sobres

Version — · Python 3.11+ · MIT

One-stop equity analysis, from the terminal.

Market data, Markowitz optimization, walk-forward backtests, saved portfolios and a web UI — one sobres command line, one SQLite file, no API key required.

Install Source on GitHub
Efficient frontier drawn from a real sobres optimize frontier run.

The product, in its own voice

A recorded transcript of the command below actually running — regenerated by site/scripts/record_figures.py, never written by hand.

What ships today

  1. Data, cached

    Prices, macro series, Fama-French factors and exchange rates from keyless providers where possible. Every observation is cached in one SQLite file, and every output says where it came from.

    sobres data prices AAPL MSFT --start 2020-01-01

  2. Optimize

    Min-variance, max-Sharpe, target return or risk, risk parity — with Ledoit-Wolf shrinkage on by default and a concentration warning when a weight runs away.

    sobres optimize markowitz --tickers AAPL MSFT NVDA --fill drop

  3. Backtest, walk-forward

    Re-solve at each rebalance on prior data only, let weights drift between them, charge 10 bps a side, and compare against equal weight on the same schedule.

    sobres optimize backtest --portfolio core --lookback 1y --rebalance quarterly

  4. Remember

    Named portfolios and watchlists, a run history with resolved parameters, and sobres run diff to see exactly what changed between two answers.

    sobres portfolio save core --tickers AAPL MSFT NVDA --weights 0.4 0.4 0.2

  5. Web UI and API

    Every command is also a form and an HTTP route, generated from the same declaration. The form shows the command line it runs; long jobs report real progress.

    sobres open

  6. One container

    The CLI is the image's entrypoint. Serve the UI, or run any command against the same volume.

    docker run -v sobres:/data aisolutionslab/sobres db info

  7. Factor analysis

    CAPM and Fama-French 3 and 5 factors plus momentum, with Newey-West errors and alpha reported next to its t-statistic, never alone.

    sobres analyze factors NVDA --model ff5

  8. Goal planning

    Retirement and FIRE, house, car and education goals; every answer carries a Monte Carlo or bootstrap success probability and the 10th–90th percentile outcomes, seed printed.

    sobres plan retire --income 120000 --expenses 60000 --simulate

  9. Econometrics

    A joint VAR/BVAR price forecast scored against no-change on held-out dates and always carrying 80% and 95% bounds; stationarity diagnostics with disagreement stated; GARCH volatility with simulated bands; robust regression with VIF.

    sobres econ forecast ticker:AAPL --horizon 20

  10. Exchange rates and PPP

    Convert once, explicitly; split a return into asset and currency; price a hedge; compare market rates with purchasing-power rates — labelled as a valuation gap, never a forecast.

    sobres ppp compare --base USD --vs EUR MXN

The number that matters is the second one

Optimize on a window, evaluate on the same window, and the Sharpe ratio looks like this:

— —

In-sample max-Sharpe versus walk-forward on the same assets and window; costs charged, benchmark on the same schedule.

sobres labels every in-sample result as in-sample, reports alpha with its t-statistic, turns shrinkage and transaction costs on by default, and says where survivorship bias enters. Defaults that flatter results are bugs.

Install

pip install sobres
docker run -p 8787:8787 -v sobres:/data aisolutionslab/sobres serve --host 0.0.0.0

Then sobres init, sobres doctor, and you are three commands in. Works with no API key; a FRED key unlocks macro series.

Roadmap — not yet available

Specified in the repository, not shipped in the version above: