pct":67.3,"confidence":0.42,"regime":"ranging","key_levels":{"support":81415,"resistance":81699},"entry_mid":81542.5,"data_time_utc":"2026-10-17T14:58:32Z","reasoning":"Over the last 60 minutes BTC climbed from a 81415 low to about 81543. On 1m and 5m, EMA9 sits above EMA21 (5m: 81527 vs 81512), price is 16 bp above the 24h VWAP, and the perp trades about 4 bp below oracle (premium -0.039%, mark 81542 vs oracle 81575). Together these give a mild upward lean. Realized volatility is very low (0.76 bp/min over 15m, 1.45 bp/min over 60m; median 1m range 1.5 bp), so the lowest allowed leverage (750x) is used to get the widest bust distance of 8.5 bp. The 12 bp target keeps about 76% after the haircut and can be reached in about 30 minutes at the 60m volatility. In a replay of the last 60 one-minute candles at these settings, longs hit the target 5-19 times, depending on target size, and were liquidated once, while shorts were liquidated 25 times.","risks":"The ask side holds more than twice the bid size within 10 bp (101 vs 43 BTC), and the 1h high of 81577 is overhead, so a rejection there plus a 8.5 bp dip to about 81473 (just above the 81460-81468 15m closes) would liquidate the trade. Thin weekend liquidity means one sweep can move the price that far."}

---

## Evidence appendix (not part of the JSON decision)

**Question:** one BTC-USD perp trade (direction, leverage 750-1000, take-profit) for PaperTrade, which fills at the Hyperliquid mid.

### Facts (fetched from https://api.hyperliquid.xyz/info at 2026-10-17T14:58:32Z)
- l2Book (nSigFigs 5): best bid 81542, best ask 81543, mid **81542.5**. Within 10 bp of mid: bids 42.98 BTC, asks 100.98 BTC. There is a 10.8 BTC ask at 81543 and a 7.6 BTC ask at 81548.
- metaAndAssetCtxs (BTC): markPx 81542, oraclePx 81575, premium -0.000392, funding 0.0000111/h, OI 41,091 BTC, 24h notional volume $1.87B, prevDayPx 81948.
- fundingHistory: the last 4 hourly prints were each 0.0000125.
- 1m candles (60m): high 81577, low 81415. EMA9 81540.9 > EMA21 81536.4. Realized volatility of 1m log returns: 0.76 / 1.37 / 1.45 bp per minute over 15 / 30 / 60 minutes, so volatility is contracting. Median 1m high-low range over the last 30 candles: 1.5 bp.
- 5m candles (6h): high 81699, low 81415. EMA9 81526.9 > EMA21 81512.0. 15-bar rate of change +10.3 bp. 6h VWAP 81545.4 (mid is -0.4 bp from it).
- 15m candles (24h): high 82384, low 80772. EMA9 81514 < EMA21 81533 (mildly bearish). 24h VWAP 81412.6 (mid is +16 bp above it).
- recentTrades: only 10 prints over 15 s (0.009 BTC). This sample is too small to give a meaningful cumulative volume delta.

### Computed (script: test/scratch/an.py, not delivered)
- Bust distance at 750x: 10000/750 - 4.8 = 8.53 bp, so bust is at about 81472.92. Take-profit at 81542.5 x 1.0012 = 81640.35.
- Net return on margin if the take-profit fills: move after the 0.2 bp deadband is 11.8 bp; scale 0.7755; x 0.98 win fee; x 750 = **67.3%**. This formula reproduces the brief's table (10 bp @1000x gives 72%).
- Replay of the last 60 one-minute candles (each candle as an entry, 30-minute horizon, bust checked before target inside a candle, results as wins/busts/neither):
  - 750x, 10 bp target: long 19/1/41, short 0/25/36.
  - 750x, 15 bp target: long 5/1/55.
  - The 5m replay over 6h favoured shorts, which reflects the earlier decline from 81699. That is one reason confidence is moderate.

### Inferences (judgement, not fact)
- Regime: "ranging" with a short-term upward lean. Higher timeframes (15m/24h) are still down from 82384.
- With no drift, the chance of reaching +12 bp before -8.5 bp is about 8.5/20.5, or 41%, so the expected value is near zero to negative. The long relies on a modest positive drift: the 1h trend, the 1m/5m EMA alignment, and the negative premium pulling mark toward oracle. Leverage is kept at the 750x floor because low volatility makes the target slow to reach, and a wider bust gives the move more time.
- A 12 bp target is a compromise. At about 1-1.5 bp/min of volatility, 20 bp or more is unlikely within 30 minutes. Below about 10 bp, the haircut takes more than 28% of the gain.

### Uncertainty and unanswered questions
- 2026-10-17 is a Saturday, so I assumed there is no scheduled US macro release in the next hour. I did not check this against a calendar, and I did not do any news or liquidation-heatmap lookup.
- The backtest sample is tiny (60 one-minute and 72 five-minute candles) and lies inside the same window used to pick the direction, so it is in-sample and optimistic.
- The heavy ask liquidity could be spoofed or pulled. The book snapshot is from a single instant.
- Confidence 0.42 reflects a weak edge: the signals agree only on short timeframes.