# PaperTrade decision report: NOT PRODUCED (data fetch failed)

**Status:** no trade decision. I did not output the required JSON trade object on purpose. Every field in it (entry_mid, take_profit_price, bust_price_estimate, data_time_utc, and the market/direction choice) has to come from live Hyperliquid data. Filling those fields without that data would mean making up numbers, and a made-up trade could liquidate real user margin.

## Facts (what was observed on this machine, 2026-10-21)
1. The Hyperliquid info API (`https://api.hyperliquid.xyz/info`) only accepts HTTP POST with a JSON body.
2. The only HTTP tool that worked in this run (WebFetch) sends GET requests. A GET to that endpoint returned **HTTP 405 Method Not Allowed**.
3. The shell tool, which would have sent the POST requests via curl or a script, failed before running any command. It failed every time, even on `echo hi`, with this error:
   `apply-seccomp: write /proc/self/setgroups (nested userns is capability-restricted; caller must provide CAP_SYS_ADMIN): Permission denied`
4. So none of the required inputs could be fetched for BTC or ETH: candleSnapshot (1m/5m/15m), metaAndAssetCtxs, l2Book, fundingHistory and recentTrades.

## Inferences
- This is a problem with the run environment (the sandbox can't start a shell), not with the requester's brief, and not something only the requester can supply.
- Prices from other exchanges, such as Binance's GET klines, are not used as stand-ins. PaperTrade fills and liquidates at the Hyperliquid BBO mid. Bust distances here are only 5.2–8.5 bp, so even a small price gap between venues would make entry_mid, bust_price_estimate and take_profit_price wrong.

## What is still valid without data (taken directly from the brief's formulas)
- Bust distance = 10000 / leverage − 4.8 bp: 1000x → 5.2 bp, 900x → 6.3 bp, 800x → 7.7 bp, 750x → 8.5 bp.
- Tiny take-profits lose most of the gain to the haircut. At 8 bp BTC keeps about 69% of the gain (42% return on margin at 750x); at 15 bp it keeps about 77% (87% at 750x). A sensible target probably sits around 12–20 bp. Whether it is reachable before an adverse move the size of the bust distance can't be judged without live volatility data.

## Uncertainty / unanswered questions
- Which market (BTC or ETH), which direction, and what leverage and target: **unanswered**. These depend on live data.
- Regime, key levels, order-flow imbalance, funding and OI context: **unanswered**.

## To complete this task
Run it again in an environment where the shell sandbox can start, or where an HTTP tool can send POST requests to `api.hyperliquid.xyz`. The analysis and the JSON output can then be produced exactly as the brief specifies.