	US 11,908,006 B2
	Message elimination in multi-model risk correlation system
Florian Huchedé, Chicago, IL (US); Robert Hayes, Naperville, IL (US); Abdoulaye Gory, London (GB); and David Henri Nicolay, London (GB)
Assigned to Chicago Mercantile Exchange Inc., Chicago, IL (US)
Filed by Chicago Mercantile Exchange Inc., Chicago, IL (US)
Filed on Dec. 20, 2018, as Appl. No. 16/227,674.
Prior Publication US 2020/0202433 A1, Jun. 25, 2020
Int. Cl. G06Q 40/04 (2012.01); G06F 17/16 (2006.01); G06Q 40/06 (2012.01)

CPC G06Q 40/04 (2013.01) [G06F 17/16 (2013.01); G06Q 40/06 (2013.01)]

16 Claims

1. A computer implemented method for cross asset correlation and margining in a trading system, comprising:

receiving, by a processor, portfolio data for a plurality of portfolios;

storing the portfolio data in memory, the memory implemented via one or more storage operations which implement local storage control via a rolling storage scheme storing portfolio data for a set of most recent transactions, thereby filtering the stored portfolio data to only the most recent transactions via the storage operations;

calculating a first plurality of simulations for the each of the plurality of portfolios;

performing a non-overlaid calculation by determining, by the processor, a single portfolio performance vector for each of the plurality of portfolios based on a portfolio specific model and results of the first plurality of simulations;

calculating a second plurality of simulations for the each of the plurality of portfolios;

performing an overlaid calculation by:

determining, by the processor, an alphanumeric code for each of the plurality of asset classes (A_k);

compiling, by the processor, a request message for an external database storing one or more joint models for the plurality of asset classes (A_k), where the request message includes the alphanumeric code for each of the plurality of asset classes (A_k), the external database facilitating non-local storage of the one or more joint models for the plurality of asset classes (A_k);

receiving, by the processor and responsive to the request message, the one or more joint models for the plurality of asset classes (A_k);

determining, by the processor and based on the one or more joint models, a joint portfolio performance vector for the plurality of portfolios based on portfolio specific models and results of the second plurality of simulations; and

determining, by the processor, for each portfolio, a portfolio specific scalar based on the single portfolio performance vector and the joint portfolio performance vector; and

generating a plurality of messages that separate the overlaid calculation from the non-overlaid calculation for an initial margin for the plurality of portfolios based on the portfolio specific model and the portfolio specific scalar, wherein the number of the plurality of messages for the trading system is reduced for margin requirements due to the portfolio specific model and the portfolio specific scalar,

wherein as asset class margin (Λ_k) for each of the plurality of asset classes (A_k) is calculated according to:

Λ_k=Percentile_Θ (P&L),

wherein Percentile is a function that returns a margin value for the asset class margin (Λ_k) according to a probability distribution for a profit and loss derived from the first plurality of simulations,

wherein the joint margin (Λ_joint) is calculated according to:

Λ_joint=Percentile_Θ (P&L),

wherein Percentile is a function that returns a margin value for the joint asset margin (Λ_joint) according to a probability distribution for the profit and loss (P&L) derived from the second plurality of simulations,

wherein a reduced initial margin (IM) is calculated according to:

IM=Λ_k−w_k·Λ_joint

wherein the asset class margin (Λ_k) is reduced by a contribution w_kportion of the joint asset margin (Λ_joint),

wherein the reduced IM decreases a computation time for the trading system to process the plurality of portfolios.