Plan a rebalance

Setup. The scripts sit beside this file. Run them from a working folder that holds your .env (E*TRADE consumer key/secret) and a data/ directory; install deps with pip install -r requirements.txt where one is included. Part of a set: etrade-pull, order-check, portfolio-review, realized-pnl, rebalance-plan, price-history, trade-journal, intrinsic-value, sec-financials, buffett-checklist.

cp targets.example.json targets.json     # then edit
python rebalance.py
python rebalance.py --dca 10 --start 2026-06-08

Needs data/portfolio.csv — see etrade-pull. Reached from portfolio-review once the decision is already made; this skill executes a decision, it doesn't reach one.

How the weights are defined

Weights are fractions of investable capital:

(market value of positions) + cash_on_hand - cash_reserve

A reserve you intend to keep is excluded from the denominator rather than quietly diluting every target. 0.15 means 15% — the script rejects a target set summing over 100% rather than producing a plan that can't be executed.

A name set to 0.0 is a full exit. sell_unlisted: true also exits every held name absent from targets.

Rules

• This prints an order list. It never places an order, and neither do you. The person types the orders. That gap is the last place a wrong premise can be caught before it costs money, and it's the reason nothing in this repo holds trading permissions.

• targets.json is gitignored — keep it that way. It is a complete statement of what someone intends to own, which is as sensitive as the positions themselves.

• The tax tag cannot see holding periods. It reads unrealized P&L, so it knows "harvest loss" from "realizes gain" and nothing more. It cannot tell short-term from long-term, and that distinction is often larger than the trade's edge. Check the lot dates before acting on any row tagged realizes gain.

• A name you don't hold needs a price in the config. There's no market value to derive one from, so the shares column comes back empty and the script says which names it couldn't size. It reports the gap rather than guessing, because a guessed price produces a share count that looks authoritative.

• Share counts are rounded; the dollar figures are the plan. For high-priced names a DCA day may be one or two shares. Keep the dollar pace.

• Re-pull positions immediately before generating a plan. Deltas computed from a stale CSV are wrong by however much the market moved.

Staging

--dca N --start YYYY-MM-DD splits the buy side evenly across N trading days, skipping weekends and any date in the config's holidays list. Sells are untouched — they're assumed to clear first and fund the buys. Add market holidays yourself; the script has no calendar and will happily schedule a purchase on Juneteenth if you don't list it.