Methods and systems for generating derived summaries of trading signals
Systems and methods are disclosed herein that compute trading signals with low latency and high throughput using highly parallelized compute resources such as integrated circuits, reconfigurable logic devices, graphics processor units (GPUs), multi-core general purpose processors, and/or chip multi-processors (CMPs). The trading signals can be summarized over time durations to create derived summaries of the trading signals. These derived summaries can be communicated with one or more data consumers.
1. A multi-processor system for derivations of trading signals with respect to financial instruments, the system comprising:
one or more first processors; and
one or more second processors, wherein the one or more second processors comprises at least one field programmable gate array (FPGA);
the at least one FPGA configured to receive data representing a plurality of financial market data messages, the messages pertaining to a plurality of financial instruments and comprising information about orders to buy and/or sell the financial instruments;
wherein the at least one FPGA comprises (1) feature compute stage circuitry arranged in parallel and configured to compute, in parallel via parallelized hardware logic on the at least one FPGA, a plurality of features of the financial market data messages in parallel based on the received data and (2) combine stage circuitry connected to the parallel feature compute stage circuitry and configured to generate real-time estimator trading signals for a plurality of the financial market data messages based on the computed features; and
wherein the one or more first processors is configured to (1) receive data comprising (i) the information about a plurality of orders to buy and/or sell a plurality of the financial instruments and (ii) the real-time estimator trading signals, (2) summarize the real-time estimator trading signals over a time duration to create a derived summary of the real-time estimator trading signals that serves as a non-real time summary of the real-time estimator trading signals for the time duration, and (3) communicate the derived summary of the real-time estimator trading signals to one or more data consumers.
2. The system of claim 1 wherein the combine stage circuitry is configured to compute a hidden liquidity size estimation as a weighted combination of the features, wherein the hidden liquidity size estimation includes an estimated size of a hidden order among the orders, wherein the real-time estimator trading signals include the computed hidden liquidity size estimation, wherein each feature has a corresponding weight for the weighted combination, and wherein the features comprise features that are correlated to and predictive of the estimated size for the hidden order.
3. The system of claim 2 wherein the one or more first processors comprises a plurality of first processors.
4. The system of claim 2 wherein the data received by the one or more first processors comprises a plurality of financial market data messages, the messages comprising the information about orders to buy and/or sell financial instruments.
5. The system of claim 1 wherein the combine stage circuitry is configured to compute a quote price duration estimation with respect to a price for a quote on a financial instrument as a weighted combination of the features, wherein the real-time estimator trading signals include the computed quote price duration estimation, wherein each feature has a corresponding weight for the weighted combination, and wherein the features comprise features that are correlated to and predictive of the quote price duration estimation.
6. The system of claim 5 wherein the one or more first processors comprises a plurality of first processors.
7. The system of claim 5 wherein the data received by the one or more first processors comprises a plurality of financial market data messages, the messages comprising the information about orders to buy and/or sell financial instruments.
8. The system of claim 1 wherein the combine stage circuitry is configured to compute a quote price direction estimation with respect to a price for a quote on a financial instrument as a weighted combination of the features, wherein the real-time estimator trading signals include the computed quote price direction estimation, wherein each feature has a corresponding weight for the weighted combination, and wherein the features comprise features that are correlated to and predictive of the quote price direction estimation.
9. The system of claim 8 wherein the one or more first processors comprises a plurality of first processors.
10. The system of claim 8 wherein the data received by the one or more first processors comprises a plurality of financial market data messages, the messages comprising the information about orders to buy and/or sell financial instruments.
11. The system of claim 1 further comprising a Tier 2 or Tier 3 trading application that serves as a member of the one or more data consumers.
12. The system of claim 11 wherein the Tier 2 or Tier 3 trading application is configured to apply the derived summary to a predictive model in combination with current financial market data to generate a plurality of financial instrument orders for communication to an exchange.
13. The system of claim 12 wherein the Tier 2 or Tier 3 trading application is configured to apply the derived summary to the predictive model in combination with current financial market data and historical financial market data to generate the financial instrument orders for communication to the exchange.
14. The system of claim 1 wherein the time duration comprises a duration in a range between 1 minute and 30 minutes.
15. The system of claim 1 wherein the time duration comprises a day.
16. The system of claim 1 wherein the one or more first processors is further configured to perform the summarize and communicate operations on a periodic basis so that the derived summaries cover different time durations.
17. The system of claim 1 wherein the data received by the one or more first processors comprises a plurality of financial market data messages, the messages comprising the information about orders to buy and/or sell financial instruments.
18. The system of claim 17 wherein the real-time estimator trading signals are synchronized with the messages to which they pertain.
19. The system of claim 17 wherein a plurality of the messages comprise a plurality of the real-time estimator trading signals.
20. The system of claim 1 wherein the one or more first processors comprises a plurality of first processors.
21. A multi-processor method for computing derivations of trading signals with respect to financial instruments, the method comprising:
performing a first plurality of operations by one or more first processors, wherein the one or more first processors comprises at least one field programmable gate array (FPGA), wherein the at least one FPGA comprises (1) feature compute stage circuitry arranged in parallel and (2) combine stage circuitry connected to the parallel feature compute stage circuitry; and
performing a second plurality of operations by one or more second processors;
wherein the first plurality of operations include (1) receiving data that represents a plurality of financial market data messages, the messages pertaining to a plurality of financial instruments and comprising information about orders to buy and/or sell the financial instruments, (2) the parallel feature compute stage circuitry computing, in parallel via parallelized hardware logic on the at least one FPGA, a plurality of features of the financial market data messages based on the received data, and (3) the combine stage circuitry generating real-time estimator trading signals for a plurality of the financial market data messages based on the computed features; and
wherein the second plurality of operations include (1) receiving data comprising (i) the information about a plurality of orders to buy and/or sell a plurality of the financial instruments and (ii) the real-time estimator trading signals, (2) summarizing the real-time estimator trading signals over a time duration to create a derived summary of the real-time estimator trading signals that serves as a non-real time summary of the real-time estimator trading signals for the time duration, and (3) communicating the derived summary of the real-time estimator trading signals to one or more data consumers.
22. A multi-processor system for derivations of trading signals with respect to financial instruments, the system comprising:
one or more first processors; and
one or more second processors, wherein the one or more second processors comprises at least one graphics processor unit (GPU); and
the at least one GPU configured to receive data representing a plurality of financial market data messages, the messages pertaining to a plurality of financial instruments and comprising information about orders to buy and/or sell the financial instruments;
wherein the at least one GPU comprises (1) feature compute stage circuitry arranged in parallel and configured to compute, in parallel via a plurality of compute cores on the GPU, a plurality of features of the financial market data messages based on the received data and (2) combine stage circuitry connected to the parallel feature compute stage circuitry and configured to generate real-time estimator trading signals for a plurality of the financial market data messages based on the computed features; and
wherein the one or more first processors is configured to (1) receive data comprising (i) the information about a plurality of orders to buy and/or sell a plurality of the financial instruments and (ii) the real-time estimator trading signals, (2) summarize the real-time estimator trading signals over a time duration to create a derived summary of the real-time estimator trading signals that serves as a non-real time summary of the real-time estimator trading signals for the time duration, and (3) communicate the derived summary of the real-time estimator trading signals to one or more data consumers.
23. The system of claim 22 wherein the real-time estimator trading signals include quote price duration estimations with respect to prices for quotes on financial instruments.
24. The system of claim 22 wherein the real-time estimator trading signals include quote price direction estimations with respect to prices for quotes on financial instruments.
25. A multi-processor system for derivations of trading signals with respect to financial instruments, the system comprising:
one or more first processors; and
one or more second processors, wherein the one or more second processors comprises at least one field programmable gate array (FPGA);
the at least one FPGA configured to (1) receive data representing a plurality of financial market data messages, the messages pertaining to a plurality of financial instruments and comprising information about orders to buy and/or sell the financial instruments and (2) process the financial market data messages through parallelized hardware logic on the at least one FPGA to compute a plurality of features of the financial market data messages in parallel;
wherein the features comprise a plurality of conditions for a logical trading signal, and wherein the at least one FPGA comprises (1) condition testing circuitry arranged in parallel and configured to test for the conditions, in parallel via parallelized hardware logic on the at least one FPGA, based on the financial market data messages and (2) combine logic circuitry connected to the parallel condition testing circuitry and configured to generate logical trading signals for a plurality of the financial market data messages in real-time based on outputs from the condition testing circuitry; and
wherein the one or more first processors is configured to (1) receive data comprising (i) the information about a plurality of orders to buy and/or sell a plurality of the financial instruments and (ii) the real-time logical trading signals, (2) summarize the real-time logical trading signals over a time duration to create a derived summary of the real-time logical trading signals that serves as a non-real time summary of the real-time logical trading signals for the time duration, and (3) communicate the derived summary of the real-time logical trading signals to one or more data consumers.
26. The system of claim 25 wherein the real-time logical trading signals include a hidden liquidity indicator that indicates if a subject order is flagged as part of a hidden order for a financial instrument that corresponds to the subject order.
27. The system of claim 26 wherein the one or more first processors is further configured to summarize the real-time logical trading signals by aggregating a plurality of hidden liquidity indicators over the time duration.
28. The system of claim 27 wherein the one or more first processors is further configured to perform the aggregation on a per financial instrument basis.
29. The system of claim 28 wherein the derived summary comprises, per applicable financial instrument, (1) a total count of hidden buy orders detected over the time duration and (2) a total count of hidden sell orders detected over the time duration.
30. The system of claim 28 wherein the derived summary comprises, per applicable financial instrument, (1) a volume weighted average price (VWAP) of hidden buy orders detected over the time duration and (2) a VWAP of hidden sell orders detected over the time duration.
31. The system of claim 28 wherein the derived summary comprises, per applicable financial instrument, (1) a notional value of hidden buy orders detected over the time duration and (2) a notional value of hidden sell orders detected over the time duration.
32. The system of claim 26 wherein the trading signals further include at least one FPGA is further configured to compute a hidden liquidity size estimation that includes an estimated size of the hidden order for the subject order's corresponding financial instrument.
33. The system of claim 32 wherein the one or more first processors is further configured to summarize the real-time logical trading signals by aggregating a plurality of hidden liquidity indicators and their corresponding hidden liquidity size estimations over the time duration.
34. The system of claim 33 wherein the one or more first processors is further configured to perform the aggregation on a per financial instrument basis.
35. The system of claim 34 wherein the derived summary comprises, per applicable financial instrument, (1) a total count of hidden buy orders detected over the time duration, (2) a total count of hidden sell orders detected over the time duration, (3) data that represents an aggregation of hidden liquidity size estimations for the total count of hidden buy orders, and (4) data that represents an aggregation of hidden liquidity size estimations for the total count of hidden sell orders.
36. The system of claim 35 wherein the data that represents an aggregation of hidden liquidity size estimations for the total count of hidden buy orders comprises a Volume Weighted Average Price (VWAP) for the hidden buy orders in the total count of hidden buy orders, and wherein the data that represents an aggregation of hidden liquidity size estimations for the total count of hidden sell orders comprises a VWAP for the hidden sell orders in the total count of hidden sell orders.
37. The system of claim 35 wherein the data that represents an aggregation of hidden liquidity size estimations for the total count of hidden buy orders comprises a notional value for all of the hidden buy orders in the total count of hidden buy orders, and wherein the data that represents an aggregation of hidden liquidity size estimations for the total count of hidden sell orders comprises a notional value for all of the hidden sell orders in the total count of hidden sell orders.
38. The system of claim 25 wherein the at least one FPGA is configured to receive the financial market data messages in a plurality of network packets; and
wherein the parallel condition testing circuitry and the combine logic circuitry cooperate to detect that a hidden order for a subject financial instrument is present if (1) a first of the messages is (i) from a first network packet and (ii) represents an order execution at a price for the subject financial instrument and (2) a second of the messages is (i) subsequent to the first of the messages, (ii) exists in the first network packet or another network packet that is near the first network packet according to defined criteria, and (iii) represents an order addition at the same price for the subject financial instrument, and (3) generate a hidden liquidity indicator for the second message if the hidden order for the subject financial instrument is detected, wherein the real-time logical trading signals include the generated hidden liquidity indicator.
39. The system of claim 38 wherein the one or more first processors comprises a plurality of first processors.
40. The system of claim 38 wherein the data received by the one or more first processors comprises a plurality of financial market data messages, the messages comprising the information about orders to buy and/or sell financial instruments.
41. A multi-processor system for derivations of trading signals with respect to financial instruments, the system comprising:
one or more first processors; and
one or more second processors, wherein the one or more second processors comprises at least one graphics processor unit (GPU);
the at least one GPU configured to (1) receive data representing a plurality of financial market data messages, the messages pertaining to a plurality of financial instruments and comprising information about orders to buy and/or sell the financial instruments and (2) process the financial market data messages through a plurality of compute cores on the at least one GPU to compute a plurality of features of the financial market data messages in parallel;
wherein the features comprise a plurality of conditions for a logical trading signal, and wherein the at least one GPU comprises (1) condition testing circuitry arranged in parallel and configured to test for the conditions, in parallel via a plurality of compute cores on the GPU, based on the financial market data messages and (2) combine logic circuitry connected to the parallel condition testing circuitry and configured to generate logical trading signals for a plurality of the financial market data messages in real-time based on outputs from the condition testing circuitry; and
wherein the one or more first processors is configured to (1) receive data comprising (i) the information about a plurality of orders to buy and/or sell a plurality of the financial instruments and (ii) the real-time trading signals, (2) summarize the real-time logical trading signals over a time duration to create a derived summary of the real-time logical trading signals that serves as a non-real time summary of the real-time logical trading signals for the time duration, and (3) communicate the derived summary of the real-time logical trading signals to one or more data consumers.