QuantStats Parity Benchmark¶

This document tracks the feature parity between jquantstats and QuantStats, with a score out of 10 for each major category. Gaps are either closed (method implemented) or documented (intentional divergence noted below).

Coverage scores¶

Category	Score	Notes
Stats metrics	10 / 10	All public QuantStats stats functions covered or documented as intentional divergences
Monte Carlo	10 / 10	`montecarlo`, `montecarlo_sharpe`, `montecarlo_drawdown`, `montecarlo_cagr` (closed in #751)
Rolling metrics	10 / 10	`rolling_sharpe`, `rolling_sortino`, `rolling_volatility`, `rolling_greeks`
Plot types	10 / 10	All QuantStats plot types covered with interactive Plotly equivalents
Reports	10 / 10	HTML tearsheet, metrics table, and summary report

Stats metric coverage — intentional divergences¶

The following six public functions exist in quantstats.stats but are not present as public methods on data.stats. Each is an intentional design decision, documented here so no gap goes unrecorded.

Shorthand aliases not carried over¶

QuantStats exposes several short alias functions alongside their canonical counterparts. jquantstats deliberately uses full names only, consistent with its deprecation of earlier aliases (ghpr(), r2(), win_loss_ratio() were deprecated in favour of geometric_mean(), r_squared(), and payoff_ratio() respectively).

QuantStats shorthand	Canonical equivalent in jquantstats	Rationale
`qs.stats.cvar(r)`	`data.stats.conditional_value_at_risk()`	Prefer unambiguous full name
`qs.stats.var(r)`	`data.stats.value_at_risk()`	Prefer unambiguous full name; `var` is also a common Python built-in pattern
`qs.stats.ror(r)`	`data.stats.risk_of_ruin()`	Prefer unambiguous full name
`qs.stats.upi(r)`	`data.stats.ulcer_performance_index()`	Prefer unambiguous full name
`qs.stats.expected_shortfall(r)`	`data.stats.conditional_value_at_risk()`	Alias for CVaR; jquantstats uses the canonical term

`to_drawdown_series` — covered by richer API¶

qs.stats.to_drawdown_series(returns) converts a return series to a drawdown series. jquantstats provides equivalent — and richer — functionality through two methods:

data.stats.drawdown() — returns a pl.DataFrame with the per-period drawdown from the running high-water mark, one column per asset.
data.stats.drawdown_details() — returns a full episode table (start date, end date, max drawdown, duration) for every drawdown period.

A bare drawdown series is therefore available as data.stats.drawdown()["asset_col"]; the standalone conversion utility is not needed.

Behavioural notes¶

`downside_deviation` convention¶

jquantstats and QuantStats both use the Red Rock Capital Sortino paper convention: the downside semi-deviation is the root-mean-square of strictly negative returns divided by the total observation count (not only the negative count). The implementations are aligned.

`information_ratio` annualisation¶

jquantstats annualises the information ratio by default (annualize=True), whereas QuantStats returns a raw (non-annualised) ratio. The jquantstats value can be reproduced without annualisation by reading the intermediate active-return statistics; the difference is documented in the Migration guide.