sipQuant

Quantitative library for physically-settled commodity markets.

---

About

sipQuant is a pure-NumPy quantitative stack for building, maintaining, and trading against physically-settled commodity price indices in thin, opaque markets. It covers the full pipeline from raw trade data to OTC structured products.

It is the internal quantitative infrastructure of SIP Global, open-sourced under GPL-3.0. All mathematics are implemented in pure NumPy.

About SIP Global

Systematic Index Partners (SIP Global) brings liquidity to overlooked markets that move the world. We apply systematic rigour and modern trading standards to create transparent pricing, real liquidity, and durable market function in physically-settled commodity markets.

Tota Ad Manum.

Visit SIP Global

---

Installation

pip install sipQuant

Requires Python 3.10+ and NumPy 1.23+. No other dependencies.

---

Quick Start

The full SIP pricing pipeline in one block:

import sipQuant as sq
import numpy as np

# 1. Ingest a physical trade
trade = sq.schema.TradeRecord(
    date='2026-03-14',
    price=187.50,
    volume=500.0,
    grade='premium_bale_14pct_moisture',
    origin='lethbridge_ab',
    destination='red_deer_ab',
    counterpartyId='CP_ANON_004',
)
assert sq.schema.validate(trade) == []

# 2. Decompose price seasonality (weekly hay prices)
prices = np.array([182.0, 184.5, 187.0, 189.0, 185.5, 183.0, 186.5])
decomp = sq.commodity.seasonality(np.arange(7), prices, period=52)

# 3. Build a local forward curve
fwd = sq.commodity.localForwardCurve(
    spotPrice=187.50,
    tenor=np.array([0.0, 0.25, 0.50, 0.75, 1.0]),
    r=0.046,
    convYield=0.028,
    storageCost=0.020,
    basisAdjustment=-2.50,
)
curve = sq.schema.ForwardCurve(
    tenors=fwd['tenors'],
    prices=fwd['forwards'],
    baseDate='2026-03-14',
    market='alberta_hay_premium',
)

# 4. Price a commodity swap against the index curve
swap = sq.otc.commoditySwap(
    fixedPrice=190.0,
    indexCurve=curve['prices'],
    notional=1000.0,
    schedule=curve['tenors'],
    r=0.046,
)
print(f"Swap NPV: ${swap['npv']:,.2f}")

# 5. Calculate an IOSCO-aligned index
spec = sq.schema.IndexSpec(
    name='SIP-AHI-001',
    version='1.0',
    constituents=['premium_bale_14pct_moisture', 'feed_grade'],
    weightsMethod='volume',
    rollRule='monthly_last_business_day',
    effectiveDate='2026-01-01',
)
result = sq.index.calculateIndex([trade], spec, '2026-03-14')
audit  = sq.index.auditTrail(result, spec)
print(f"Index value: {result['indexValue']:.2f}  Checksum: {audit['checksum']}")

---

Design Philosophy

• Pure NumPy — all mathematics implemented without pandas, scipy, or statsmodels
• Vectorised operations — performance-first for simulation-heavy workflows
• Schema-validated inputs — every function accepts validated dict objects from sipQuant.schema
• Lightweight outputs — all functions return plain dicts or NumPy arrays; convert to DataFrame as needed
• Production-ready — IOSCO-aligned audit trail, restatement procedures, and margin estimation built in

---

Module Reference

Core Commodity Stack

schema — Data Contract Layer

The foundation of sipQuant. Every pricing, risk, and index function accepts schema objects.

Object	Purpose
PriceSeries	Regular price time series (broker quotes, exchange settlements)
SparsePriceSeries	Irregular physical trade observations (thin/weekly markets)
TradeRecord	Single physical trade (price, volume, grade, origin, destination)
QuoteSheet	OTC broker quote with bid/ask/mid and tenor
ForwardCurve	Forward curve for pricing and hedging
OTCPosition	Live OTC position for dealer book aggregation
IndexSpec	Immutable index methodology specification
validate()	Universal validation — returns list of error strings

commodity — Physical Pricing Primitives

Function	Use Case
seasonality()	STL-like decomposition: trend + seasonal + residual
convenienceYield()	Implied convenience yield from spot/futures pair
basis()	Local cash price minus benchmark reference
gradeAdjustment()	Quality-adjusted price from grade factor dict
transportDifferential()	Delivered price including logistics costs
localForwardCurve()	Forward curve from spot, carry, and convenience yield
rollingRollCost()	Roll cost estimated from forward curve shape

index — IOSCO-Aligned Index Infrastructure

Function	Use Case
calculateIndex()	VWAP-weighted index value from trade records
auditTrail()	IOSCO-aligned audit record with checksum
restatement()	Documented index correction procedure
rollSchedule()	Roll date generation (monthly/weekly/quarterly)
proxyRegression()	OLS or Huber regression for sparse-market proxy values
backtestIndex()	Historical index values, returns, volatility, drawdown

otc — OTC Instrument Pricing

Function	Use Case
commoditySwap()	Fixed-float commodity swap NPV and Greeks
asianSwap()	Arithmetic average price swap NPV
collar()	Long cap + short floor (buyer protection)
physicalForward()	PV of a physically-settled forward contract
swaption()	Black's formula swaption on forward swap rate
asianOption()	Monte Carlo Asian option with antithetic variates

book — Dealer Book Management

Function	Use Case
netGreeks()	Aggregate Greeks across all positions
hedgeRatios()	Delta-hedge units to flatten book delta
pnlAttribution()	Taylor expansion P&L decomposition
scenarioShock()	Pre-event scenario analysis (WASDE, weather, quota)
bookSummary()	Position summary by instrument type
marginEstimate()	Initial and variation margin estimation

liquidity — Thin-Market Risk

Function	Use Case
liquidityAdjustedVar()	LVAR following Bangia et al.
marketImpact()	Almgren-Chriss permanent + temporary impact
optimalExecution()	AC optimal liquidation schedule
thinMarketScore()	Liquidity score [0,1] for market quality assessment
concentrationRisk()	HHI position concentration
optimalLiquidation()	TWAP/VWAP cost estimation

---

Quantitative Toolkit

sim — Price Process Simulation

gbm, ou, levyOu, ar1, arma, markovSwitching, arch, garch, heston, cir, vasicek, poisson, compoundPoisson, simulate

fit — Model Calibration

fitGbm, fitOu, fitLevyOU, fitAr1, fitArma, fitGarch, fitHeston, fitCir, fitVasicek, fitJumpDiffusion, fitCopula, fitDistributions, aic, bic

options — Option Pricing

blackScholes, binomial, trinomial, asian, binary, spread, barrier, simulate, impliedVol, buildForwardCurve, bootstrapCurve

econometrics — Statistical Testing

ols, whiteTest, breuschPaganTest, durbinWatson, ljungBox, adfTest, kpssTest, grangerCausality

bootstrap — Curve & Surface Construction

curve, surface, zeroRateCurve, forwardCurve, discountCurve, yieldCurve, volSurface, creditCurve, fxForwardCurve, inflationCurve

---

Portfolio & Risk

portfolio — Portfolio Optimisation

blackLitterman, meanVariance, minVariance, riskParity, equalWeight, maxDiversification, tangency, maxSharpe, efficientFrontier, hierarchicalRiskParity, minCvar

risk — Risk Metrics

parametricVar, historicalVar, parametricCvar, historicalCvar, expectedShortfall, drawdown, maxDrawdownDuration, calmarRatio, sharpeRatio, sortinoRatio, omegaRatio, treynorRatio, informationRatio, modifiedVar, hillTailIndex, beta, downsideDeviation, ulcerIndex, painIndex, tailRatio, capturRatio, stabilityRatio

distributions — Distribution Fitting

fitNormal, fitLognormal, fitExponential, fitGamma, fitBeta, fitT, fitMixture, moments, ksTest, adTest, klDivergence, jsDivergence, quantile, qqPlot

dimension — Dimensionality Reduction

pca, lda, tsne, ica, nmf, kernelPca, mds, isomap

factor — Factor Models

famaFrench3, carhart4, apt, capm, estimateBeta, estimateFactorLoading, rollingBeta, pcaFactors, factorMimicking, jensenAlpha, treynorMazuy, multifactor

ml — Machine Learning

regressionTree, decisionTree, predictTree, isolationForest, anomalyScore, kmeans, knn, naiveBayes, randomForest, gradientBoosting, pca, lda, logisticRegression

---

Converting to Pandas or Polars

All functions return NumPy arrays or dicts. Convert as needed:

import pandas as pd
import polars as pl
from sipQuant import sim

paths = sim.gbm(mu=0.1, sigma=0.2, nSteps=252, nSims=100)
df_pandas = pd.DataFrame(paths.T, columns=[f'sim_{i}' for i in range(100)])
df_polars = pl.DataFrame(paths.T)

---

Documentation

Full documentation: sipquant.readthedocs.io

---

License

This project is licensed under GPL-3.0. See LICENSE for details.

---

Contact

For inquiries: packages@sipglobally.com

Brayden Boyko — brayden@sipglobally.com