# BTC-USD perp trade decision (PaperTrade / Hyperliquid mid)

## Decision

```json
{"direction":"long","leverage":750,"take_profit_bp":20,"take_profit_price":81702.58,"bust_distance_bp":8.53,"bust_price_estimate":81469.92,"expected_net_return_on_margin_pct":119.5,"confidence":0.52,"regime":"volatile_choppy","key_levels":{"support":81478,"resistance":81729},"entry_mid":81539.5,"data_time_utc":"2026-10-16T20:14:30Z","reasoning":"Price is pulling back to the rising 5m EMA9 (81505) inside a 6h recovery off the 80772 low (5m EMA9 > EMA21, +52 bp over 75 min), and two 1m candles bottomed at 81530 and held. 1m realized vol is about 5-6 bp/min, so a 20 bp target (81703, below the 1h high of 81729) can be reached in under 30 minutes. 750x is the lowest leverage allowed and gives the widest bust distance (8.5 bp). In a driftless model, bust_distance x leverage x keep is largest at 750x, and a 20 bp move keeps about 80% of the raw gain after the haircut, against about 49-61% for 3-5 bp targets. In the last hour of 1m data, a long at 750x/20 bp hit the target before liquidation 30 times and was liquidated 27 times; a short won only 7 times and was liquidated 41 times.","risks":"The 1m trend is down (-12 bp over 5 min, price below the 1m EMAs), sells dominate the latest prints, the book is ask-heavy within 10 bp, and the perp trades at a -3.4 bp premium to the oracle. If 81478-81530 breaks, the 8.5 bp bust (81470) is roughly 1.5 one-minute standard deviations away."}
```

## Evidence (facts — Hyperliquid public API `https://api.hyperliquid.xyz/info`, fetched 2026-10-16 20:13:51Z and refreshed 20:14:30Z)

| Item | Value | Source query |
|---|---|---|
| Best bid / ask / mid | 81539 / 81540 / **81539.5** | `l2Book` nSigFigs 5 |
| markPx / oraclePx / premium | 81537 / 81568 / −0.000343 (−3.4 bp) | `metaAndAssetCtxs` |
| Funding (hourly) | 0.0000125, stable over 24h | `metaAndAssetCtxs`, `fundingHistory` |
| Open interest | 41,280.7 BTC (flat between the two fetches) | `metaAndAssetCtxs` |
| 24h notional volume | ≈ $2.90B; prevDayPx 82289 | `metaAndAssetCtxs` |
| 24h high / low (15m) | 82898 / 80772 → mid sits at 36% of range | `candleSnapshot` 15m |
| 24h VWAP (15m typical price) | 81769 → mid is −28 bp below | computed |
| 6h high / low (5m) | 82479 / 80772 | `candleSnapshot` 5m |
| 1h high / low (1m) | 81729 / 81278 | `candleSnapshot` 1m |
| 1m EMA9 / EMA21 | 81582 / 81566 (price below both) | computed |
| 5m EMA9 / EMA21 | 81505 / 81420 (price above both, bullish stack) | computed |
| 15m EMA9 / EMA21 | 81410 / 81486 (bearish stack, price above both) | computed |
| ROC | 1m: −12 bp / 5 bars; 5m: +52 bp / 15 bars; 15m: −11 bp / 15 bars | computed |
| 1m realized vol per minute | 6.0 (15m), 5.0 (30m), 5.1 (60m) bp, so expanding a little | computed |
| Mean 1m excursion from open | up 3.2 bp, down 2.6 bp; mean 1m range 5.9 bp; mean 5m range 21.4 bp | computed |
| Book within 10 bp | bids 58 BTC vs asks 74 BTC (asks heavier) | `l2Book` |
| Latest 10 prints | 1.14 BTC sold vs 0.006 BTC bought over 2.7 s (very small sample) | `recentTrades` |

### First-passage backtest (last 61 one-minute candles, each start bar, 30-bar horizon, the adverse move is checked first within each candle, so this is conservative)

| Lev | Bust bp | TP bp | Long win/bust | Short win/bust |
|---|---|---|---|---|
| 750 | 8.53 | 10 | 33/26 | 22/36 |
| 750 | 8.53 | 20 | 30/27 | 7/41 |
| 1000 | 5.2 | 20 | 24/37 | 6/45 |

On 5m bars over 6h the same test is worse for longs (16/54 at 750x/20 bp) and better for shorts (32/39). A whole 5m bar's range is about 21 bp, so this coarse test overstates busts at these bust distances. It also shows that the 6h path contained a large drawdown.

### Payoff (from the brief's formula)
For the 20 bp target: move after the 0.2 bp deadband = 19.8 bp, and scale × 0.98 = 0.805. Return on margin = 0.00198 × 750 × 0.805 = **119.5%**. A liquidation loses −100%. The trade breaks even at a win rate of about 45.6%.

## Inferences (judgement, not fact)
- **Direction:** I chose long. The 5m structure is recovering, with higher lows since 80772, and the 81530 level was defended twice. The 1h backtest was strongly asymmetric in favour of longs. This is a pullback-continuation bet against a bearish 1m reading.
- **Leverage:** With no drift, EV ∝ (bust × lev × keep − 100·…). Since bust × lev = 10000 − 4.8·lev, the lowest leverage gives the best odds. 750x is the floor.
- **Target:** I picked 20 bp rather than 10 bp. At 10 bp the trade keeps only 72% of the gain and the payoff (≈54%) needs a win rate of about 65%. At 20 bp the payoff needs about 46%, and the 1h sample showed ≈53% for longs. 20 bp also stays below the 81729 resistance.
- **Confidence 0.52:** The signals conflict: the 5m trend is up, while 1m momentum, order flow, book skew and the negative premium all lean down.

## Uncertainty and unanswered questions
- I did not check news, the macro calendar, liquidation heatmaps or order flow on other exchanges, because of time limits. It is 20:14 UTC on a Friday, after the US cash close. I did not verify any scheduled events.
- The backtests are in-sample, cover only 61 one-minute bars, and assume an ordering of high and low inside each candle. They are not predictive guarantees.
- The trade data is a 10-print snapshot, which is too small to measure CVD meaningfully.
- The open-interest change could not be measured beyond the 40-second gap between the two fetches.
- The price will have moved after 20:14:30Z. The TP and bust prices above are fixed to entry_mid 81539.5.