Sixty-two transparent research strategies for Pacifica perpetual markets:

• 10 directional, relative-value, funding, and order-flow baselines
• 15 grid variants with reproducible Pacifica historical simulations
• 7 review-only copy-trading policies sourcing Hyperliquid, Pacifica, or both
• 30 independently implemented TradingView-origin research translations

The repository publishes the research code used by Neptune Strategy Desk. The strategies, indicators, public-data adapters, cost model, funding treatment, and result generation are dependency-free JavaScript so they can be inspected and reproduced easily.

Research and proposal generation only. This repository does not place, sign, cancel, or modify orders. Copy and grid outputs must be validated against live Pacifica data, the Neptune-bound account, margin, precision, and risk limits before a user approves execution.

These are close-rebalanced target-position grids. A grid level crossed at an hourly close changes the target on the following bar. This deliberately avoids pretending that mark candles can reconstruct passive queue fills.

The adapters use documented public endpoints:

• Hyperliquid POST https://api.hyperliquid.xyz/info with clearinghouseState and userFillsByTime
• Pacifica GET /api/v1/positions, GET /api/v1/account, and GET /api/v1/trades/history

Leader reads require public account addresses only. They never require a leader private key. Proposal execution is intentionally absent; Neptune must use the currently bound Pacifica account and managed signing after explicit approval.

These entries retain TradingView as their source, link to the original author, and are backed by separate JavaScript implementations for Pacifica data. No Pine source is redistributed. See TRADINGVIEW_ATTRIBUTION.md for source-by-source attribution and the important translation caveat.

These numbers are not TradingView Strategy Tester results. They use the same Pacifica mark-to-mark engine as the other catalog entries: hourly bars, actual Pacifica funding, next-bar targets, and 5 bps per 1x one-way turnover. The translation intentionally keeps losing results visible.

The committed result uses a fixed, reproducible window:

• Environment: Pacifica mainnet
• Window: 2026-01-01 00:00 UTC through 2026-06-30 23:00 UTC
• Interval: 1 hour
• Aligned bars: 4,344
• Markets: BTC, ETH, SOL
• Execution: signal at close, target active on the following bar
• Cost stress: 5 bps per 1x one-way turnover
• Funding: actual Pacifica hourly historical funding rates
• Exposure: maximum 1x gross, without compounding leverage

Grid results are intentionally published even when negative:

Copy policies do not receive a performance backtest without a selected leader and a complete, timestamped leader history. Their published evidence is adapter normalization, direction handling, staleness checks, scaling, gross caps, per-symbol caps, and consensus tests. This avoids presenting an arbitrary leader as representative.

Requirements: Node.js 22 or newer and outbound access to the Pacifica public API.

npm test
npm run backtest

npm run backtest downloads mark-price candles and funding history directly from Pacifica, runs the 55 historically testable strategies, prints a summary, and rewrites results/backtests.json.

No account, wallet, API key, private key, or environment file is required for the test suite or historical simulations.

import { fetchHyperliquidSnapshot } from "./src/copy/adapters.mjs";
import { buildPositionMirrorProposal } from "./src/copy/policies.mjs";

const snapshot = await fetchHyperliquidSnapshot({
  account: "0xLEADER_ACCOUNT",
});

const proposal = buildPositionMirrorProposal({
  snapshot,
  followerEquityUsd: 1_000,
  copyRatio: 0.25,
  maxGross: 0.75,
  maxSymbolWeight: 0.25,
  symbolMap: { BTC: "BTC", ETH: "ETH", SOL: "SOL" },
});

console.log(proposal); // review-only; no order is sent

src/engine.mjs              Pacifica data client, indicators, and P&L engine
src/strategies.mjs          Original signals plus grid catalog export
src/grid/strategies.mjs     Fifteen grid target-position implementations
src/copy/adapters.mjs       Hyperliquid and Pacifica read-only source adapters
src/copy/policies.mjs       Position, fill, and consensus copy policies
src/copy/catalog.mjs        Seven copy-trading definitions
src/tradingview/            Thirty attributed clean-room research translations
src/catalog.mjs             Complete 62-strategy catalog
scripts/run-backtests.mjs   Reproducible result generator
results/backtests.json      Committed metrics and weekly equity curves
test/                       Catalog, adapter, and policy tests

• Hourly grid simulations do not reconstruct order-book queue position, passive fill probability, latency, or intrabar path ordering.
• Copy trading introduces source latency, symbol-mapping, leader liquidation, fee, leverage, and follower slippage differences.
• Hyperliquid exposes only a bounded recent fill history through the standard info API; Pacifica history is cursor-paginated. Completeness must be checked.
• Public leader performance can be survivorship-biased and may include deposits, withdrawals, subaccounts, hedges, or off-platform positions.
• A fixed historical window is not walk-forward validation and does not predict future profitability.
• TradingView-origin translations can differ materially from their Pine references because source code is not redistributed and Pacifica uses a different market, dataset, funding stream, execution model, and cost model.

No license is granted at this time. The repository is public for transparency and review; public visibility alone does not grant permission to copy, modify, or redistribute the code.