Testing Algo Engines

, , ,
1

Imported from previous forum

2

Hello,

  1. Does anyone offer algo-specific testing tools?

    I checked the vendor/product page and didn’t see
    any algo-specific test tools.

  2. Is there a need for algo-specific testing tools?

    For the past few years I’ve designed and implemented
    repeatable, deterministic, overnight testing of algo engines.
    It’s made testing much faster and more reliable.

    I’m wondering if other people see the speed of algo testing
    (or the lack of speed) as an important issue within their
    organization.

Thoughts?

Thanks a bunch,
Scott

3

[ original email was from Carfield Yim - carfield.w.yim@jpmorgan.com ]
I think algo issue is more on the integration side , as this is very sensitive to the input, like market data, OMS response, etc etc. Quite a lot of issue are coming from what algo expecting is difference from what market data / oms provide .

Hello,

  1. Does anyone offer algo-specific testing tools?

    I checked the vendor/product page and didn’t see any algo-specific
    test tools.

  2. Is there a need for algo-specific testing tools?

    For the past few years I’ve designed and implemented repeatable,
    deterministic, overnight testing of algo engines. It’s made testing
    much faster and more reliable.

    I’m wondering if other people see the speed of algo testing (or the
    lack of speed) as an important issue within their organization.

Thoughts?

Thanks a bunch, Scott

4

I think algo issue is more on the integration side , as this is very
sensitive to the input, like market data, OMS response, etc etc. Quite a
lot of issue are coming from what algo expecting is difference from what
market data / oms provide .

Hello,

  1. Does anyone offer algo-specific testing tools?

    I checked the vendor/product page and didn’t see any algo-specific
    test tools.

  2. Is there a need for algo-specific testing tools?

    For the past few years I’ve designed and implemented repeatable,
    deterministic, overnight testing of algo engines. It’s made
    testing much faster and more reliable.

    I’m wondering if other people see the speed of algo testing (or the
    lack of speed) as an important issue within their organization.

Thoughts?

Thanks a bunch, Scott

Scott,

Don’t know if this is directly on your point but wanted to comment that even with FIXatdl there is still a need for initial testing. Sure, a BD can whip up a new algo strategy, create an updated FIXatdl xml file that describes it, send it out, and many OMS can drop that in and automatically render the screens for traders to start trading it. HOWEVER, there is still a need to test. Hopefully that will be greatly reduced over time especially for minor changes.

At FIXatdl the WG is in the final stages of producing an updated version - will be v1.1 when finally released. We would VERY MUCH LOVE to have a testing expert help insure there are no surprises and all the edges are tested.

Where FIXatdl is very likely to break currently standing OMS systems is in complex trades. For instance a BD may create a multi product strategy (e.g. several insterest rate futures against an underlying treasury) and that all goes off as a single new order message. However the OMS isn’t aware of that and when it gets back multiple (single product) executions its most likely to be in a DK (don’t know) situation.

Having a tester’s cautious eye would be very helpful to help define where extensive testing (and retesting) will be needed, and to test our XSD files to be sure they do exactly what we intend them to.

Rick