- portfolio._rebalance_dates: Grouper labels groups with period-END
timestamps (Sundays, month/quarter/year ends) that never exist in a
trading-day index, so 1W/1ME/QE/YE silently never rebalanced (only
1D and buy-and-hold worked). Now takes the first trading day of each
group.
- metrics.beta_alpha: guard var(bench) == 0 (flat benchmark) instead of
dividing by zero.
- data._read_csv_clean: the downloader has written Go nil tokens
('%!f(<nil>)') into price cells (89 symbols, one bad day each), which
made pandas infer the whole column as string and crash every
pct_change on any portfolio containing them. Coerce all columns to
numeric (nil tokens -> NaN); parquet cache rebuilt (0 string cols).
- new tests/test_portfolio.py (18 checks: rebalance dates land on
trading days and actually fire, flat-benchmark beta, nil-token
columns), wired into run_tests.sh (now 5 sections).
- full suite: 30 + 22 + 18 + 38 AppTest + 12 browser e2e, all passing
The old model valued holdings in ADJ (total-return index) units but
computed distribution flows as raw per-share dollars — so every
distribution, and its tax, was overstated by the raw/adj ratio
(JLPSX: 30.10/11.25 = 2.7x; the Dec-2020 cap-gain tax showed as 12.1%
of the account instead of the true 4.5%). The wiggle in the after-tax
curve was this bug, not a convention issue.
tax.py now:
- values holdings in RAW share units on close prices (bundle.close);
- receives the per-share distribution on its event date, pays the tax
(recorded in TaxResult.taxes), and reinvests the after-tax remainder
at the same day's raw close — the tax's effect lives in the (smaller)
reinvested units and is NOT also deducted from cash (double-count
caught and fixed in review);
- recomputes market value after the reinvestment so equity[t] is the
post-event liquidation value.
With the fix, the 'as-if-liquidated' equity on JLPSX's ex-div day drops
by exactly the true tax cost (4.72% vs 12.1% before); the -22.9% price
drop is offset by the distribution kept.
Also:
- app.py passes bundle.close to the after-tax model (pre-tax
portfolio_returns still uses adj);
- JLPSX/JLPYX: the 2020-12-11 6.824 capital-gain distribution is moved
to the true ex-div date 2020-12-14 (remove/add correction ops), so
the reinvestment prices at the post-drop close;
- tests/test_tax.py: 8 synthetic regression tests (tax magnitude,
reinvestment MV, no double-count, ex-div equity step, per-component
rates); run_tests.sh now runs it.
The cache now tracks every file in the data dir (mtime_ns + size) in
.cache/manifest.json. On load, a directory scan is compared against the
manifest:
- changed/added files are re-read and merged into the parquet panels
(one read + one concat + one write per touched panel; new values
win where present, old values kept where the new file is short)
- removed files drop their symbols (and names)
- an up-to-date cache is a ~30 ms memo hit
Measured on the real 4k-symbol set: full build 54 s, refresh of
5 modified + 1 added + 1 removed files 3.4 s. No scan TTL (a scan is
a few ms); a previous 5 s scan cache masked data updates.
Tests: tests/test_data.py (11 checks) added as step 1 of run_tests.sh.
- single spec grammar for symbol and benchmark fields: commas join one
portfolio (MSFT:0.6,V:0.4), spaces separate distinct symbols/portfolios;
both fields accept one or many entries
- benchmarks simulated with the same scheme/cost/tax rules; per-benchmark
beta/alpha columns; after-tax benchmark curves
- global Curve mode (pre/after/both) above the tabs; clean names in
single-curve mode
- live updates: field commits on Enter/blur, page recomputes per rerun;
portfolio+tax sims cached (st.cache_data); plotly.js from CDN (4.6MB ->
browser-cached) with F_INLINE_PLOTLY=1 offline fallback
- chart: legend underneath, solid lines, pan sticks to data edges
(width-preserving), zoom edge-clamped
- inputs persist in settings.json across reloads/restarts/devices
- tests: tests/test_app.py (AppTest) + tests/test_e2e_browser.py
(Playwright) via ./run_tests.sh
