Treasury futures algo plainly visible-dumb or smart?

Discussion in 'Automated Trading' started by jelite, Apr 12, 2012.

  1. jelite

    jelite

    I see almost on a daily basis an algorithm that is usually active around 2pm CST that keeps buying/selling a fixed number of lots in treasury futures at what seems like fixed time intervals. It can be seen on cumulative delta chart (you have to go down to one second resolution). Attached pictures shows cumulative delta for UBM2 (ultrabond June 12 futures contract) on 4/4/2012. It's so obvious that it almost seems the algo 'wants to' be seen. Sometimes (not always) there is simultaneous selling of another treasury futures contract (for example buying ZN and selling ZB, so probably playing the spread). Was wondering if someone else (most likely technically savvy...) noticed this too? Maybe there is something to exploit here (by following the algo). Feel free to pm me to discuss my/your observations.
     
  2. probably a scout algo - sent out obviously losing.
    as soon as the algo making money out of it is seen the bigger algo steps in and wipes it out.

    pretty obvious bullshit algo trading in all honesty.

    destroy any natural business and liquidity and provide no readl genuine business to trade off.

    hence exchange volume falling off a cliff.
     
  3. jelite

    jelite

    Maybe you are right (although the post was not quite clear to me), the 'true' algo may be hidden. But the 'visible' algo is loading up on thousands of contracts-not peanuts by any means-doesn't seem like a scout to me (would be a huge scout).
     
  4. jelite,

    May I know where you got that chart?

    Just want to confirm whether you may be reading too much into a simple buy order spread over 30 minutes.
     
  5. Bob111

    Bob111

    i've seen a lot of dumb algos, but not on those securities.
     
  6. jelite

    jelite

    A 'simple' spread order of ten thousand contracts (that's 1 billion notional of ZN I am talking about here...)? Hedge funds usually work the spread at a predetermined level (of spread) and may use icebergs on the legs of the spread but don't just simply spray the order flow with buys/sells for a number of minutes with fixed size (aggressively). My chart is my own production but that's not important where the chart comes from, this information is visible to anyone with sufficiently fine resolution of order flow (down to one second).
     
  7. +1

    He didn't mean "spread" in terms of difference between two assets.

    He meant a buy order spread across 30 minutes/seconds, as in I need to buy 30 lots and I automate a program to buy 1 lot every minutes/seconds - spreading the buy order across time. It's a basic buy-side algorithm designed to minimize market impact.
     
  8. jelite

    jelite

    I didn't read that one correctly... but still seems a little too simplistic to me to buy that many contracts so visibly, why wouldn't you at least randomize the lot size/ time interval?
     
  9. It's hard to tell their total volume or motivation. It could be closing a position, opening a new one, hedging a huge book of some sort, anything.

    It's far from nefarious however.
     
  10. At "1 second" resolution, it matters very much where the chart comes from.
     
    #10     Apr 13, 2012