bsm.guru is a local-first screener that uses Black-Scholes-Merton-inspired lognormal assumptions as a baseline to quantify how unusual a perp move is vs realized volatility. The UI ranks Hyperliquid markets, suggests a small set of positions, and can optionally auto-enter/auto-exit based on your thresholds.
A volatility-normalized perp dashboard + trading helper for Hyperliquid. It syncs live mids and candles, computes realized volatility, and turns the last 24h move into a simple "how many sigmas was that?" score.
Not a hosted SaaS, not financial advice, and not a guarantee of edge. The "BSM" here is a baseline model (lognormal + constant vol) used to measure surprise, not to claim the market is "wrong."
The app is designed to run on a machine you control. Wallet keystores live on disk and trading routes are restricted to localhost by default.
./run.shpnpm lint
pnpm test
pnpm build1) Create a wallet in the app (stored on disk). 2) Fund the wallet on Arbitrum with USDC. 3) Deposit USDC into Hyperliquid from the wallet page. 4) Use the dashboard to enter positions (or enable auto-enter/auto-exit with conservative thresholds).
The app periodically pulls Hyperliquid market data and candles, and stores derived metrics in a local SQLite database. This powers fast UI refreshes and consistent calculations.
If you want to run the tool from source, clone the repo and use the included runner script.
git clone https://github.com/pierce403/bsm.guru.git
cd bsm.guru
./run.shIf you prefer, you can also run the usual Next.js commands (pnpm dev), but ./run.sh handles the local sync loop.
Black-Scholes-Merton assumes a lognormal price process with constant volatility. Crypto is not constant-vol and not perfectly lognormal - but the model still gives a convenient unit: "sigmas."
Markets move around. A 3% move is boring for some assets and extreme for others. If you normalize moves by realized volatility, you can compare "how unusual" different markets are on the same scale, then decide whether to fade the move, follow momentum, or avoid it.
These are computed locally from Hyperliquid candles and market context. They are intentionally simple, inspectable, and fast.
| Metric | Meaning | How it is computed |
|---|---|---|
| Realized σ | Annualized realized volatility estimated from recent log-returns. | Take log returns from candle closes, compute sample stdev per period, then annualize by sqrt(yearSeconds / periodSeconds). |
| Sigma move (24h) | How many "sigmas" the last 24h move represents. | log(mid / prevDayPx) / (sigma * sqrt(1/365)). Positive is up, negative is down. |
| Tail p (24h) | Two-sided normal tail probability for that sigma move (smaller = rarer). | 2 * (1 - Phi(|sigmaMove|)), where Phi is the standard normal CDF. |
| Position health | A "still good?" score based on how the market's current sigma move compares to what you entered on. It can go down even if PnL is up. | The app tracks entry context and recomputes the current sigma move. Health is tied to the persistence of the original signal, not just profit. |
Auto trading is deliberately constrained: the app exits first, then enters up to the cap.
By default, the app keeps a maximum of 5 concurrent positions and requires your thresholds to be met at refresh time. Use conservative settings first and verify behavior with small notionals.
After each refresh cycle, auto-exit evaluates open positions against your exit threshold. Auto-enter looks at the current ranking and enters the best eligible position(s) until the cap is reached.
This tool touches real money if you enable trading. Treat it like a small trading system, not a dashboard toy.
The sigma metrics assume normality of log returns and a stable realized vol estimate. Crypto frequently violates both assumptions (fat tails, regime shifts, jumps). Use these metrics as context, not certainty.
The app writes human-readable logs under ./logs/ (trade, wallet, analysis, errors). Private keys are not logged. Every trade action aims to leave a proof link via Hypurrscan / Dexly for external verification.
Contributions are welcome. If you have ideas for better signals, safer execution, or clearer UX, please jump in.
Improve signal definitions, add risk controls (cooldowns, max daily trades), add better observability (dashboards, structured logs), and expand Playwright coverage for wallet and trading flows.
Open an issue describing the change, then submit a PR. Please keep commits small, run the quality gates, and include a short note on how you tested.
pnpm lint
pnpm test
pnpm build
pnpm playwright test