Risk

Compute delta, gamma, and bucket-level risk sensitivities from calibrated curves using automatic differentiation. The Solver provides AD-based delta and gamma, while IRImpliedCurve offers tenor-bucketed risk reporting.

Setup -- Calibrate with Labeled Instruments

Calibrate a DiscountCurve using the 5-tuple instrument format. Labels are required for delta and gamma output to identify which calibrating instrument drives each sensitivity.

import datetime
from vade import DiscountCurve, IRS, Solver

effective = datetime.date(2025, 6, 16)

nodes = {
    effective: 1.0,
    datetime.date(2026, 6, 16): 1.0,
    datetime.date(2027, 6, 16): 1.0,
    datetime.date(2028, 6, 16): 1.0,
    datetime.date(2030, 6, 16): 1.0,
    datetime.date(2035, 6, 16): 1.0,
}
curve = DiscountCurve(
    nodes, interpolation="log_linear", convention="act360", id="sofr"
)

irs_1y = IRS(effective=effective, termination="1Y", frequency="a", fixed_rate=4.00, convention="act360", float_convention="act360")
irs_2y = IRS(effective=effective, termination="2Y", frequency="a", fixed_rate=3.85, convention="act360", float_convention="act360")
irs_3y = IRS(effective=effective, termination="3Y", frequency="a", fixed_rate=3.75, convention="act360", float_convention="act360")
irs_5y = IRS(effective=effective, termination="5Y", frequency="a", fixed_rate=3.70, convention="act360", float_convention="act360")
irs_10y = IRS(effective=effective, termination="10Y", frequency="a", fixed_rate=3.85, convention="act360", float_convention="act360")

instruments = [
    (irs_1y, 4.00, 0, "1Y_IRS", "USD"),
    (irs_2y, 3.85, 0, "2Y_IRS", "USD"),
    (irs_3y, 3.75, 0, "3Y_IRS", "USD"),
    (irs_5y, 3.70, 0, "5Y_IRS", "USD"),
    (irs_10y, 3.85, 0, "10Y_IRS", "USD"),
]

solver = Solver(curves=[curve], instruments=instruments)
result = solver.iterate()
result.converged  # True

Delta -- First-Order Sensitivities

Delta measures how an instrument's NPV changes per unit move in each calibration instrument rate. The solver uses first-order AD (Dual numbers) to compute exact analytic sensitivities without finite differences.

import datetime
from vade import DiscountCurve, IRS, Solver

effective = datetime.date(2025, 6, 16)
nodes = {
    effective: 1.0,
    datetime.date(2026, 6, 16): 1.0,
    datetime.date(2027, 6, 16): 1.0,
    datetime.date(2028, 6, 16): 1.0,
    datetime.date(2030, 6, 16): 1.0,
    datetime.date(2035, 6, 16): 1.0,
}
curve = DiscountCurve(nodes, interpolation="log_linear", convention="act360", id="sofr")

irs_1y = IRS(effective=effective, termination="1Y", frequency="a", fixed_rate=4.00, convention="act360", float_convention="act360")
irs_2y = IRS(effective=effective, termination="2Y", frequency="a", fixed_rate=3.85, convention="act360", float_convention="act360")
irs_3y = IRS(effective=effective, termination="3Y", frequency="a", fixed_rate=3.75, convention="act360", float_convention="act360")
irs_5y = IRS(effective=effective, termination="5Y", frequency="a", fixed_rate=3.70, convention="act360", float_convention="act360")
irs_10y = IRS(effective=effective, termination="10Y", frequency="a", fixed_rate=3.85, convention="act360", float_convention="act360")

instruments = [
    (irs_1y, 4.00, 0, "1Y_IRS", "USD"),
    (irs_2y, 3.85, 0, "2Y_IRS", "USD"),
    (irs_3y, 3.75, 0, "3Y_IRS", "USD"),
    (irs_5y, 3.70, 0, "5Y_IRS", "USD"),
    (irs_10y, 3.85, 0, "10Y_IRS", "USD"),
]

solver = Solver(curves=[curve], instruments=instruments)
result = solver.iterate()

# Price a 5Y IRS slightly off-market
test_irs = IRS(effective=effective, termination="5Y", frequency="a", fixed_rate=3.80, convention="act360", float_convention="act360")

delta_df = solver.delta(test_irs, result)
delta_df.columns  # ['instrument_label', 'tenor', 'sensitivity', 'currency']
len(delta_df)  # 5
any(abs(s) > 1e-10 for s in delta_df["sensitivity"].to_list())  # True

Portfolio Delta

Pass a list of instruments to compute aggregated portfolio-level delta. The solver sums sensitivities across all instruments in the list.

import datetime
from vade import DiscountCurve, IRS, Solver

effective = datetime.date(2025, 6, 16)
nodes = {
    effective: 1.0,
    datetime.date(2026, 6, 16): 1.0,
    datetime.date(2027, 6, 16): 1.0,
    datetime.date(2028, 6, 16): 1.0,
    datetime.date(2030, 6, 16): 1.0,
    datetime.date(2035, 6, 16): 1.0,
}
curve = DiscountCurve(nodes, interpolation="log_linear", convention="act360", id="sofr")

irs_1y = IRS(effective=effective, termination="1Y", frequency="a", fixed_rate=4.00, convention="act360", float_convention="act360")
irs_2y = IRS(effective=effective, termination="2Y", frequency="a", fixed_rate=3.85, convention="act360", float_convention="act360")
irs_3y = IRS(effective=effective, termination="3Y", frequency="a", fixed_rate=3.75, convention="act360", float_convention="act360")
irs_5y = IRS(effective=effective, termination="5Y", frequency="a", fixed_rate=3.70, convention="act360", float_convention="act360")
irs_10y = IRS(effective=effective, termination="10Y", frequency="a", fixed_rate=3.85, convention="act360", float_convention="act360")

instruments = [
    (irs_1y, 4.00, 0, "1Y_IRS", "USD"),
    (irs_2y, 3.85, 0, "2Y_IRS", "USD"),
    (irs_3y, 3.75, 0, "3Y_IRS", "USD"),
    (irs_5y, 3.70, 0, "5Y_IRS", "USD"),
    (irs_10y, 3.85, 0, "10Y_IRS", "USD"),
]

solver = Solver(curves=[curve], instruments=instruments)
result = solver.iterate()

# Portfolio of two off-market swaps
port_irs1 = IRS(effective=effective, termination="3Y", frequency="a", fixed_rate=3.80, convention="act360", float_convention="act360")
port_irs2 = IRS(effective=effective, termination="7Y", frequency="a", fixed_rate=3.80, convention="act360", float_convention="act360")

portfolio_delta = solver.delta([port_irs1, port_irs2], result)
len(portfolio_delta)  # 5
portfolio_delta.columns  # ['instrument_label', 'tenor', 'sensitivity', 'currency']

Gamma -- Second-Order Sensitivities

Gamma measures second-order (convexity) risk using Dual2 automatic differentiation. The result is a cross-gamma matrix showing how delta changes with respect to each calibrating instrument rate.

import datetime
from vade import DiscountCurve, IRS, Solver

effective = datetime.date(2025, 6, 16)
nodes = {
    effective: 1.0,
    datetime.date(2026, 6, 16): 1.0,
    datetime.date(2027, 6, 16): 1.0,
    datetime.date(2028, 6, 16): 1.0,
    datetime.date(2030, 6, 16): 1.0,
    datetime.date(2035, 6, 16): 1.0,
}
curve = DiscountCurve(nodes, interpolation="log_linear", convention="act360", id="sofr")

irs_1y = IRS(effective=effective, termination="1Y", frequency="a", fixed_rate=4.00, convention="act360", float_convention="act360")
irs_2y = IRS(effective=effective, termination="2Y", frequency="a", fixed_rate=3.85, convention="act360", float_convention="act360")
irs_3y = IRS(effective=effective, termination="3Y", frequency="a", fixed_rate=3.75, convention="act360", float_convention="act360")
irs_5y = IRS(effective=effective, termination="5Y", frequency="a", fixed_rate=3.70, convention="act360", float_convention="act360")
irs_10y = IRS(effective=effective, termination="10Y", frequency="a", fixed_rate=3.85, convention="act360", float_convention="act360")

instruments = [
    (irs_1y, 4.00, 0, "1Y_IRS", "USD"),
    (irs_2y, 3.85, 0, "2Y_IRS", "USD"),
    (irs_3y, 3.75, 0, "3Y_IRS", "USD"),
    (irs_5y, 3.70, 0, "5Y_IRS", "USD"),
    (irs_10y, 3.85, 0, "10Y_IRS", "USD"),
]

solver = Solver(curves=[curve], instruments=instruments)
result = solver.iterate()

test_irs = IRS(effective=effective, termination="5Y", frequency="a", fixed_rate=3.80, convention="act360", float_convention="act360")

gamma_df = solver.gamma(test_irs, result)
"label" in gamma_df.columns  # True
len(gamma_df)  # 5

Bucket-Level Risk

IRImpliedCurve provides tenor-bucketed DV01 sensitivities by re-parameterizing a calibrated curve into standard tenor buckets. This gives a risk report aligned with market-standard tenor points.

import datetime
import math
from vade import DiscountCurve, IRImpliedCurve, BusinessCalendar, FixedRateBond

effective = datetime.date(2024, 1, 1)

# Build a curve with known DFs for bucket risk
nodes = {effective: 1.0}
for y in range(1, 11):
    nodes[datetime.date(2024 + y, 1, 1)] = math.exp(-0.03 * y)

source = DiscountCurve(
    nodes, interpolation="log_linear", convention="act365f", id="test_curve"
)

cal = BusinessCalendar("NYC")
tenors = ["1Y", "2Y", "5Y", "10Y"]
implied = IRImpliedCurve(source, effective, tenors, cal)
implied.tenor_labels()  # ['0D', '1Y', '2Y', '5Y', '10Y']

float(implied.discount_factor(datetime.date(2025, 1, 1))) > 0.95  # True

bond = FixedRateBond(effective=effective, termination="5Y", coupon=0.04, frequency="s")
bucket_df = implied.bucket_delta(bond)
"tenor" in bucket_df.columns  # True
"delta" in bucket_df.columns  # True
any(abs(float(d)) > 1e-12 for d in bucket_df["delta"].to_list())  # True

Next Steps

Calibration -- multi-curve frameworks and solver algorithm comparison
Curve Building -- interpolation methods and curve types