Repository Public MIT
1#!/usr/bin/env python32# ORIX Calc3# File: Packages/ORIXFinancial/Reference/generate_reference.py4# Description: Independent reference values for the ORIXFinancial test suite (brute force / high precision, no Swift code involved).5#6# Author: Simon-Pierre Boucher7# Contact: contact@spboucher.ai8#9# Usage: python3 Reference/generate_reference.py10# Writes: Reference/reference.json and a copy in Tests/ORIXFinancialTests/Resources/reference.json11#12# Method: every value is obtained by a route different from the Swift implementation —13# * TVM: period-by-period compounding of the balance in 50-digit arithmetic (mpmath), never the closed form;14# * cash flows: term-by-term discounting; IRRs from the polynomial roots (mpmath.polyroots / findroot);15# * loans: closed forms in 50 digits; schedules simulated with Python's decimal module (exact cents, ROUND_HALF_UP);16# * bonds: term-by-term discounting with dates from the standard library; durations/convexity by high-precision17# numerical differentiation of the price function (mpmath.diff), not by the analytic sums;18# * options: Black-Scholes-Merton with mpmath.ncdf; Greeks by mpmath.diff of the price; implied volatility by19# scipy.optimize.brentq; CRR trees with numpy.20import calendar21import datetime as dt22import json23import os24import shutil25from decimal import Decimal, ROUND_HALF_UP, getcontext2627import mpmath as mp28import numpy as np29import scipy30from scipy.optimize import brentq31from scipy.stats import norm3233mp.mp.dps = 5034getcontext().prec = 6035F = float36out = {"_meta": {"generator": "generate_reference.py", "mpmath_dps": 50, "scipy": scipy.__version__, "numpy": np.__version__}}3738# ---------------------------------------------------------------- TVM39def tvm_fv(n, iy, pv, pmt, py, cy, begin):40 i = (1 + mp.mpf(iy) / 100 / cy) ** (mp.mpf(cy) / py) - 141 bal = mp.mpf(pv)42 for _ in range(n):43 bal = (bal + pmt) * (1 + i) if begin else bal * (1 + i) + pmt44 return -bal, i4546tvm_cases = [47 dict(name="us_mortgage", n=360, iy="6", pv="200000", pmt="-1199.10", py=12, cy=12, begin=False),48 dict(name="lump_sum", n=10, iy="5", pv="-1000", pmt="0", py=1, cy=1, begin=False),49 dict(name="annuity_due", n=20, iy="7", pv="0", pmt="-500", py=1, cy=1, begin=True),50 dict(name="canadian_py12_cy2", n=300, iy="5", pv="300000", pmt="-1744.81", py=12, cy=2, begin=False),51 dict(name="zero_rate", n=48, iy="0", pv="12000", pmt="-250", py=12, cy=12, begin=False),52 dict(name="negative_rate", n=10, iy="-2", pv="-1000", pmt="-100", py=1, cy=1, begin=False),53 dict(name="quarterly_pay_monthly_comp_begin", n=40, iy="8", pv="-5000", pmt="-200", py=4, cy=12, begin=True),54 dict(name="daily_compounding", n=24, iy="3.5", pv="10000", pmt="-300", py=12, cy=365, begin=False),55 dict(name="tiny_rate", n=120, iy="0.0000001", pv="-100", pmt="-10", py=12, cy=12, begin=False),56 dict(name="savings_begin_semiannual_comp", n=180, iy="4.25", pv="-2500", pmt="-150", py=12, cy=2, begin=True),57 dict(name="lease_with_residual", n=48, iy="3.9", pv="35000", pmt="-450", py=12, cy=12, begin=True),58]59out["tvm"] = []60for c in tvm_cases:61 fv, i = tvm_fv(c["n"], c["iy"], mp.mpf(c["pv"]), mp.mpf(c["pmt"]), c["py"], c["cy"], c["begin"])62 out["tvm"].append(dict(name=c["name"], n=c["n"], iy=F(c["iy"]), pv=F(c["pv"]), pmt=F(c["pmt"]), fv=F(fv),63 py=c["py"], cy=c["cy"], begin=c["begin"], periodic_rate=F(i)))6465# rate conversions66out["rates"] = dict(67 eff_6_monthly=F((1 + mp.mpf("0.06") / 12) ** 12 - 1),68 eff_6_continuous=F(mp.e ** mp.mpf("0.06") - 1),69 nom_semi_from_eff_5=F(2 * ((1 + mp.mpf("0.05")) ** (mp.mpf(1) / 2) - 1)),70 nom_cont_from_eff_5=F(mp.log(1 + mp.mpf("0.05"))),71 monthly_from_semi_5=F(12 * ((1 + mp.mpf("0.05") / 2) ** (mp.mpf(1) / 6) - 1)),72)7374# ---------------------------------------------------------------- Cash flows75def npv(flows, r):76 return sum(mp.mpf(c) / (1 + mp.mpf(r)) ** k for k, c in enumerate(flows))7778def real_irrs(flows):79 # roots of sum c_k x^k in x = 1/(1+r), x > 080 coeffs = [mp.mpf(c) for c in reversed(flows)] # highest degree first81 while coeffs and coeffs[0] == 0:82 coeffs.pop(0)83 roots = mp.polyroots(coeffs, maxsteps=2000, extraprec=400)84 rs = []85 for x in roots:86 if abs(mp.im(x)) < mp.mpf(10) ** -25 and mp.re(x) > 0:87 rs.append(1 / mp.re(x) - 1)88 return sorted(F(r) for r in rs)8990def payback(flows):91 cum = mp.mpf(flows[0])92 if cum >= 0:93 return 0.094 for k in range(1, len(flows)):95 nxt = cum + flows[k]96 if nxt >= 0:97 return F(k - 1 + (-cum) / flows[k])98 cum = nxt99 return None100101def mirr(flows, fin, re):102 n = len(flows) - 1103 pvneg = sum(mp.mpf(c) / (1 + mp.mpf(fin)) ** k for k, c in enumerate(flows) if c < 0)104 fvpos = sum(mp.mpf(c) * (1 + mp.mpf(re)) ** (n - k) for k, c in enumerate(flows) if c > 0)105 return F((fvpos / -pvneg) ** (mp.mpf(1) / n) - 1)106107cf_cases = [108 dict(name="conventional", flows=[-10000, 3000, 4200, 6800], rate="0.10"),109 dict(name="two_irrs", flows=[-1000, 2300, -1320], rate="0.15"),110 dict(name="no_real_irr", flows=[1000, -3000, 2500], rate="0.10"),111 dict(name="grouped", flows=[-50000] + [12000] * 5 + [20000], rate="0.08"),112 dict(name="negative_irr", flows=[-1000, 200, 300, 300], rate="0.05"),113 dict(name="three_irrs", flows=[-1000, 3600, -4310, 1716], rate="0.25"),114 dict(name="mine_with_cleanup", flows=[-100, 50, 50, 50, 50, -150], rate="0.07"),115 dict(name="negative_discount_rate", flows=[-500, 100, 150, 200, 120], rate="-0.03"),116]117out["cashflows"] = []118for c in cf_cases:119 fl, r = c["flows"], mp.mpf(c["rate"])120 n = len(fl) - 1121 v = npv(fl, r)122 disc = [mp.mpf(x) / (1 + r) ** k for k, x in enumerate(fl)]123 entry = dict(name=c["name"], flows=fl, rate=F(r), npv=F(v), nfv=F(v * (1 + r) ** n), irrs=real_irrs(fl),124 payback=payback(fl), discounted_payback=payback(disc))125 if any(x < 0 for x in fl) and any(x > 0 for x in fl):126 entry["mirr_fin"], entry["mirr_re"] = 0.06, 0.09127 entry["mirr"] = mirr(fl, "0.06", "0.09")128 if fl[0] < 0:129 entry["pi"] = F((v - fl[0]) / -fl[0])130 out["cashflows"].append(entry)131# long series: one IRR, found by a high-precision root search132long_flows = [-100000] + [599.55] * 360133r_long = mp.findroot(lambda r: npv(long_flows, r), mp.mpf("0.005"))134out["cashflow_long"] = dict(initial=-100000, amount=599.55, count=360, irr=F(r_long), npv_at_0004=F(npv(long_flows, "0.004")))135136# ---------------------------------------------------------------- Loans137PPY = dict(monthly=12, semiMonthly=24, biweekly=26, acceleratedBiweekly=26, weekly=52, acceleratedWeekly=52)138CY = dict(usMonthly=12, canadianFixed=2, canadianVariable=12)139140def periodic(rate, comp, ppy):141 rate = mp.mpf(rate)142 if comp == "continuous":143 return mp.e ** (rate / ppy) - 1144 m = CY[comp] if comp in CY else int(comp)145 return (1 + rate / m) ** (mp.mpf(m) / ppy) - 1146147def level(P, i, n):148 return P / n if i == 0 else P * i / (1 - (1 + i) ** (-n))149150def raw_payment(P, rate, comp, freq, remaining):151 ppy = PPY[freq]152 if freq.startswith("accelerated"):153 im = periodic(rate, comp, 12)154 return level(mp.mpf(P), im, mp.mpf(remaining) * 12 / ppy) / (2 if freq == "acceleratedBiweekly" else 4)155 return level(mp.mpf(P), periodic(rate, comp, ppy), remaining)156157def cents(x):158 return Decimal(mp.nstr(x, 40)).quantize(Decimal("0.01"), rounding=ROUND_HALF_UP)159160pay_cases = [161 dict(name="canadian_300k_5_25", P="300000", rate="0.05", years=25, freq="monthly", comp="canadianFixed"),162 dict(name="us_250k_65_30", P="250000", rate="0.065", years=30, freq="monthly", comp="usMonthly"),163 dict(name="cdn_var_semi_monthly", P="425000", rate="0.0595", years=25, freq="semiMonthly", comp="canadianVariable"),164 dict(name="cdn_biweekly", P="400000", rate="0.0479", years=25, freq="biweekly", comp="canadianFixed"),165 dict(name="cdn_acc_biweekly", P="400000", rate="0.0479", years=25, freq="acceleratedBiweekly", comp="canadianFixed"),166 dict(name="cdn_weekly", P="400000", rate="0.0479", years=25, freq="weekly", comp="canadianFixed"),167 dict(name="cdn_acc_weekly", P="400000", rate="0.0479", years=25, freq="acceleratedWeekly", comp="canadianFixed"),168 dict(name="quarterly_comp", P="80000", rate="0.072", years=7, freq="monthly", comp="4"),169 dict(name="continuous", P="80000", rate="0.072", years=7, freq="monthly", comp="continuous"),170 dict(name="zero_rate", P="12000", rate="0", years=4, freq="monthly", comp="usMonthly"),171]172out["loan_payments"] = []173for c in pay_cases:174 ppy = PPY[c["freq"]]175 raw = raw_payment(c["P"], c["rate"], c["comp"], c["freq"], c["years"] * ppy)176 out["loan_payments"].append(dict(c, payment_unrounded=F(raw), payment=F(cents(raw)), periodic_rate=F(periodic(c["rate"], c["comp"], ppy))))177178def schedule(P, rate, comp, freq, years, lumps=None, recurring=None, changes=None, override=None, stop=None):179 """Cent-exact simulation with the decimal module. changes: {period: (rate, comp_or_None, 'recalc'|'keep')}."""180 ppy = PPY[freq]181 n = years * ppy182 lumps, recurring, changes = lumps or {}, recurring or [], changes or {}183 accelerated = freq.startswith("accelerated")184 i = periodic(rate, comp, ppy)185 if override is not None:186 pmt, forced = cents(mp.mpf(override)), None187 else:188 pmt, forced = cents(raw_payment(P, rate, comp, freq, n)), (None if accelerated else n)189 bal = Decimal(P)190 rows, k = [], 0191 first_payment = pmt192 while bal > 0:193 k += 1194 if k in changes:195 rate, newcomp, policy = changes[k]196 comp = newcomp or comp197 i = periodic(rate, comp, ppy)198 if policy == "recalc":199 pmt = cents(raw_payment(mp.mpf(str(bal)), rate, comp, freq, max(1, n - k + 1)))200 forced = None if accelerated else max(n, k)201 else:202 forced = None203 interest = cents(mp.mpf(str(bal)) * i)204 if k == forced or bal + interest <= pmt:205 principal, paid, bal = bal, bal + interest, Decimal(0)206 else:207 assert pmt > interest208 principal, paid = pmt - interest, pmt209 bal -= principal210 extra = Decimal(0)211 if k in lumps:212 extra += Decimal(lumps[k])213 for (start, end, every, amount) in recurring:214 if k >= start and (end is None or k <= end) and (k - start) % every == 0:215 extra += Decimal(amount)216 extra = min(extra, bal)217 bal -= extra218 rows.append((k, paid, interest, principal, extra, bal))219 if stop is not None and k == stop:220 break221 return rows, first_payment222223def summarize(rows, first_payment, probes=()):224 return dict(225 payments=len(rows), payment=F(first_payment), final_payment=F(rows[-1][1]),226 total_interest=F(sum(r[2] for r in rows)), total_principal=F(sum(r[3] for r in rows)),227 total_prepayments=F(sum(r[4] for r in rows)), total_paid=F(sum(r[1] + r[4] for r in rows)),228 ending_balance=F(rows[-1][5]),229 probes=[dict(period=p, interest=F(rows[p - 1][2]), principal=F(rows[p - 1][3]), balance=F(rows[p - 1][5])) for p in probes if p <= len(rows)],230 )231232sched_cases = [233 dict(name="canadian_300k", args=dict(P="300000", rate="0.05", comp="canadianFixed", freq="monthly", years=25), probes=(1, 2, 60, 150, 299, 300)),234 dict(name="us_250k", args=dict(P="250000", rate="0.065", comp="usMonthly", freq="monthly", years=30), probes=(1, 12, 180, 359, 360)),235 dict(name="acc_biweekly", args=dict(P="400000", rate="0.0479", comp="canadianFixed", freq="acceleratedBiweekly", years=25), probes=(1, 26, 260)),236 dict(name="weekly", args=dict(P="150000", rate="0.039", comp="usMonthly", freq="weekly", years=10), probes=(1, 52, 519, 520)),237 dict(name="semi_monthly_quarterly_comp", args=dict(P="95000", rate="0.0825", comp="4", freq="semiMonthly", years=5), probes=(1, 60, 120)),238 dict(name="zero_rate", args=dict(P="12000", rate="0", comp="usMonthly", freq="monthly", years=4), probes=(1, 48)),239 dict(name="prepayments", args=dict(P="250000", rate="0.065", comp="usMonthly", freq="monthly", years=30,240 lumps={24: "10000"}, recurring=[(1, None, 1, "200"), (12, 120, 12, "1500")]), probes=(1, 24, 25, 120)),241 dict(name="rate_changes", args=dict(P="300000", rate="0.05", comp="canadianFixed", freq="monthly", years=25,242 changes={61: ("0.065", None, "recalc"), 121: ("0.04", None, "keep")}), probes=(60, 61, 120, 121, 200)),243 dict(name="rate_up_keep_payment", args=dict(P="200000", rate="0.03", comp="canadianVariable", freq="monthly", years=25,244 changes={13: ("0.045", None, "keep")}), probes=(12, 13, 300)),245 dict(name="payment_override", args=dict(P="20000", rate="0.0699", comp="usMonthly", freq="monthly", years=5, override="500"), probes=(1, 12)),246 dict(name="term_5y", args=dict(P="300000", rate="0.05", comp="canadianFixed", freq="monthly", years=25, stop=60), probes=(60,)),247]248out["loan_schedules"] = []249for c in sched_cases:250 rows, pmt = schedule(**c["args"])251 a = dict(c["args"])252 a["lumps"] = [[k, F(v)] for k, v in (a.get("lumps") or {}).items()]253 a["recurring"] = [[s, e if e is not None else -1, ev, F(am)] for (s, e, ev, am) in (a.get("recurring") or [])]254 a["changes"] = [[k, F(v[0]), v[2]] for k, v in (a.get("changes") or {}).items()]255 out["loan_schedules"].append(dict(name=c["name"], args=a, **summarize(rows, pmt, c["probes"])))256257# refinancing: 240k left at 6.5% US over 25 years, new 5.25% over 25 years, 3000 costs + 1500 penalty paid in cash258def refi(financed):259 costs = Decimal("4500")260 old, pold = schedule("240000", "0.065", "usMonthly", "monthly", 25)261 new, pnew = schedule(str(Decimal("240000") + (costs if financed else 0)), "0.0525", "usMonthly", "monthly", 25)262 cash = Decimal(0) if financed else costs263 savings = pold - pnew264 be = None265 po = pn = Decimal(0)266 for k in range(1, max(len(old), len(new)) + 1):267 if k <= len(old): po += old[k - 1][1]268 if k <= len(new): pn += new[k - 1][1]269 bo = old[k - 1][5] if k <= len(old) else Decimal(0)270 bn = new[k - 1][5] if k <= len(new) else Decimal(0)271 if pn + bn + cash <= po + bo:272 be = k273 break274 return dict(financed=financed, current_payment=F(pold), new_payment=F(pnew), savings=F(savings),275 simple_break_even=F(costs / savings), net_break_even=be,276 current_interest=F(sum(r[2] for r in old)), new_interest=F(sum(r[2] for r in new)),277 lifetime_savings=F(sum(r[1] for r in old) - sum(r[1] for r in new) - cash))278out["refinance"] = [refi(False), refi(True)]279280# affordability: income 120k, tax 4200, heat 1500, condo 300 (50%), debts 650, limits 39% / 44%, 5.25% Canadian fixed, 25 years281inc = mp.mpf(120000) / 12282other = mp.mpf(4200) / 12 + mp.mpf(1500) / 12 + 300 * mp.mpf("0.5")283by_h, by_t = mp.mpf("0.39") * inc - other, mp.mpf("0.44") * inc - other - 650284i_aff = periodic("0.0525", "canadianFixed", 12)285pm = min(by_h, by_t)286out["affordability"] = dict(by_housing=F(by_h), by_total=F(by_t), max_payment=F(pm), max_loan=F(pm * (1 - (1 + i_aff) ** -300) / i_aff))287288# ---------------------------------------------------------------- Day counts and bonds289def eom(d):290 return d.day == calendar.monthrange(d.year, d.month)[1]291292def days_30_360_us(a, b):293 d1, d2 = a.day, b.day294 fa, fb = (a.month == 2 and eom(a)), (b.month == 2 and eom(b))295 if fa and fb: d2 = 30296 if fa: d1 = 30297 if d2 == 31 and d1 >= 30: d2 = 30298 if d1 == 31: d1 = 30299 return 360 * (b.year - a.year) + 30 * (b.month - a.month) + (d2 - d1)300301def days_30e_360(a, b):302 return 360 * (b.year - a.year) + 30 * (b.month - a.month) + (min(b.day, 30) - min(a.day, 30))303304def daycount(conv, a, b):305 if conv == "thirty360US": return days_30_360_us(a, b)306 if conv == "thirtyE360": return days_30e_360(a, b)307 return (b - a).days308309D = dt.date310pairs = [(D(2024, 1, 31), D(2024, 2, 29)), (D(2023, 2, 28), D(2024, 2, 29)), (D(2024, 2, 29), D(2024, 8, 31)),311 (D(2025, 1, 30), D(2025, 3, 31)), (D(2025, 1, 31), D(2025, 3, 31)), (D(2024, 12, 15), D(2025, 6, 15)),312 (D(2023, 8, 31), D(2024, 2, 28)), (D(2020, 2, 29), D(2021, 2, 28)), (D(1999, 12, 31), D(2000, 3, 1)),313 (D(2024, 5, 31), D(2024, 6, 30))]314CONVS = ["thirty360US", "thirtyE360", "actual360", "actual365Fixed", "actualActualICMA"]315out["daycounts"] = [dict(a=a.isoformat(), b=b.isoformat(), **{c: daycount(c, a, b) for c in CONVS}) for a, b in pairs]316317def add_months(d, n, eom_rule):318 t = d.year * 12 + d.month - 1 + n319 y, m = divmod(t, 12)320 m += 1321 dim = calendar.monthrange(y, m)[1]322 return D(y, m, dim if (eom_rule and eom(d)) else min(d.day, dim))323324def coupon_dates(settle, maturity, f, eom_rule=True):325 rem, k = [], 0326 while True:327 d = add_months(maturity, -k * (12 // f), eom_rule)328 if d <= settle:329 return d, list(reversed(rem))330 rem.append(d)331 k += 1332333def bond_setup(face, cpn, f, redemption, conv, settle, maturity):334 prev, rem = coupon_dates(settle, maturity, f)335 nxt = rem[0]336 C = mp.mpf(face) * mp.mpf(cpn) / f337 A = mp.mpf(daycount(conv, prev, settle))338 if conv == "actualActualICMA":339 E = mp.mpf((nxt - prev).days); w = mp.mpf((nxt - settle).days) / E340 elif conv in ("thirty360US", "thirtyE360"):341 E = mp.mpf(360) / f; w = (E - A) / E342 elif conv == "actual360":343 E = mp.mpf(360) / f; w = mp.mpf((nxt - settle).days) / E344 else:345 E = mp.mpf(365) / f; w = mp.mpf((nxt - settle).days) / E346 flows = [(w + k, C + (mp.mpf(redemption) if k == len(rem) - 1 else 0)) for k in range(len(rem))]347 return flows, C * A / E, A, E, w, rem348349def dirty(flows, y, f):350 return sum(a / (1 + mp.mpf(y) / f) ** t for t, a in flows)351352def bond_metrics(flows, accrued, y, f):353 y = mp.mpf(y)354 P = dirty(flows, y, f)355 dP = mp.diff(lambda x: dirty(flows, x, f), y)356 d2P = mp.diff(lambda x: dirty(flows, x, f), y, 2)357 mod = -dP / P358 return dict(dirty=F(P), clean=F(P - accrued), accrued=F(accrued), modified=F(mod), macaulay=F(mod * (1 + y / f)), convexity=F(d2P / P),359 ytm_from_clean=F(brentq(lambda x: float(dirty(flows, x, f) - P), -0.5, 5.0, xtol=1e-15, rtol=8.9e-16, maxiter=500)))360361bond_cases = [362 dict(name="excel_price_example", face=100, cpn="0.0575", f=2, red=100, conv="thirty360US", settle=D(2008, 2, 15), mat=D(2017, 11, 15), y="0.065"),363 dict(name="icma_semi", face=100, cpn="0.04375", f=2, red=100, conv="actualActualICMA", settle=D(2025, 3, 10), mat=D(2034, 11, 15), y="0.0412"),364 dict(name="e30_annual", face=1000, cpn="0.03", f=1, red=1000, conv="thirtyE360", settle=D(2024, 7, 31), mat=D(2031, 1, 31), y="0.0355"),365 dict(name="act360_quarterly", face=100, cpn="0.06", f=4, red=100, conv="actual360", settle=D(2025, 1, 20), mat=D(2027, 12, 15), y="0.052"),366 dict(name="act365_monthly", face=100, cpn="0.048", f=12, red=100, conv="actual365Fixed", settle=D(2025, 5, 7), mat=D(2026, 9, 30), y="0.07"),367 dict(name="eom_feb_maturity", face=100, cpn="0.05", f=2, red=100, conv="thirty360US", settle=D(2024, 10, 15), mat=D(2029, 2, 28), y="0.045"),368 dict(name="zero_coupon", face=100, cpn="0", f=2, red=100, conv="actualActualICMA", settle=D(2025, 6, 1), mat=D(2035, 6, 1), y="0.04"),369 dict(name="negative_yield", face=100, cpn="0.005", f=1, red=100, conv="actualActualICMA", settle=D(2021, 4, 12), mat=D(2026, 8, 15), y="-0.004"),370 dict(name="on_coupon_date", face=100, cpn="0.06", f=2, red=100, conv="thirty360US", settle=D(2025, 6, 15), mat=D(2030, 6, 15), y="0.05"),371 dict(name="premium_redemption", face=100, cpn="0.07", f=2, red=103, conv="thirty360US", settle=D(2025, 2, 3), mat=D(2028, 9, 1), y="0.058"),372]373out["bonds"] = []374for c in bond_cases:375 flows, acc, A, E, w, rem = bond_setup(c["face"], c["cpn"], c["f"], c["red"], c["conv"], c["settle"], c["mat"])376 m = bond_metrics(flows, acc, c["y"], c["f"])377 out["bonds"].append(dict(name=c["name"], face=c["face"], coupon=F(c["cpn"]), frequency=c["f"], redemption=c["red"], convention=c["conv"],378 settlement=c["settle"].isoformat(), maturity=c["mat"].isoformat(), y=F(c["y"]),379 accrued_days=F(A), period_days=F(E), w=F(w), remaining=len(rem),380 first_coupon=rem[0].isoformat(), **m))381382out["bonds_simple"] = []383for (face, cpn, f, n, y) in [(100, "0.08", 2, 20, "0.10"), (1000, "0.05", 1, 10, "0.05"), (100, "0.03", 4, 37, "0.021"), (100, "0.09", 12, 60, "0.12")]:384 C = mp.mpf(face) * mp.mpf(cpn) / f385 flows = [(mp.mpf(k), C + (face if k == n else 0)) for k in range(1, n + 1)]386 out["bonds_simple"].append(dict(face=face, coupon=F(cpn), frequency=f, periods=n, y=F(y), **bond_metrics(flows, mp.mpf(0), y, f)))387388# yield to call: 8% semi-annual, callable 2029-06-15 at 104, clean price 106.50, settlement 2025-09-02, 30/360 US389flows, acc, *_ = bond_setup(100, "0.08", 2, 104, "thirty360US", D(2025, 9, 2), D(2029, 6, 15))390out["yield_to_call"] = dict(clean=106.5, call_price=104, accrued=F(acc),391 ytc=F(mp.findroot(lambda x: dirty(flows, x, 2) - (mp.mpf("106.5") + acc), mp.mpf("0.06"))))392393# ---------------------------------------------------------------- Investment394def xirr(flows):395 d0 = min(d for d, _ in flows)396 f = lambda r: sum(mp.mpf(a) / (1 + mp.mpf(r)) ** (mp.mpf((d - d0).days) / 365) for d, a in flows)397 guess = brentq(lambda r: float(f(r)), -0.99, 10)398 return F(mp.findroot(f, mp.mpf(guess)))399400xirr_cases = [401 dict(name="excel_xirr_example", flows=[(D(2008, 1, 1), -10000), (D(2008, 3, 1), 2750), (D(2008, 10, 30), 4250), (D(2009, 2, 15), 3250), (D(2009, 4, 1), 2750)]),402 dict(name="portfolio_with_contributions", flows=[(D(2022, 1, 3), -25000), (D(2022, 6, 15), -5000), (D(2023, 2, 1), 3000), (D(2023, 9, 30), -4000), (D(2024, 12, 31), 36250.75)]),403 dict(name="loss", flows=[(D(2023, 5, 1), -10000), (D(2024, 5, 1), 400), (D(2025, 5, 1), 8200)]),404]405out["xirr"] = [dict(name=c["name"], flows=[[d.isoformat(), a] for d, a in c["flows"]], rate=xirr(c["flows"])) for c in xirr_cases]406407def growing_pv(c, r, g, n, due):408 r, g = mp.mpf(r), mp.mpf(g)409 if due:410 return sum(mp.mpf(c) * (1 + g) ** k / (1 + r) ** k for k in range(n))411 return sum(mp.mpf(c) * (1 + g) ** (k - 1) / (1 + r) ** k for k in range(1, n + 1))412413out["growing_annuities"] = []414for (c, r, g, n, due) in [(1000, "0.08", "0.03", 25, False), (1000, "0.05", "0.05", 30, False), (1000, "0.05", "0.0500000001", 30, False),415 (500, "0.04", "0.07", 15, True), (250, "0.06", "-0.02", 40, False), (1000, "0.05", "0.05", 12, True)]:416 pv = growing_pv(c, r, g, n, due)417 out["growing_annuities"].append(dict(c=c, r=F(r), g=F(g), n=n, due=due, pv=F(pv), fv=F(pv * (1 + mp.mpf(r)) ** n)))418419rets = ["0.12", "-0.08", "0.25", "0.03", "-0.15", "0.19"]420prod = mp.mpf(1)421for x in rets: prod *= 1 + mp.mpf(x)422out["returns"] = dict(423 series=[F(x) for x in rets], arithmetic=F(sum(mp.mpf(x) for x in rets) / len(rets)), geometric=F(prod ** (mp.mpf(1) / len(rets)) - 1),424 cagr_10000_18500_7y=F((mp.mpf(18500) / 10000) ** (mp.mpf(1) / 7) - 1),425 annualized_35pct_2_5y=F(mp.mpf("1.35") ** (1 / mp.mpf("2.5")) - 1),426 real_7_nominal_3_inflation=F(mp.mpf("1.07") / mp.mpf("1.03") - 1),427 deflate_1000_25y_2pct=F(1000 / mp.mpf("1.02") ** 25),428 twr=F((mp.mpf(105000) / 100000) * (mp.mpf(118000) / 115000) * (mp.mpf(102000) / 108000) - 1),429 pv_annuity_due_500_6pct_15=F(sum(500 / mp.mpf("1.06") ** k for k in range(15))),430 fv_annuity_ord_500_6pct_15=F(sum(500 * mp.mpf("1.06") ** k for k in range(15))),431 pv_annuity_tiny_rate=F(sum(100 / (1 + mp.mpf("1e-10")) ** k for k in range(1, 361))),432)433434# ---------------------------------------------------------------- Options435def bsm(kind, S, K, r, q, s, T):436 S, K, r, q, s, T = map(mp.mpf, (S, K, r, q, s, T))437 d1 = (mp.log(S / K) + (r - q + s * s / 2) * T) / (s * mp.sqrt(T))438 d2 = d1 - s * mp.sqrt(T)439 if kind == "call":440 return S * mp.e ** (-q * T) * mp.ncdf(d1) - K * mp.e ** (-r * T) * mp.ncdf(d2)441 return K * mp.e ** (-r * T) * mp.ncdf(-d2) - S * mp.e ** (-q * T) * mp.ncdf(-d1)442443def bsm_float(kind, S, K, r, q, s, T): # second route: scipy.stats.norm in double precision444 d1 = (np.log(S / K) + (r - q + s * s / 2) * T) / (s * np.sqrt(T))445 d2 = d1 - s * np.sqrt(T)446 if kind == "call":447 return S * np.exp(-q * T) * norm.cdf(d1) - K * np.exp(-r * T) * norm.cdf(d2)448 return K * np.exp(-r * T) * norm.cdf(-d2) - S * np.exp(-q * T) * norm.cdf(-d1)449450opt_cases = [("42", "40", "0.10", "0", "0.20", "0.5"), ("100", "100", "0.05", "0.02", "0.25", "1"), ("100", "130", "0.03", "0.01", "0.35", "0.25"),451 ("100", "70", "0.03", "0.04", "0.15", "2"), ("50", "50", "-0.01", "0", "0.6", "0.08"), ("3500", "3600", "0.045", "0.015", "0.18", "1.5"),452 ("10", "25", "0.02", "0", "0.4", "0.5"), ("100", "100", "0.05", "0", "0.01", "1")]453out["bsm"] = []454for (S, K, r, q, s, T) in opt_cases:455 for kind in ("call", "put"):456 args = (S, K, r, q, s, T)457 p = bsm(kind, *args)458 g = dict(459 delta=mp.diff(lambda x: bsm(kind, x, K, r, q, s, T), mp.mpf(S)),460 gamma=mp.diff(lambda x: bsm(kind, x, K, r, q, s, T), mp.mpf(S), 2),461 vega=mp.diff(lambda x: bsm(kind, S, K, r, q, x, T), mp.mpf(s)),462 theta=-mp.diff(lambda x: bsm(kind, S, K, r, q, s, x), mp.mpf(T)),463 rho=mp.diff(lambda x: bsm(kind, S, K, x, q, s, T), mp.mpf(r)),464 )465 fl = tuple(map(float, args))466 pf = bsm_float(kind, *fl)467 assert abs(pf - F(p)) < 1e-9 * max(1.0, F(p)), (kind, args)468 lo = max((fl[0] * np.exp(-fl[3] * fl[5]) - fl[1] * np.exp(-fl[2] * fl[5])) * (1 if kind == "call" else -1), 0)469 iv = None470 if F(p) - lo > 1e-6: # implied volatility is only well conditioned away from the intrinsic bound471 iv = brentq(lambda v: bsm_float(kind, fl[0], fl[1], fl[2], fl[3], v, fl[5]) - F(p), 1e-6, 10, xtol=1e-15, rtol=8.9e-16)472 out["bsm"].append(dict(kind=kind, S=F(S), K=F(K), r=F(r), q=F(q), sigma=F(s), T=F(T), price=F(p), implied_vol=iv,473 **{k: F(v) for k, v in g.items()}))474475def crr(kind, american, S, K, r, q, s, T, n):476 dt_ = T / n477 u = np.exp(s * np.sqrt(dt_)); d = 1 / u478 p = (np.exp((r - q) * dt_) - d) / (u - d)479 disc = np.exp(-r * dt_)480 sign = 1 if kind == "call" else -1481 j = np.arange(n + 1)482 v = np.maximum(sign * (S * u ** (2.0 * j - n) - K), 0)483 for step in range(n - 1, -1, -1):484 v = disc * (p * v[1:step + 2] + (1 - p) * v[0:step + 1])485 if american:486 jj = np.arange(step + 1)487 v = np.maximum(v, sign * (S * u ** (2.0 * jj - step) - K))488 return float(v[0])489490out["crr"] = []491for (S, K, r, q, s, T, n) in [(50, 50, 0.10, 0.0, 0.40, 5 / 12, 5), (50, 50, 0.10, 0.0, 0.40, 5 / 12, 500), (42, 40, 0.10, 0.0, 0.20, 0.5, 1000),492 (100, 110, 0.03, 0.06, 0.30, 1.0, 400), (100, 95, 0.05, 0.02, 0.25, 0.75, 250), (36, 40, 0.06, 0.0, 0.20, 1.0, 2000)]:493 for kind in ("call", "put"):494 out["crr"].append(dict(kind=kind, S=S, K=K, r=r, q=q, sigma=s, T=T, steps=n,495 european=crr(kind, False, S, K, r, q, s, T, n), american=crr(kind, True, S, K, r, q, s, T, n),496 bsm=float(bsm_float(kind, S, K, r, q, s, T))))497498# ---------------------------------------------------------------- write499here = os.path.dirname(os.path.abspath(__file__))500path = os.path.join(here, "reference.json")501with open(path, "w") as fh:502 json.dump(out, fh, indent=1, sort_keys=True)503 fh.write("\n")504shutil.copyfile(path, os.path.join(here, "..", "Tests", "ORIXFinancialTests", "Resources", "reference.json"))505506# Human-readable spot checks of the textbook figures quoted in VALIDATION.md.507print("Canadian mortgage 300000 @5% semi-annual, 25y monthly:", out["loan_payments"][0]["payment"])508print("Hull BSM S=42 K=40 r=10%% sigma=20%% T=0.5: call %.4f put %.4f" % (out["bsm"][0]["price"], out["bsm"][1]["price"]))509print("Hull CRR American put 5 steps:", out["crr"][1]["american"])510print("Excel PRICE example:", out["bonds"][0]["clean"], " Excel XIRR example:", out["xirr"][0]["rate"])511print("IRRs two_irrs:", out["cashflows"][1]["irrs"], " three_irrs:", out["cashflows"][5]["irrs"], " none:", out["cashflows"][2]["irrs"])512print("cases:", {k: (len(v) if isinstance(v, list) else 1) for k, v in out.items() if k != "_meta"})513