Strata, Open Source Market Risk
3 likes1,058 views
Strata is an open-source market risk library designed for use in the finance industry, emphasizing market risk analytics, derivatives, and trading. It offers a lightweight Java library under the Apache v2 license and features a robust API for pricing and risk calculations, supports various asset classes, and provides built-in conventions. Developed by OpenGamma, Strata aims to bring open-source values to finance and includes features for cloud-based calculations and commercial support options.
Strata, Open Source Market Risk
- 1. STRATA Open Source Market Risk
- 2. STEPHEN COLEBOURNE • Java Champion, regular conference speaker • Best known for date & time - Joda-Time and JSR-310 • More Joda projects - http://www.joda.org • Major contributions in Apache Commons • Blog - http://blog.joda.org • Worked at OpenGamma for 6 years
- 3. OPENGAMMA • Founded in 2009 • Financed by Venture Capital • Mission to bring Open Source values to finance industry • Focus on Market Risk Analytics, Derivatives and Clearing • Cloud-based offering coming soon
- 4. MARKET RISK
- 5. MARKET RISK • Providing pricing and analytics on a financial portfolio • “present value (NPV) of an interest rate swap” • “PV01 of a forward rate agreement (FRA)” • “vega and gamma of an FX vanilla option” • “examine the portfolio against a set of scenarios”
- 6. COMPONENTS FOR MARKET RISK • Pricing/Analytic models • Trade representations - for each supported asset class • Market data representations - quotes, curves, surfaces, etc. • Calibration - curves, surfaces, etc. • Market data management and Scenario creation • Reference data - holiday calendars, securities • Basics - schedules, day counts, currencies, etc.
- 7. BUILD, BUY OR OPEN SOURCE? • Build it in-house • Buy from a vendor • Open Source • QuantLib (C++, with exports to other languages) • Strata (Java/JVM)
- 8. STRATA
- 9. OVERVIEW • Open Source - Apache v2 license • Lightweight, easy-to-use library • Just jar files - no servers or databases needed • Released in Maven Central • Foundation for OpenGamma commercial products
- 10. Commons Trade Model Market Data Pricing & Analytics Measures Data Mgmt Business day conventions, currencies, day counts, holidays, identifiers, indices, roll conventions, time-series Products, trades, securities Built-in pricers Sensitivities Curve framework Interpolators Calculation API, scenarios, job execution Calc functions Market Data API Data sourcing Scenarios Live data providers Data sources Loaders Source integration Holiday calendars Day counts Currencies Asset classes Indices Pricers Market data types Perturbations Config types Filters Measures Reporting Simple, yet powerful, reporting facilities Report templates
- 11. QUICK START • Built in examples • Up and running in minutes • Hard coded reference data • Holidays, indices, conventions • Simple command line reporting
- 13. JAVA 8 • Early decision to use Java 8 for Strata • Features very beneficial for Market Risk Analytics • Date and Time • Streams and Lambdas • Methods on interfaces • Affects both macro-level design and micro-level code
- 14. DEPENDENCIES • Guava • Joda-Beans • Joda-Convert • SLF4J • Commons-Math • Colt
- 15. MODULES • strata-collect - low-level, arrays, time-series, IO, tuples • strata-basics - holidays, schedules, indices, reference data • strata-data - market data containers • strata-market - market data structures - curves, surfaces, etc • strata-product - trades, products, securities • strata-loader - data loaders from csv and xml • strata-pricer - analytic pricers • strata-calc - calculation engine, scenarios, market data building • strata-measure - high-level measures, potentially multi-scenario
- 16. TRADES
- 17. TRADES • Trades are simple immutable beans (data objects) • Built using Joda-Beans • Use builder or static factory to create • Real properties and methods with Javadoc
- 18. JODA-BEANS • Source code generator/regenerator • Just write the fields and add a couple of annotations • Joda-Beans generates additional high-quality source code • Mutable and immutable beans • Provides C# style properties • Easy and fast serialization to XML, JSON, Binary
- 19. JODA-BEANS @BeanDefinition public final class Person implements ImmutableBean { @PropertyDefinition(validate = “notNull”) private final String forename; @PropertyDefinition(validate = “notNull”) private final String surname; // autogenerated getters, equals, hashCode, // toString, builder, metabean }
- 20. TRADES AND PRODUCTS • Trade • Trade date • Trade identifier • Counterparty • Product • Financial details of the trade • Effective/Termination date, notional, rate, index, etc.
- 22. FRA // create Fra using builder Fra fra = Fra.builder() .buySell(BuySell.BUY) // Buy .index(IborIndices.GBP_LIBOR_3M) // GBP LIBOR 3M .notional(10_000_000) // 10 million GBP .fixedRate(0.0085) // 0.85% .startDate(LocalDate.of(2016, 9, 14)) .endDate(LocalDate.of(2016, 12, 14)) .build();
- 23. FRA // create FraTrade Fra fra = ... TradeInfo info = TradeInfo.builder() .tradeDate(LocalDate.of(2016, 6, 14)) .id(StandardId.of(“Trade”, “123456”)) .counterparty(StandardId.of(“Party”, “654321”)) .build(); FraTrade trade = FraTrade.of(info, fra);
- 24. CONVENTIONS • Most OTC trades follow market conventions • Strata includes definitions of some of these conventions • Avoids repetitive code
- 25. FRA CONVENTION // create FRA from a convention FraTrade trade = FraConvention.of(IborIndices.GBP_LIBOR_3M) .createTrade( LocalDate.of(2015, 7, 14), // Trade date Period.ofMonths(2), // start in 2 months BuySell.BUY, // Buy 10_000_000, // 10 million GBP 0.012, // 1.2% ReferenceData.standard()); // Holiday calendars
- 26. NOTES • Built in conventions for FRAs • Built in conventions for 18 ibor-like indices, such as GBP LIBOR • Built in conventions for 12 overnight indices, such as GBP SONIA • Built in conventions for currencies • Built in holiday calendar data, suitable for evaluation
- 27. SWAP • Flexible interest rate swap • Fixed legs support variable interest rates and known amounts • Float legs support Ibor, Overnight and Inflation • Stubs support fixed, floating, interpolated and known amount • Support for variable notional, gearing and spread • Conventions and Templates available
- 30. SWAP // create swap from a convention SwapTrade trade = FixedIborSwapConventions.GBP_FIXED_1Y_LIBOR_3M .createTrade( LocalDate.of(2015, 7, 14), // Trade date Tenor.TENOR_10Y, // 10 year swap BuySell.BUY, // Buy 10_000_000, // 10 million GBP 0.014, // 1.4% ReferenceData.standard()); // Holiday calendars
- 31. SECURITIES • Two approaches supported • Define security as and when needed • Setup reference data map of securities • Provides ability to use, or avoid, a big security master
- 32. SECURITIES AS PRODUCTS IborFutureOption Trade IborFutureOption IborFuture Trade ProductProductTrade
- 33. SECURITIES AS REFERENCE DATA Trade IborFutureOption IborFuture Security IborFutureOption Security IborFuture Security SecurityTrade ReferenceData used to lookup SecurityId ReferenceData used to convert Security to Product
- 34. ASSET CLASSES • Swaps, Swaptions, DSF, CMS, Cap/Floor • FRA, STIR futures, STIR future options • Bond, Bond futures, Bond future options • FX forward, NDF, FX swap, vanilla option, single barrier option • CDS • Term deposit, Bullet payment • Generic security, ETD future, ETD option • Additional asset classes may be for commercial customers only
- 35. PRICERS
- 36. PRICERS • Lower-level analytics API • Provides ability to calculate PV, sensitivities, greeks, etc. • Explain facility to understand how result was calculated • Operates on resolved trades/products
- 37. RESOLVING • Resolving the trade requires reference data • Locks in dates to the current holiday calendar rules • Standard reference data contains hard coded holiday rules // resolve a swap SwapTrade trade = … ReferenceData refData = ReferenceData.standard(); ResolvedSwapTrade resolved = trade.resolve(refData);
- 38. RESOLVING Swap SwapLegSwapTrade Trade resolved using ReferenceData Resolved SwapTrade Resolved Swap Resolved SwapLeg Resolved leg contains full list of accrual/payment periods, stubs and notional exchange
- 39. USING A PRICER • Stateless - takes resolved trade and any necessary market data • Calculates for one trade and one set of market data • Can usually price at trade or product level
- 40. USING A PRICER // obtain the swap and market data to price against ResolvedSwapTrade trade = ... RatesProvider market = ... // calculate the present value MultiCurrencyAmount pv = DiscountingSwapTradePricer.DEFAULT .presentValue(trade, market);
- 41. MARKET DATA
- 42. MARKET DATA • Support for all kinds of market data • Built in classes for FX, quotes, curves, surfaces, etc. • Many types can be loaded from CSV • Multi-curve rates calibration • Scenarios, stored efficiently as arrays
- 43. RATES PROVIDER • RatesProvider is a single, coherent, set of market data • FX rates, Discount factors, Ibor rates • Overnight rates, Inflation price indices, Historic fixings // get discount factors for GBP DiscountFactors df = ratesProvider.discountFactors(GBP); double factor = df.discountFactor(date);
- 44. MARKET DATA • MarketData is a container of market data • Hash-map like • Keys are MarketDataId<T> // get curve by identifier CurveId id = CurveId.of(“Default”, “USD-DSC”); Curve curve = marketData.getValue(id);
- 45. SCENARIO MARKET DATA • ScenarioMarketData is a container of scenario data • Hash-map like, where values are arrays • Keys are MarketDataId<T> // get curves by identifier CurveId id = CurveId.of(“Default”, “USD-DSC”); MarketDataBox<Curve> curves = scenarioData.getValue(id); // process using a stream (for example) curves.stream().forEach( curve -> … );
- 46. MARKET DATA LOOKUP • Market data containers hold arbitrary sets of market data • May hold multiple USD discounting curves • RatesMarketDataLookup is used to select a coherent set CurveId usdDscId = CurveId.of(“Default”, “USD-DSC”); CurveId usdLiborId = CurveId.of(“Default”, “USD-LIBOR”); // map currency/index to curve RatesMarketDataLookup md = RatesMarketDataLookup.of( ImmutableMap.of(Currency.USD, usdDscId), ImmutableMap.of( IborIndices.USD_LIBOR_3M, usdLiborId), IborIndices.USD_LIBOR_6M, usdLiborId));
- 47. COMBINATIONS MarketData RatesProviderRatesMarketDataLookup+ = • Many curves • Keyed by curve ID • Map currency to curve ID • Map index to curve ID • DF by currency+date • Rate by index+date
- 48. MARKET DATA BUILDING • Can create market data manually, loading from CSV or by factory • MarketDataFactory can • Query quotes from a simple provider interface • Query time-series from a simple provider interface • Calibrate • Create scenarios by shifting/bumping • See SwapPricingWithCalibrationExample
- 49. MEASURES
- 50. MEASURE-LEVEL API • Higher-level than pricers • Stateless - takes resolved trade and any necessary market data • Calculates for one trade and one or more sets of market data • ie. supports scenarios • Only operates on trades, not products • Scaled output • eg. PV01 in basis points
- 51. USING THE MEASURE-LEVEL API // obtain the swap and market data to price against ResolvedSwapTrade trade = ... RatesProvider market = ... // calculate the present value MultiCurrencyAmount pv = SwapTradeCalculations.DEFAULT .presentValue(trade, market);
- 52. USING THE MEASURE-LEVEL API // obtain the swap and market data to price against ResolvedSwapTrade trade = ... RatesMarketDataLookup lookup = ... ScenarioMarketData market = ... // calculate the present value for many scenarios MultiCurrencyScenarioArray scenarioPv = SwapTradeCalculations.DEFAULT .presentValue(trade, lookup, market);
- 53. CALCULATIONS
- 54. CALCULATION-LEVEL API • Highest-level API • Calculates for many trades and one or more sets of market data • ie. supports scenarios • Optional currency conversion • Multi-threaded • Results can be received asynchronously
- 55. CALCULATION-LEVEL API • Calculation API result is a grid • Rows are trades, positions, or similar • Columns are measures, such as PV, PV01, Par rate • Mixed portfolio of trades (PV for Swap, FRA and future in one call) NPV (USD) NPV (GBP) PV01 (USD) Par rate Trade 1 - Swap 13,487.25 10,176.72 12.7365 0.23 Trade 2 - Swap -34,276.73 -27,273.28 76.2725 0.24 Trade 3 - FRA 12,835.26 9,263.75 -26.8367 0.31 Trade 4 - STIR -965.76 -754.23 1.2676 0.52
- 56. CALCULATIONS • CalculationRunner is entry to Calculation API • Provides a multi-threaded executor • Also allow callers to use their own executor // obtains a multithreaded runner try (CalculationRunner runner = CalculationRunner.ofMultiThreaded()) { // use the runner }
- 57. RULES • CalculationRules defines how to calculate • Functions mapping from trade type to code • Reporting currency • Market data lookup // setup the rules CalculationRules rules = CalculationRules.of( StandardComponents.calculationFunctions(), Currency.USD, ratesMarketDataLookup);
- 58. CALCULATIONS • Each column defined by Column • Measure specifies what to calculate • Can control reporting currency per column // specify the columns List<Column> columns = ImmutableList.of( Column.of(Measure.PRESENT_VALUE), Column.of(Measure.PRESENT_VALUE, Currency.GBP), Column.of(Measure.PV01_CALIBRATED_SUM), Column.of(Measure.PAR_RATE));
- 59. CALCULATIONS • Calculation runner is stateless • Pass in all inputs, get back results • Separate API allows results to be received asynchronously // calculate the results Results results = runner.calculate( rules, // How to calculate trades, // Trades to process columns, // Columns, eg PV, PV01, par rate marketData, // Market data referenceData); // Reference data
- 60. CALCULATION-LEVEL API NPV (USD) NPV (GBP) PV01 (USD) Par rate Trade 1 - Swap 13,487.25 10,176.72 12.7365 0.23 Trade 2 - Swap -34,276.73 -27,273.28 76.2725 0.24 Trade 3 - FRA 12,835.26 9,263.75 -26.8367 0.31 Trade 4 - STIR -965.76 -754.23 1.2676 0.52 ... ... ... ... ...
- 61. SUMMARY
- 62. STRATA v1.2 • Trades are immutable beans • Pricing/risk logic is stateless, and separate from the trades • Three levels of pricing/risk API • Pricer - one trade, one set of market data • Measure - one trade, one or many sets of market data • Calc - many trades, one or many sets of market data
- 63. STRATA v1.2 • Modern market risk library in Java 8 • Lightweight and easy-to-use, lots of examples • Good asset class coverage, lots of built-in conventions • Open source, Apache v2 license • Foundation of commercial products • cloud-based calculations for derivatives trading • Commercial support available from OpenGamma http://strata.opengamma.io/
- 64. THANK YOU