cedardev-fixed-income 1.0.0

Fixed Income Risk Analytics — US Treasuries, Agency MBS, Treasury Futures, VaR, Margining (SEC Rule 4210)

cedardev-fixed-income

Industrial-grade fixed income risk analytics for quantitative researchers and risk infrastructure teams.

Built by CedarDev Capital Management LLC — Market Risk Infrastructure desk.

---

What's in the box

Module	What it does
ust_pricer	US Treasury pricing: clean/dirty price, YTM, duration, DV01, convexity, zero-curve bootstrap, Z-spread
mbs_analytics	Agency MBS: PSA prepayment model, cash flow generation, WAL, OAS, PSA sensitivity
treasury_futures	CME Treasury futures: CTD identification, conversion factor, implied repo, basis
var_engine	VaR: Historical Simulation (delta-gamma), Parametric (EWMA), Expected Shortfall, Component VaR
margin_engine	SEC Rule 4210 margining pipeline: extract → categorize → net → cross-margin → validate
model_validation	Backtesting: Kupiec POF, Christoffersen independence, Basel traffic light, FRTB PLA test

---

Installation

pip install cedardev-fixed-income

---

Quick Start

Price a US Treasury bond

from datetime import date
from cedardev.fixed_income import USTBond, USTBondPricer, ZeroCurve

# Define the bond
bond = USTBond(
    cusip         = "91282CEX5",
    issue_date    = date(2021, 8, 15),
    maturity_date = date(2031, 7, 31),
    coupon_rate   = 0.0125,   # 1.25%
)

# Build a zero curve (from on-the-run par yields)
par_tenors = [0.25, 0.5, 1.0, 2.0, 3.0, 5.0, 7.0, 10.0, 20.0, 30.0]
par_yields = [0.0530, 0.0528, 0.0520, 0.0490, 0.0475, 0.0450,
              0.0440, 0.0430, 0.0445, 0.0435]
zero_curve = ZeroCurve.bootstrap(par_tenors, par_yields)

# Price and get risk measures
pricer     = USTBondPricer()
settlement = date(2024, 7, 1)
analytics  = pricer.full_analytics(bond, clean_price_100=84.10, settlement=settlement,
                                   zero_curve=zero_curve)

print(f"YTM             : {analytics.ytm:.4%}")
print(f"Modified Duration: {analytics.modified_duration:.4f} years")
print(f"DV01            : ${analytics.dv01 * 1_000_000:,.0f} per $1MM face per bp")
print(f"Z-spread        : {analytics.z_spread:.1f} bps")

Price an Agency MBS pool

from datetime import date
from cedardev.fixed_income import MBSPool, MBSPricer, AgencyType, CollateralType
import numpy as np

# Zero curve
zc_tenors = np.array([0.25, 0.5, 1, 2, 3, 5, 7, 10, 15, 20, 25, 30])
zc_rates  = np.array([0.0528, 0.0526, 0.0518, 0.0488, 0.0472, 0.0447,
                      0.0437, 0.0427, 0.0432, 0.0440, 0.0437, 0.0433])

pool = MBSPool(
    pool_number      = "MA3456",
    agency           = AgencyType.FNMA,
    collateral_type  = CollateralType.FIXED_RATE,
    original_balance = 50_000_000,
    current_factor   = 0.85,
    gross_coupon     = 0.065,
    net_coupon       = 0.0625,
    wam              = 320,
    wala             = 40,
    issue_date       = date(2021, 3, 1),
    settlement_date  = date(2024, 7, 1),
)

pricer    = MBSPricer(zc_tenors, zc_rates)
analytics = pricer.full_analytics(pool, price_100=101.25, psa_speed=150.0)

print(f"OAS    : {analytics.oas_bps:.1f} bps")
print(f"WAL    : {analytics.wal_years:.2f} years")
print(f"OA Dur : {analytics.modified_duration:.3f}")
print(f"DV01   : ${analytics.dv01 * pool.current_balance:,.0f}")

Run the SEC Rule 4210 Margin Engine

from cedardev.fixed_income import MarginPipeline
from pathlib import Path

pipeline = MarginPipeline(
    use_synthetic = True,       # swap for False in production (uses Snowflake)
    output_dir    = Path("./margin_output"),
)
results_df, validation = pipeline.run()

# results_df columns:
# portfolio_id | bucket | gross_margin | net_margin | cross_margin_credit | final_margin

Backtest a VaR model

from cedardev.fixed_income import (
    ModelValidator, BacktestWindow, KupiecTest, BaselTrafficLightAssigner
)
import numpy as np
from datetime import date, timedelta

# Simulated backtest data (replace with real desk P&L)
np.random.seed(42)
n          = 500
var_series = np.full(n, 500_000 * 2.326)   # 99% VaR estimate
actual_pnl = np.random.normal(0, 500_000, n)

backtest = [
    BacktestWindow(
        date         = date(2023, 1, 1) + timedelta(days=i),
        var_estimate = var_series[i],
        actual_pnl   = actual_pnl[i],
    )
    for i in range(n)
]

validator = ModelValidator()
report    = validator.run(
    backtest_data    = backtest,
    hypothetical_pnl = actual_pnl * 0.95,
    actual_pnl       = actual_pnl,
)
validator.print_report(report)
# → Kupiec, Christoffersen, Basel traffic light, PLA test results

---

Design principles

• No black boxes. Every calculation is traceable to its formula. Comments reference Fabozzi, Basel documents, and SEC releases.
• No ML. Traditional quantitative methods only — Pandas, NumPy, SciPy. Exactly what regulators expect.
• Audit trails. Every MarginResult carries audit_notes with the full dollar breakdown.
• Regulatory-first. Day-count conventions, bucket boundaries, margin rates, and correlation matrices all sourced from published SEC/FINRA/CME documentation.
• Snowflake-native. The margin engine queries RATES_DW.POSITIONS.OPEN_POSITIONS_V by default, with a --synthetic flag for testing without a live DB connection.

---

Regulatory coverage

Standard	Where implemented
SEC Rule 4210 (FINRA)	margin_engine.py
Basel II.5 / III internal models	var_engine.py
Basel backtesting framework (1996)	model_validation.py
FRTB P&L Attribution test	model_validation.py
CME Treasury futures delivery specs	treasury_futures.py
SIFMA PSA prepayment standard	mbs_analytics.py
ACT/ACT ICMA (ISMA Rule 251)	ust_pricer.py

---

Requirements

• Python 3.11+
• pandas, numpy, scipy, pydantic, tabulate, colorlog
• snowflake-connector-python (only needed for live DB extraction)

---

License

MIT — see LICENSE.

---

Disclaimer

This library is intended for internal quantitative research and risk infrastructure use. All regulatory table values are illustrative and based on publicly available SEC/FINRA documentation. Always verify against your firm's current regulatory schedule before using in production.