# BTC-USD perpetual trade decision for PaperTrade

Data time: 2026-10-16 21:00:28 UTC (Friday). Decision: **short, 750x, take-profit 12 bp.**

## 1. Decision (the deliverable)

```json
{
  "direction": "short",
  "leverage": 750,
  "take_profit_bp": 12,
  "take_profit_price": 81515.47,
  "bust_distance_bp": 8.5333,
  "bust_price_estimate": 81683.05,
  "expected_net_return_on_margin_pct": 67.3,
  "confidence": 0.42,
  "regime": "volatile_choppy",
  "key_levels": { "support": 81480, "resistance": 81710 },
  "entry_mid": 81613.405,
  "data_time_utc": "2026-10-16T21:00:28Z",
  "reasoning": "BTC is 0.8% down on the day after a hot US CPI print and ETF outflows, and the 45-minute bounce from the 80,800 double bottom has stalled at 81,682, a lower high beneath the 81,710 spike high from 19:00 UTC, with a high-volume rejection wick, perps trading at a small discount to index on OKX and Hyperliquid, and price roughly 50 bp above the 6-hour VWAP into a thin Friday evening session. 750x is the lowest leverage allowed and gives the widest survivable adverse move (8.5 bp) against 1-minute realized volatility of roughly 3 to 4 bp per minute, while a 12 bp target sits just below the 81,540 minor support so a break of that level fills it, clears the haircut at about 75% of the raw gain, and is about one 30-minute standard deviation away. Entry mid is a Coinbase BBO proxy because the Hyperliquid POST API could not be reached from this environment; re-derive prices from the live Hyperliquid mid using the bp values.",
  "risks": "The bust price sits almost exactly at the 20:52 UTC high of 81,682, so any retest of that high, a short squeeze through 81,710 toward 81,860, or a single 1-minute candle of average-plus range liquidates the position before the target is reached."
}
```

Price arithmetic (entry_mid = (81613.40 + 81613.41) / 2 = 81613.405):

| Quantity | Formula | Value |
|---|---|---|
| Bust distance | 10000/750 − 4.8 | 8.5333 bp |
| Bust price (short) | 81613.405 × (1 + 0.00085333) | 81683.05 |
| Take-profit price (short) | 81613.405 × (1 − 0.0012) | 81515.47 |
| Haircut scale at 11.8 bp (after 0.2 bp deadband) | 0.90 / (1 + 1/(0.00118×15000) + 100000/(1e6×0.00118×814.598)) | 0.7755 |
| Net kept move | 11.8 × 0.7755 × 0.98 | 8.97 bp (74.7% of the raw 12 bp) |
| Net return on margin | 8.97 bp × 750 / 10000 | 67.3% |

## 2. What was and was not fetched (important limitation)

**Fact.** The shell sandbox in this run refused to start any command (seccomp/user-namespace error on every call), so nothing could be executed: no `curl`, no Python. The only network client available performs HTTP GET.

**Fact.** The Hyperliquid `/info` endpoint accepts only POST. A GET to `https://api.hyperliquid.xyz/info` returned HTTP 405. Therefore the Hyperliquid `candleSnapshot`, `metaAndAssetCtxs`, and `l2Book` calls specified in the brief were **not** made. The `entry_mid` above is the Coinbase Exchange BTC-USD best bid/ask mid, used as a proxy. Hyperliquid's mid at the same moment is expected to be within a few basis points of it but is not known. **Before placing the trade, replace `entry_mid` with the live Hyperliquid mid and recompute the two prices from the bp values.** The direction, leverage, and bp numbers do not depend on that substitution.

**Fact.** Binance endpoints returned HTTP 451 from this location and were not used.

Sources actually used (all public GET, fetched 20:57–21:00 UTC):

- Coinbase Exchange: 1m/5m/15m candles, level-1 and level-2 book, last 100 trades, 24h stats, server time (`api.exchange.coinbase.com/products/BTC-USD/...`, `/time`).
- OKX: `BTC-USDT-SWAP` ticker and funding (`www.okx.com/api/v5/market/ticker`, `/api/v5/public/funding-rate`).
- Deribit: `BTC-PERPETUAL` ticker (`www.deribit.com/api/v2/public/ticker`).
- CoinGecko derivatives feed for Hyperliquid BTC-USD (lagged by several minutes): price, index, basis, funding, open interest (`api.coingecko.com/api/v3/derivatives` and `/derivatives/exchanges/hyperliquid`).
- Web search for news and the US macro calendar (secondary sources, listed in section 6).

All metrics below were computed by hand from the fetched tables because no code could run. They are labelled as estimates where rounding or sampling matters.

## 3. Market facts at decision time

| Item | Value | Source |
|---|---|---|
| Coinbase best bid / ask, 21:00:28 UTC | 81613.40 (0.0094 BTC) / 81613.41 (0.1138 BTC) | Coinbase L1 book |
| Deribit BTC-PERPETUAL bid / ask / index, 21:00:34 UTC | 81627.0 / 81627.5 / index 81611.53 | Deribit ticker |
| OKX BTC-USDT-SWAP last, 20:58:10 UTC | 81674.5 (USDT-quoted, about +8 to +10 bp vs USD venues) | OKX ticker |
| Hyperliquid BTC-USD (CoinGecko, last trade 20:51:32 UTC) | price 81571.86, index 81670, basis +0.028%, funding 0.00125%/h, OI $3.369B, 24h vol $2.91B | CoinGecko |
| Coinbase 24h: open / high / low / last | 82022.8 / 82831.01 / 80693.69 / 81591.87 | Coinbase stats |
| OKX 24h: high / low / start-of-day (UTC) | 82898 / 80770.7 / 82192.9 | OKX ticker |
| Deribit 24h price change | −0.51% | Deribit ticker |
| OKX funding (8h) / premium | +0.01% / −0.036% (perp below index) | OKX funding |
| Deribit funding 8h | +0.0034% (neutral) | Deribit ticker |
| Hyperliquid OI change between two snapshots ~10 min apart | $3.374B → $3.369B (−0.15%) | CoinGecko |

Session context (fact): Friday 21:00 UTC. US cash equities closed at 20:00 UTC. The only US releases today were import/export prices (12:30 UTC) and industrial production (13:15 UTC), both long past. No scheduled US macro release falls in the next hour. Hyperliquid funding accrues hourly; the next accrual is at 22:00 UTC and does not affect the mid.

## 4. Price structure (facts from Coinbase candles, Unix times converted to UTC)

- 24h: BTC opened the UTC day near 82,200, printed the high 82,831 around 11:15 UTC, and sold off to 80,694 at about 17:15 UTC, then 80,797 at about 18:45 UTC (double bottom, 5m candles with 190 and 142 BTC of volume, the heaviest of the day).
- From 18:45 to 20:52 UTC price rallied to 81,681.8 (+108 bp off the low). The 20:52 1m candle (4.57 BTC, large for the hour) spiked to 81,681.8 and closed at 81,668; the next candle faded to 81,654; by 20:58 price traded 81,592 and by 21:00 81,613.
- Lower-high sequence on 15m closes/highs since the day high: 82,831 → 82,384 → 82,315 → 81,879 → 81,710 (19:00 UTC spike) → 81,682 (20:52). The current bounce has not exceeded the 19:00 UTC spike high.
- Position in the 24h range: (81,613 − 80,694) / (82,831 − 80,694) ≈ 43%.
- Minor support: 81,542 (three 1m lows at 81,542.3–81,542.4 between 20:35 and 20:47 UTC). Stronger support: 81,465–81,483 (5m lows 20:15–20:30 UTC) and the round 81,500. Below that 81,400 and 81,300.
- Resistance: 81,682 (20:52 high), 81,710 (19:00 spike high), then 81,860–81,890 (15m candle at 15:00 UTC) and 81,986.

## 5. Volatility and excursion metrics vs. bust distance (hand-computed estimates)

Bust distances: 1000x → 5.2 bp, 900x → 6.3 bp, 800x → 7.7 bp, 750x → 8.53 bp.

- **1m realized volatility (std of close-to-close moves):** roughly 3.3 bp/min over the last 20 minutes, roughly 4 bp/min over the last 60 minutes. Volatility is **contracting**: 1m ranges were 5–15 bp between 19:49 and 20:10 UTC and 2–8 bp between 20:30 and 20:53 UTC. Estimate; computed by hand from 65 candles.
- **1m candle range (high − low):** mean ≈ 5.2 bp, median ≈ 5 bp over the last 65 candles. About half of 1m candles had a range at least as large as the 1000x bust distance; about 6 of 65 (≈9%) had a range at least as large as the 750x bust distance (9.8, 8.7, 13.6, 10.6, 14.9, 8.7 bp).
- **Per-candle adverse excursion for a short (high − open):** in the last 65 minutes about 5 candles exceeded 8.5 bp (14.3, 9.9, 8.7, 8.2, 8.2 bp). For a long (open − low): about 1–2 candles (13.3, 8.2 bp). This asymmetry reflects the rally that just happened, not a forecast.
- **5m candle range:** 5–18 bp in the last hour (largest: 81,599 to 81,483 = 14 bp at 20:00; 81,559 to 81,432 = 16 bp at 19:50). Every 5m candle in the last two hours has a range larger than the 750x bust distance, so holding through any 5m window carries material bust risk regardless of direction.
- **Random-walk reference:** with σ ≈ 3.3 bp/min, a 30-minute standard deviation is about 18 bp. A 12 bp target is about 0.7σ at 30 minutes, reachable; an 8.5 bp bust is about 0.5σ.
- **Driftless probability of take-profit before bust:** bust / (bust + TP) = 8.53 / 20.53 ≈ 41.5%. Break-even win probability given a +67.3% / −100% payoff is 1 / 1.673 ≈ 59.8%. **Inference:** the trade has negative expectancy unless the directional read is right well over half the time. This is structural for every allowed leverage and target: with keep ≤ 0.90 and bust × leverage/10000 = 1 − 0.00048 × leverage, expected value under zero drift is always negative. Larger targets reduce the edge required (at 20 bp the break-even is ≈45.6% vs 29.9% driftless; at 12 bp it is 59.8% vs 41.5%), but a 20 bp move is a full 30-minute standard deviation in current contracting volatility.
- **Session VWAP (6h):** estimated 81,150–81,250 (the heaviest volume printed at 80,800–81,300 between 18:45 and 19:15 UTC). Price is ≈ 45–55 bp above it. Estimate.
- **EMAs:** price at 21:00 is slightly below the 1m EMA(9) (≈81,615–81,620) and above the 5m EMA(20) (≈81,530). 1m momentum has rolled over; 5m uptrend intact; 15m and 1h trend down. Timeframes disagree, hence `regime: volatile_choppy`.
- **Order flow:** last 100 Coinbase trades (20:57:47–20:58:10 UTC, 23 seconds): maker-buy (taker sell) 0.271 BTC vs maker-sell (taker buy) 0.035 BTC, average trade 0.003 BTC. Taker selling dominated, but the sample is tiny. Level-1 book at 21:00: ask size 12x bid size at the touch (0.114 vs 0.009 BTC), a mild resting-sell signal, also tiny.
- **Derivatives positioning:** OKX premium −3.6 bp and the Hyperliquid perp trading below its index (CoinGecko basis +0.028%) mean perps are not leading the bounce. Deribit's perp mark is ~15 bp above its index, which partly contradicts this. Funding is neutral on all three venues. Open interest on Hyperliquid was flat (−0.15%) across the bounce, i.e. the rally was not driven by new longs piling in.

## 6. Reasoning: why short, why 750x, why 12 bp

**Inference (direction).** Evidence for short: daily downtrend with intact lower highs; macro backdrop is bearish (hotter-than-expected US CPI on 14 Oct, roughly $434M of crypto fund outflows reported by CoinShares, Fed hike odds rising); high-volume rejection at 81,682 under the 81,710 lower high; price extended ~50 bp above the 6h VWAP into a thin Friday evening session; perps at a small discount to index; latest prints taker-sell heavy. Evidence for long: 45-minute momentum of +43 bp/hour, higher 5m lows, dip from 81,592 bought back to 81,613 within two minutes, flat OI (no crowded long to squeeze out). I judge the short case slightly stronger, hence confidence 0.42 (below even odds relative to the 59.8% break-even).

**Inference (leverage).** 750x is the minimum allowed and gives the widest survivable adverse move (8.5 bp). With median 1m ranges near 5 bp and 9% of 1m candles exceeding 8.5 bp, 900–1000x (5.2–6.3 bp bust) would be liquidated by routine noise within minutes; the extra return at 1000x (+33% more on margin) does not compensate for roughly doubling the per-minute bust frequency.

**Inference (target).** 12 bp puts the target at 81,515, just below the 81,542 minor support and above the 81,465–81,500 cluster, so a break of 81,542 should fill it before buyers at 81,480 react. It keeps ≈75% of the raw gain after the haircut (67% on margin), versus 69% kept at 8 bp (41% on margin) and 80% at 20 bp (119% on margin, but twice as far in a contracting tape).

## 7. Risks and what would cause liquidation

- Bust price 81,683 is essentially the 20:52 high (81,681.8). Any retest of that high ends the trade. This is the single most likely failure.
- A squeeze through 81,710 would open 81,860–81,890; shorts entered against the lower high would be the fuel.
- Late-Friday thin books make 8–15 bp 1m candles more likely in both directions.
- Hyperliquid mid can deviate from the Coinbase proxy by several bp; the actual bust and target must be recomputed from the live Hyperliquid mid.

## 8. Uncertainty and unanswered questions

- **Not fetched:** Hyperliquid candles, `metaAndAssetCtxs` (markPx, oraclePx, premium, funding, OI, dayNtlVlm), `l2Book`, `recentTrades`, `fundingHistory`. Hyperliquid book imbalance within 10 bp of mid and resting liquidity between entry and target/bust are unknown.
- **Estimates:** all volatility, range, VWAP, and EMA figures were computed by hand from the fetched tables, without code. Expect ±10–20% error on the volatility and VWAP numbers; the candle highs/lows and prices are exact as fetched.
- **Lag:** CoinGecko's Hyperliquid snapshot was 9 minutes old at decision time. Two snapshots disagreed on the Hyperliquid–Coinbase spread (+10 bp vs −6 bp), so the direction of the Hyperliquid premium is unresolved.
- **Macro calendar:** entries come from fedratecalc.com and smartcalendars.ai, not BLS/Fed pages. No release in the next hour was found; a verification against bls.gov was not possible here.
- **News:** the "hot CPI / ETF outflow / hike odds" narrative is from search snippets (StreetInsider, CoinDesk/Decrypt items via search), not full articles.
- **Open question for the user:** what is the live Hyperliquid mid at placement time? The bp values stand; the two price fields must be recomputed from it.

## 9. Sources

- Coinbase Exchange API: https://api.exchange.coinbase.com/products/BTC-USD/candles?granularity=60 (and 300, 900); https://api.exchange.coinbase.com/products/BTC-USD/book?level=1; .../book?level=2; .../trades?limit=100; .../stats; https://api.exchange.coinbase.com/time
- OKX API: https://www.okx.com/api/v5/market/ticker?instId=BTC-USDT-SWAP ; https://www.okx.com/api/v5/public/funding-rate?instId=BTC-USDT-SWAP
- Deribit API: https://www.deribit.com/api/v2/public/ticker?instrument_name=BTC-PERPETUAL
- CoinGecko: https://api.coingecko.com/api/v3/derivatives/exchanges/hyperliquid?include_tickers=unexpired ; https://api.coingecko.com/api/v3/derivatives
- Hyperliquid (attempted, HTTP 405 on GET): https://api.hyperliquid.xyz/info
- Binance (attempted, HTTP 451): https://api.binance.com/api/v3/klines?symbol=BTCUSDT&interval=1m&limit=60
- Macro calendar (secondary): https://fedratecalc.com/us-economic-calendar/october-2026/ ; https://smartcalendars.ai/en/feeds/us-economic-calendar
- News (secondary, via web search): https://www.coindesk.com/tag/bitcoin ; https://www.streetinsider.com/MarketMediaWire/Bitcoin+News+Today%3A+BTC+Recovers+Above+%2482K+as+Remittix%E2%80%99s+November+Debut+Adds+a+New+Payments+Catalyst/27196921.html (promotional, used only for the price context)