BioComp Dakota with Swamp Technology

Discussion in 'Trading Software' started by chifai2, May 19, 2006.

Thread Status:
Not open for further replies.
  1. James, just for the curious, is this system of yours using Flocking or PID equity control for its adaptation, and is there a reason why you chose one approach over the other?

    Cheers.
     
    #151     Aug 4, 2008
  2. jalsck

    jalsck

    The systems are using the equity controller for swarm adaptation. However, I have just finished a 4 month project that included the development of a new partical swarm adaptation library and a series of new swarms that are based on the k-Nearest Neighbor algorithm. Signals that I post ET are now based on Sp_z0021 that incorporates the latest research.

    Kind Regards,

    James
     
    #152     Aug 4, 2008
  3. James, can you post an equity curve for the Sp_z0021 model?

    Thanks.

    Your track record here would look a lot better if you took the trades on the open and put a reasonable (2.5% or 3%) stop on. Quite good actually.
     
    #153     Aug 4, 2008
  4. Thanks James.
     
    #154     Aug 4, 2008
  5. jalsck

    jalsck

    The equity curve for Sp_z0021 is attached. The parameter ranges for the 39 Dakota systems that contributed to Sp_z0020 have not been changed.

    Kind Regards,

    James
     
    #155     Aug 4, 2008
  6. jalsck

    jalsck

    An graph of the swarm for Sp_z0021 is attached. The swarm adaption routine used by Sp_z0021 is a hybrid that switches to individual hill climbers when a given percentage of the trade bots are losing and then changes to a flocking type swarm when a profitable zone is found again.

    Kind Regards,

    James
     
    #156     Aug 4, 2008
  7. Hi James,

    Can you post a shot of the trade stats screen for system 21, so we can compare them with the similar stats for sytems 19 and 20, which you previously posted?

    Thanks.
     
    #157     Aug 4, 2008
  8. jalsck

    jalsck

    The statistics for Sp_z0021 are attached.

    Kind Regards,

    James
     
    #158     Aug 4, 2008
  9. Thank you for posting those.

    Comparing the new system stats to those of the two original systems, system 21 seems to have a worse % perfect, a worse W/L point ratio, and substantially less total equity. So I'm curious as to why you made the switch. Can you shed some light on this?

    Thanks.
     
    #159     Aug 4, 2008
  10. jalsck

    jalsck

    The individual Dakota systems are adapting to different market conditions better than before. I think we are seeing the effect of adding more systems. I plan on selecting a subset of systems to use. To do this properly I will need to keep a couple of years for out of sample validation.

    Another approach is to use one of the scripts that selects the Top N systems on a walk-forward basis. I will keep you posted.

    Kind Regards,

    James
     
    #160     Aug 4, 2008
Thread Status:
Not open for further replies.