What's New

This page provides a brief overview of the user-facing new features and improvements in the latest version of Atoti Market Risk.

For bug fixes and developer-facing changes, see our Changelog.

5.4.0-BETA 2


New features

Market Data API redesign

This release introduces a refined Market Data API, featuring enhancements that streamline the underlying data model for improved customization efficiency. This update significantly reduces the code needed to handle market data, improving overall development efficiency and implementation ease.

For details, see Market Data API.


DirectQuery support for connecting to a Databricks database has been added. Read more about DirectQuery and how to connect to a Databricks Database.

New engine

Atoti Market Risk 5.4.0 is powered by the latest version of Atoti Server and Spring. The result is a faster, safer application, which you can customize using the most modern stack available.

New adjustments in the VaR cube

You can now apply PnL vector adjustments for individual scenarios in the VaR cube. To request an adjustment, follow the usual sign-off process on the VaR cube and build a pivot table to display the PnL vector expanded by Scenario.



New features

MS SQL Server support for DirectQuery

We are excited to announce expanded support for data sources by now including MS SQL Server as a DirectQuery option. This enhancement empowers you to seamlessly tap into data stored within this widely used warehouse technology, enabling trend and historical analysis directly from Atoti Market Risk, bypassing the in-memory loading step. Read more about DirectQuery and how to connect to an MS SQL Server Database.

Simplified API and configuration

We continue working on improving the data model and customization options in Atoti Market Risk and in this release we are delivering a simplified data model for market data retrieval, as well as market data interpolation service. For details, see Market Data APIs.

Virtual hierarchies

You can now use the virtual hierarchy feature on Atoti Market Risk directly from the configuration.

Here’s how to define virtual hierarchies in the application.yaml file:

            - Trades@Booking


Analysis hierarchies can’t be virtual.



  • Upgraded to Atoti Server 6.0.9.



  • Fixed ladder computation: The ladder formula has been fixed. This changes the result of PnL and VaR computation that uses a sensitivity based on Ladders.
  • Added Volga Taylor VaR measures: Volga Taylor VaR measures have been added to the Solution.
  • Added statistical measures for interpolated market shift measures: The minimum, maximum, average, and percentile measures have been added for interpolated market shift measures.
  • Made ladder-based sensitivity measures optional: Ladder-based sensitivity measures can now be excluded from the configuration. If the configuration classes are excluded, no visible measures will be present in the cube. If the Solution is configured to use ladders as an input to PnL Explain and Taylor VaR calculations, the result will be NaN.
  • DirectQuery cube refresh is now available by REST service: It can be done with a post at the URL ‘http://localhost:10010/mr-application/directquery/refresh’.



  • Upgraded to Atoti Server 6.0.7.
  • Bug fixes and improvements.



Improved Sensitivity Cube configuration for testing: store, dimension, and measure configuration classes have been split to allow targeted testing of measure chains.



New features

DirectQuery support for Atoti Market Risk

DirectQuery support has been introduced for the VaR/ES, Sensitivity, and PnL Cubes. This feature empowers you to effortlessly access historical data stored in your warehouse directly from Atoti Market Risk.

The technology allows you to delve into the depths of your data and analyze trends, bypassing the in-memory loading step. This way, DirectQuery not only enables historical analysis but also optimizes system resources.

For more information, see DirectQuery.

  • Currently, DirectQuery does not support What-Ifs, Sign-Off, or data updates. These features are only available for the in-memory Cubes.
  • Today, we support integration with Snowflake. Stay tuned as we expand our compatibility with other warehouse technologies.


Atoti UI update

Atoti UI has been updated to 5.1. For more details on this version, see the Atoti UI documentation.

Sensitivity input file formats update

The SecondOrderLadder field has been removed from the Sensitivities files to simplify their structure:



  • Upgraded to Atoti Server 6.0.3.
  • Change to the way FX conversion is computed for PnL values in the VaR-ES cube.
  • Taylor VaR enhancements
  • Updated datasets
  • New bookmarks
  • TradeKey field has been introduced for Summary data.
  • Terminology change in documentation

New features

Taylor VaR enhancements

You can now use market shift data from a different date to the asOfDate for the calculation of Taylor measures. New measures display minimum, maximum, and average values of market shift vectors, and the value at a user-specified percentile of the vector.

VaR FX Conversion

This version changes the method by which PnL vectors are converted to a different currency within the VaR-ES cube. The mark-to-market (MTM) value of the trade is now included in the calculation to ensure the FX risk associated with the trade is taken into account. Details of the calculation can be seen here. This change impacts all measures that are based on the PnL vectors, including VaR, VaE, Expected Shortfall, and so on.


New sets of bookmarks have been added:

  • How It Works bookmarks show interim steps to help you understand calculations and input data, and find useful measures.
  • Daily Jobs bookmarks illustrate demo stories.
  • Additional Story-Telling Target Views bookmarks to compliment the other two sets.


The dataset has been updated:

  • Dataset moved 5 years forward (from 2018 to 2023) for AsOfDate, maturities, trade dates, any dates.
  • Full revaluation VaR is aligned with Taylor VaR.
  • Full revaluation PnL is close to PnL Explain.
  • Market data shifts are regenerated so that T-1 vector and T1 vector contain the same values except 1 value
  • Risk factors are consistent between Taylor VaR and Full revaluation VaR.
  • Taylor VaR market data adjusted.
  • Market data anomaly to support backtesting exception story.
  • Commodity risk factor shocks.

Terminology change

Product names have been updated across the documentation.

Why are we introducing these changes:

  • To make our product offering more readable.
  • To bring naming consistency to the software platform as a whole.
  • To pave the way for future products and features.


The changes in names do not have any impact on the content or the substance of the product. Only the designation changes.



  • Implementation of additional Taylor measures : Many of the VaR and similar measures that are available in the VaR-ES cube are now available in the Sensitivities Cube. These measures are derived from PnL vectors created with a Taylor expansion of sensitivities and market data.
    • For each combination of sensitivity - Cash, Correlation, CrossGamma, Delta, Gamma, Vanna, Vega, and Aggregate - and measure type - Expected Shortfall (ES), Expected Tail Gain (ETG), Value at Earnings (VaE), Value at Risk (VaR), Weighted Value at Earnings (WVaE), Weighted Value at Risk (WVaR) - the following measures are available:
      • Basic measure
      • Measure from the previous day
      • Day-to-day difference of the measure
      • Day-to-day difference of the measure expressed as a percentage
      • LEstimated version of the measure
    • For each of these measures, there is a version for which the confidence is controlled with a context value. There are also versions that use a fixed confidence. These fixed confidences are set by the confidence.levels application property.
    • The Sensitivies Cube now also has a Tail VaR measure (VaR with 100% confidence) and a Tail VaE measure (VaE with 100% confidence).
  • Removal of IGreekDescription interface : Removed IGreekDescription interface and replaced it with source- and hierarchy-specific configuration.

Known issues

PIVOT-5759 - Attempting to create partitions for reference stores causes a deadlock.



This is a UI-only release that has updated Accelerator-sdk and ActiveUI dependencies.

For details, see the Release notes.

Known issues

PIVOT-5759 - Attempting to create partitions for reference stores causes a deadlock.



  • New accelerator architecture: The structure of the MR Accelerator has been updated to enable greater flexibility and ease of use. New modules have been created encapsulating different calculation models, such as VaR, or PnL. These modules can be used as standalone libraries or the Accelerator can still be used as a cohesive customizable project. For an overview of the modules, see Maven modules. For details on the class changes, see Classes.
  • Accelerator name changed from ‘MRA’ to ‘MR’: The application is now called ‘MR’ (Market Risk) instead of ‘MRA’ (Market Risk Accelerator) to be aligned with the naming convention of the other ActiveViam accelerators.
  • MRACombinedCube renamed to MRCombinedCube To be in line with the name change of the application, the combined cube has been renamed.
  • Bookmarks updated to reflect the use of ‘MR’ instead of ‘MRA’: The ‘MRA’ folder and the ‘MRA’ server name have been replaced with ‘MR’ in the bookmarks.

Known issues

PIVOT-5759 - Attempting to create partitions for reference stores causes a deadlock.



  • Added Market Data Cube to display the market data, market shift values, and FX rate.
  • Improved Sign-off scopes to handle multiple hierarchies and level depths higher than 1.
  • Added What-If File Upload widget.
  • Replaced Spring profiles with properties in the application.yaml file.
  • Moved the cube toggling properties from the mr.properties file to the application.yaml file and renamed them.
  • Added support for Limits Module integration

Known issues

  • Updated 2022-09-12: PIVOT-5759 - Attempting to create partitions for reference stores causes a deadlock.

New features

Replaced Spring profiles with properties

The six Spring profiles used for controlling application modes have been removed and replaced with application properties:

  • azure-dataloading
  • local-file-system
  • standard
  • scalar-sensi
  • aggregated
  • enable-signoff

See application.yaml for a list of replacement properties.

Moved cube toggling properties to application.yaml file

The cube toggling properties in the mr.properties file have been removed:

Property Name Comment
enable.var.cube Moved to application.yaml and renamed.
enable.var.summary.cube Moved to application.yaml and renamed.
enable.sensi.cube Moved to application.yaml and renamed.
enable.sensi.summary.cube Moved to application.yaml and renamed.
enable.pnl.cube Moved to application.yaml and renamed.
enable.pnl.summary.cube Moved to application.yaml and renamed.
enable.common.cube Moved to application.yaml and renamed.
enable.market.cube Moved to application.yaml and renamed.

See application.yaml or the Migration Notes for a list of replacement properties.

Market Data Cube

The Market Data Cube has been added to display the market data, market shift values, and FX rate. Unlike the Sensitivities Cube that needs to have a risk factor associated to a sensitivity, the facts are directly linked to the risk factors.

See the Market Data Cube for details.

File Upload

The File Upload widget has been added, allowing you to create or append a WhatIf scenario by uploading CSV files into the cube.

See File Upload


  • Ease modification of Context values: The context values are regrouped in a single configuration for better customisation.

  • Handle multiple hierarchies and level depths higher than 1 for sign-off scopes: Until version 3.1.0, the scope of a Sign-off task could only be defined for the first level of one hierarchy. Now the scope definition can contain multiple hierarchies, and within each hierarchy, a level at a depth that can be higher than 1.


    The hierarchies and levels that can be used in the scopes of the Sign-off task must be part of the ones defined in configuration properties. See details in signoff.properties.

  • Possibility to define Sign-off DEE export templates by cube-measure pairs On top of being defined by cube names, DEE templates used for Sign-off export can now be defined by cube name and measures.



The MR 3.0.0-ui-update-1 release is a UI only release, and no server-side changes were made.


  • ActiveUI: Upgrade to 5.0.21.



  • Summary cubes that can be started along with the standard cubes have been created.
  • Harrell-Davis VaR is available on VaR-ES cube with incremental and L estimators.
  • Extraction templates are now used for exports instead of a programmatic export definition.

Known issues

  • Updated 2022-09-12: PIVOT-5759 - Attempting to create partitions for reference stores causes a deadlock.
  • Update 2022-06-07: We’ve identified a bug in the Frequent Value Compression ActivePivot feature (new since AP 5.10). MRA 3.0.0 depends on ActivePivot 5.10.6 and is therefore impacted. We strongly recommend disabling Frequent Value Compression by setting the following property when launching ActivePivot: -Dactiveviam.chunk.compression.enabledFrequentValue=0. The next maintenance release of MRA will include an upgrade to ActivePivot, which resolves this issue.

New features

Real time feature

Cubes are updated in real time when any of the following occur:

  • a currency change
  • a market data change
  • any main store change

Summary cubes

Import cube profiles are now replaced by summary cubes. Principally, the same application can run the default and import/summary cubes together. See Activating and de-activating cubes.

For details of the ISO configuration for migrating the spring profiles, see the Migration Notes.

Context values

The following context values are now replaced by specific slicing hierarchies:

  • ReferenceCurrency
  • DynamicTenorSets
  • DynamicMaturitySets
  • DynamicMoneynessSets

For details, see the Migration Notes.


  • Bucket-level comparator: Doubles are now used for the level comparator for buckets instead of integers to define the number of dates per time period (year, month, week)
  • Active-Pivot 5.10: The Accelerator is upgraded to the 5.10.6 version.



This release follows the developments of the 2.1.0 Alpha for the Sign-off features.

2.1.0 Alpha

Known issues

  • Updated 2022-09-12: PIVOT-5759 - Attempting to create partitions for reference stores causes a deadlock.
Issue Key Details
MRA-1138 When a cube-level adjustment request is sent to the MRA, the slicing hierarchies not explicitly present in the filters are not handled.
MRA-1091 Create Harrell-Davis VaR
MRA-1139 Roll-over of PnL vectors: The Pnl values are not shifted properly.


  • Azure data loading support: MRA supports data loading from Azure Blob storage.
  • Cube configuration: VaR, sensitivities and PnL cubes can now be disabled using properties.
  • Cube-level adjustments: MRA handles adjustments of aggregated measures.
  • Fact-level adjustments: Overrides and scaling are now handled in addition to add-ons.
  • New approval mechanism for approvals: The sign-off approval mechanism now uses a specific analysis hierarchy instead of the epoch hierarchy.
  • Harrell-Davis VaR: Harrell-Davis VaR has been implemented.
  • Day count computation: The day count computation is now computed differently.

New features

Calendar Improvement

The day count computation now uses MraSimpleTenorConverter and DayCount30on360ISDA instead of using LegacyTenorConverter and DayCountA365Fixed. This change results in almost the same duration.

Component Use Behavior Convention
LegacyTenorConverter Tenor to Date converter Convert a tenor by using 30 days for a month and 360 days for a year. 30/360
MraSimpleTenorConverter Tenor to Date converter Convert a tenor using the exact computation. Exact
DayCountA365Fixed Date interval to days converter Convert a date interval to an exact number of days. A001
DayCount30on360ISDA Date interval to days converter Convert a date interval to a number of days by using 30 days for a month and 360 days for a year. A005

The two conventions produce almost the same result, except for the pillars represented by:

  • A number of days above 30D.
  • A number of weeks above 4W.
  • An exact date.
  • The end of month (31st is assumed to be a 30th, February end of month is assumed to be a 28th (or 29th)).

The year fraction will be interpreted as Year / Month / Day by assuming that 1M = 1/12Y such that:

  • 0.5Y=6M
  • 0.25Y=3M
  • 1.5Y=1Y6M

See Indicator: Method of Interest Computation Indicator for the convention.

This change highly impacts the Theta metric.

You can change the convention if required.

Azure data loading support

MRA supports loading data from Azure Blob storage. The azureCredentials.properties file defines the necessary credential information for connecting to an Azure Blob container.

Cubes enabling/disabling

VaR, sensitivities and PnL cubes can now be activated or disabled from the configuration file mr.properties.

Cube-level adjustments

Cube-level adjustments are now supported. The logic relies on the concept of location digest and uses a look-up in the datastore to retrieve the overriding value for a measure. The overridden measures have the “_Adjusted” suffix.

Fact-level adjustments

Overrides and roll-over are now supported.

Harrell-Davis VaR

Harrell-Davis VaR is now implemented.

New approval mechanism for approvals

The sign-off approval mechanism now uses a specific analysis hierarchy instead of the epoch hierarchy.


Known issues

  • Updated 2022-09-12: PIVOT-5759 - Attempting to create partitions for reference stores causes a deadlock.

Fixed Issues


New features

Usage of the common library

Several class are now shared with other Accelerators. They are located in the file ‘shared*.jar’.

The Maven dependency is:






The package name is : com.activeviam.accelerator.common

Sign-off profile

The enable-signoff Spring Profile has been added to the Market Risk Accelerator, grouping all Sign-off functionality in a single toggled configuration.

Adjustments 2.0

Interfaces have been integrated allowing interaction with the Sign-Off Module.

For details about supported adjustments, see Supported Sign-off Adjustments.

Corporate Actions

The dividends, coupons, stock splits or merge are now taken in account for PnL calculations.

For details, see Corporate Actions

New measures


The correlation measure is a two-risk factor sensitivity that relies on a single market data. For a correlation RF1 * RF2, the market data name will be RF1/RF2. The market data separator is defined by the property correlation.risk.level.split. Apart from these specifics, it behaves like a delta metric. It can handle one maturity.


The cross-gamma measure is a two-risk factor sensitivity that relies on two distinct market data. The PnL explain is calculated by applying the two PnL formulas PnL = f_RF2(f_RF1(crossGamma, MD_1_d, MD_1_d-1), MD_2_d, MD_2_d-1) sequentially. This measure does not handle any tenors, maturities or moneyness.


  • Upgrade to AP5.9 : The Accelerator now uses Active Pivot 5.9.x, which also means that only Java 11 is supported.
  • JUnit 5: Most of the tests now use JUnit 5.
  • Data Connectors: The DEE feature is now integrated into an import/export toolbox library called “Data Connectors” prefixed with ‘com.activeviam.io’.
  • Datastore Helper 2.0 : The Datastore Helper tool for defining and customizing the datastore schema has been upgraded to version 2.0.
  • Spring configuration : The Spring configuration templates for cube and datastore definition are in common with FRTB.

Previous releases


Known issues

  • Updated 2022-09-12: PIVOT-5759 - Attempting to create partitions for reference stores causes a deadlock.
Issue Key Details
MRA-806 Real-time updates in MarketData.csv file are not taken into account.
MRA-1028 MRA SDK has an unmet dependency that causes build to fail. It is fixed now and the new version is uploaded. The old one is marked as broken.

New features


Bug fixes

  • Bug fix in export feature: In cube and aggregated exports, cube measures were not put in the output columns. This has been fixed.
  • Parent/child data load: The system now waits for parent-child processing to complete before loading main stores.


Known issues

  • Updated 2022-09-12: PIVOT-5759 - Attempting to create partitions for reference stores causes a deadlock.

New features



  • Minor bug fixes
  • Configuration of sparse vectors: Sparse vectors can be used on any double[] vector and activated below a defined threshold.


Known issues

  • Updated 2022-09-12: PIVOT-5759 - Attempting to create partitions for reference stores causes a deadlock.
  • The Accelerator doesn’t handle the real-time update on market data.

New features


The sign-off is handled by an external application, only the export services stay under the responsibility of MRA.


The common Accelerator export tool is embedded in MRA.

Market Data Retrieval Service

The Market Data computation and retrieval is handled by a specific spring bean that is customisable at will by the user. For details, see Market Data Retrieval Service.

Cash metric

A new metric that computes the risk linked to a foreign currency position.

Theta metric

A new metric that takes in account the time decay on the PnL.

Incremental metrics

The whole incremental metrics has been reviewed and is customizable from the property file, with new reference metrics introduced: - Compared to the top location - Compared to a specific level - Slicing on any hierarchy.

Active Pivot Core

Upgrade to AP 5.8.14


Known issues


New features

Product Control

Product control features were added in a new separate module which can be used if those features are required.




Known issues


New features



Back-port bug fixes from MRA 1.8.0


Known issues


New features

Data Load Controller

The Data Load Controller has now been integrated into the Accelerator.  It provides the ability to better manage loading (and unloading) data.  A REST service is provided, so that your data orchestration service (the process that is responsible for delivering data to the Accelerator), can trigger the loading of data into ActivePivot.

The REST service will allow loading (and unloading) of topics (or groups of topics, called aliases) by scope.  Currently the scope allows the specification of an as-of date, however it can easily be customized to support, for example, desks or legal entities.  The scope helps the data load controller find the data when loading, and becomes the criteria for removing data when unloading.  For example, it is possible to load (or unload) all Delta sensitivities (the topic alias) for the as-of date 2019-08-09 (the scope).




Known issues


New features



Fixed weighted VaR and weighted ES computations in the case PnL vectors are loaded from the oldest to the latest date


Known issues


Calculation error for weighted measures

There is a calculation error for the weighted measures when the PnL data are present from oldest to newest in the input vectors (MRA-783)

Best scenario measure is incorrectly implemented

The best scenario measure is incorrectly implemented (MRA-785)

ActivePivot incompatible with latest OpenJDK ActivePivot

5.8.7 (and all earlier versions) are incompatible with the latest version of OpenJDK 11 (Version 11.0.6, released 2020-01-15).  This is due to a breaking change in OpenJDK (JDK-8211919)

New features

Support for Java 8 and Java 11

Starting from version 1.6.0, the MRA is delivered in 2 versions, one that supports Java 8 and one that supports Java 11.

Taylor VaR

First version of Taylor VaR to compute VaR from market data shifts and sensitivities.