Robot Trading on futures - ES et al

Discussion in 'Journals' started by iuykcif, Sep 3, 2009.

  1. Curious as to how a neutral ETF Pair would perform without constraints.

    For futures without constraints what % retracement is required to closeout?

    What size positions would the accounts be carrying during a 20 point rally?

    Can you set the constraint dynamically based on a dollar value as the account grows?

    ie. Start with 2 contract constraint approx $10K and increase dynamically using 50% of the systems profits.

    IB has the most conservative margins for trading futures. May want to take a look at the plethora of future brokers offering $500 intraday margins and either go flat or constrain the system to scale down to the accounts max supported position size for overnight.

    As far as I know IB does not support Net Margining of sub accounts. With Net Margining the cash requirements seem reasonable.


     
    #31     Sep 4, 2009
  2. iuykcif

    iuykcif

    > Curious as to how a neutral ETF Pair would perform without constraints.

    Well I think that while futures are entertaining, the best way to use this approach is on ETF pairs.

    I would probably pick SPY/SH

    I believe that the game can be relatively much more efficient because of the smaller "granularity".

    I mean while 1ES = 500 SPY, with etf i can make smaller packets. For instance 200 shares at a time. And this should increase efficiency.

    Maybe we sometime could start a journal with ETF pairs.


    >For futures without constraints what % retracement is required to closeout?

    Do not have enough data to tell precisely for futures. But I guess experienced future traders can tell.
    Options may also be of aid to keep the problem under control


    >What size positions would the accounts be carrying during a 20 point rally?


    This is an important point.

    It depends on the scalp range.
    Assume the smallest one: 1 point.

    Theoretical max is as follows:


    UnConstrained:

    Rally UP

    Acct Long N+1
    Acct Short -2N


    Rally DOWN

    Acct Long 2N
    Acct Short - (N+1)

    with N<=20

    But this max requires a rally with no small oscillations: say like a 45 degree line :))

    Constrained:

    It's much lower, but for a formula have to think about it. (The constraint causes the scalpers to "slide" on rallies.)


    >Can you set the constraint dynamically based on a dollar value as the account grows?

    Actually that's a very good idea. I think I am going to add this possibility.


     
    #32     Sep 4, 2009
  3. iuykcif

    iuykcif

    _

    Actually gamma scalping could be integrated well in the strategy, giving a "double" action strategy for which thinking of a losing scenario requires fervid imagination.


    <b>STRATEGY VARIANT (+option scalping)</b>

    Assume, for instance, ES is 1000. We start this:

    10 ES CALL (S account) ---------------------------------------------------------------------------------- 1010

    ES ROBOT SCALPING (max 5 contracts difference) --------------------------------------------------- 1000

    10 ES PUT (L account) ---------------------------------------------------------------------------------- 900


    Denote P_ES the P&L due to ES scalping only
    Denote P_tot the global, P&L, including options

    Assume the following algorithm:

    - If the ROBOT is not "frozen" continue scalping ES only (closing normally every time you see a scalp profit, P_ES > 190 $ for instance)

    - if the ROBOT is "frozen" (rally, 5 contract difference reached) close when P_tot > 190 $ (or whatever)
    (this might happen anytime either because one option is ITM or because price chances on options)

    - If all is closed. Begin again from the current price with a new straddle and new automated inner scalping


    In other words, while the Robot makes a <b>massacre within </b> the straddle (with a little constraint of 5 contracts difference between accounts), in case of big rallies it is protected by options, which are, in turn, scalped if profitable.

    If not, it means the price is returning within the straddle, and a boatload of money is being made, due to the averaging Up/Down of the robot.
    (Remember that when rallying Up, the robot loads short, and vice versa).

    This is anyway an additional layer of security, since the robot is already doing "hedging" in a neutral strategy.

    I am quite positive that this scalping action cannot be improved, due to the peculiar nature of the scalping algorithm, however further improvements can be achieve with appropriate use of options.

    Do you have suggestions to improve this approach ?


    Tom


     
    #33     Sep 6, 2009
  4. mockney

    mockney

    I'm not clear why you need to maintain an open long and short position concurrently when this is the same delta as being flat.

    when you take one side off or add one, its just the same as opening a position from flat, is it not? so why not do that and reduce transaction costs for un-necessary trades?
     
    #34     Sep 6, 2009
  5. It is not the same regarding control of contracts and it would require extremely precise execution to achieve a specific position using a single account.

    No such thing as a fractional ES contract and fractional ticks. If the Algo requires the net account position to be net Long 2.33 contracts at 1007.125 when the market touches 1008 how could you achieve such position in sequence using a single account?

    There are also delays between placing orders and getting confirmed fills. When a Fill comes back outside of what was expected the position is easier to adjust when the long component and short components are separated.

    There is also a benefit of net margining and control of the two sets of contracts that remain open. If a typical trading session is 400 - 500 round turns you will have built up positional control of 800 - 1000 contracts with a net margin for only 5. With control of these contracts you have the ability to write options that are covered. At any time you can instruct your broker to offset and close the contracts out against each other. No trip back through the market, no spread, no slip, no commissions.
     
    #35     Sep 6, 2009
  6. Have you ever seen the 180 degree line last year when FED cut rates 75 basis points ? You should do your backtesting VERY carefully. ;=)
     
    #36     Sep 6, 2009

  7. Hey ASusilovic,

    just wondering if you know what date that was so I can look at it on historical charts ? :)
     
    #37     Sep 6, 2009
  8. 22nd January 2008. One day after SocGen had to unload Jerome Kerviels famous 5 billion EUR trades-at least we have all been informed about the " computer genius" status of "Jerome"...
     
    #38     Sep 6, 2009
  9. Marked area...
     
    #39     Sep 6, 2009
  10. iuykcif

    iuykcif

    Just one concrete example why look seriously into options when doing this game (and, probably, any similar game).

    In fact, I imagine your word of caution holds for many trading strategies, and not specifically for what we are discussing here.

    [Clearly not even options would save us from falling meteors, though :)) ]


    Tom

     
    #40     Sep 6, 2009