Source code for quantdata.dysanos.tangent_levy.evaluation

"""Daily fit evaluation against SANOS and retained SPX quote intervals."""

from __future__ import annotations

from dataclasses import dataclass
from datetime import date

import numpy as np
from cdxcore.err import verify

from .config import StaticFitConfig
from .double_exponential import call_prices
from .markets import TangentLevyMarketDay
from .metrics import (
    IntervalErrorMetrics,
    PriceErrorMetrics,
    implied_volatility_error,
    interval_error_metrics,
    price_error_metrics,
)
from .static_fit import DayFit


[docs] @dataclass(frozen=True) class DailyFitEvaluation: """Outright and executable-interval metrics for one calibrated day.""" market_date: date maturities: int optimizer_success_fraction: float tangent_vs_sanos_price: PriceErrorMetrics tangent_vs_sanos_iv: PriceErrorMetrics tangent_pure_interval: IntervalErrorMetrics sanos_pure_interval: IntervalErrorMetrics tangent_call_interval: IntervalErrorMetrics | None tangent_put_interval: IntervalErrorMetrics | None
def _optional_interval_metrics( model: np.ndarray, bids: np.ndarray, asks: np.ndarray, mask: np.ndarray, *, half_spread_floor: float, ) -> IntervalErrorMetrics | None: active = ( mask & np.isfinite(model) & np.isfinite(bids) & np.isfinite(asks) & (asks > bids) ) if not np.any(active): return None return interval_error_metrics( model, bids, asks, mask=active, half_spread_floor=half_spread_floor, )
[docs] def evaluate_day_fit( market: TangentLevyMarketDay, day_fit: DayFit, *, config: StaticFitConfig = StaticFitConfig(), ) -> DailyFitEvaluation: """Score a calibrated DETL day at the original retained market strikes.""" verify( market.market_date == day_fit.market_date, lambda: ( f"market date {market.market_date} does not match fit date " f"{day_fit.market_date}" ), ) model_parts: list[np.ndarray] = [] target_parts: list[np.ndarray] = [] strike_parts: list[np.ndarray] = [] tau_parts: list[np.ndarray] = [] pure_bid_parts: list[np.ndarray] = [] pure_ask_parts: list[np.ndarray] = [] model_mask_parts: list[np.ndarray] = [] call_bid_parts: list[np.ndarray] = [] call_ask_parts: list[np.ndarray] = [] put_bid_parts: list[np.ndarray] = [] put_ask_parts: list[np.ndarray] = [] call_active_parts: list[np.ndarray] = [] put_active_parts: list[np.ndarray] = [] for fit in day_fit.fits: matches = np.flatnonzero( np.isclose(market.expiries, fit.tau, rtol=0.0, atol=1.0e-8) ) verify( matches.size == 1, lambda fit=fit: ( f"fit maturity {fit.tau} does not uniquely match the market grid" ), ) index = int(matches[0]) strikes = market.strikes[index] model = call_prices( tau=fit.tau, log_strikes=np.log(strikes), parameters=fit.parameters, config=config.pricing, ) model_parts.append(model) target_parts.append(market.sanos_calls[index]) strike_parts.append(strikes) tau_parts.append(np.full(strikes.shape, fit.tau)) pure_bid_parts.append(market.pure_bids[index]) pure_ask_parts.append(market.pure_asks[index]) model_mask_parts.append(market.model_masks[index]) call_bid_parts.append(market.call_bids[index]) call_ask_parts.append(market.call_asks[index]) put_bid_parts.append(market.put_bids[index]) put_ask_parts.append(market.put_asks[index]) call_active_parts.append(market.call_active[index]) put_active_parts.append(market.put_active[index]) model = np.concatenate(model_parts) target = np.concatenate(target_parts) strikes = np.concatenate(strike_parts) tau = np.concatenate(tau_parts) pure_bids = np.concatenate(pure_bid_parts) pure_asks = np.concatenate(pure_ask_parts) model_mask = np.concatenate(model_mask_parts) half_spread = np.maximum( 0.5 * (pure_asks - pure_bids), config.half_spread_floor, ) call_bids = np.concatenate(call_bid_parts) call_asks = np.concatenate(call_ask_parts) put_bids = np.concatenate(put_bid_parts) put_asks = np.concatenate(put_ask_parts) call_active = np.concatenate(call_active_parts) put_active = np.concatenate(put_active_parts) model_puts = model - 1.0 + strikes return DailyFitEvaluation( market_date=market.market_date, maturities=len(day_fit.fits), optimizer_success_fraction=float( np.mean([fit.success for fit in day_fit.fits]) ), tangent_vs_sanos_price=price_error_metrics( model, target, half_spread=half_spread, mask=model_mask, ), tangent_vs_sanos_iv=implied_volatility_error( tau=tau, strikes=strikes, model_calls=model, target_calls=target, mask=model_mask, ), tangent_pure_interval=interval_error_metrics( model, pure_bids, pure_asks, mask=model_mask, half_spread_floor=config.half_spread_floor, ), sanos_pure_interval=interval_error_metrics( target, pure_bids, pure_asks, mask=model_mask, half_spread_floor=config.half_spread_floor, ), tangent_call_interval=_optional_interval_metrics( model, call_bids, call_asks, call_active, half_spread_floor=config.half_spread_floor, ), tangent_put_interval=_optional_interval_metrics( model_puts, put_bids, put_asks, put_active, half_spread_floor=config.half_spread_floor, ), )