Author Topic: Questions re Execution Platform  (Read 3765 times)

Alexander Horn

  • Newbie
  • *
  • Posts: 2
  • Karma: +0/-0
Questions re Execution Platform
« on: September 26, 2014, 08:06:28 pm »
Just questions out of curiosity from own coding, not suggesting these for implementation.

 - Have you tried running advanced / combined orders in IB for the pairs? E.g. bracket orders / pair hedge orders to trigger intraday entries/exits in IB, or do you rely on MKT orders to be triggered by PLT?

 - Have you tried setting up stop loss orders in IB? Noted one post on the forum that you rather sit out losses and rely on stop bars, and smoothen hits running many pairs… but still think this hurts if pairs drift?

 - Very much like your “slot-based” allocation as it allows optimum exposure / weighting between pairs. But still, how to you determine priority for pairs to be entered if several signals are triggered or over time? As they cross thresholds, by profit potential?
« Last Edit: September 26, 2014, 08:09:15 pm by Alexander Horn »

admin

  • Administrator
  • Sr. Member
  • *****
  • Posts: 315
  • Karma: +12/-0
    • Pair Trading Lab
Re: Questions re Execution Platform
« Reply #1 on: September 27, 2014, 10:26:16 am »
Just questions out of curiosity from own coding, not suggesting these for implementation.

 - Have you tried running advanced / combined orders in IB for the pairs? E.g. bracket orders / pair hedge orders to trigger intraday entries/exits in IB, or do you rely on MKT orders to be triggered by PLT?

I did not, some of my friends tried to use IB combo orders for this but they were not too happy about the reliability of these IB execution algos. They reported some problems also to IB, but I don't know details. However this was few years ago, maybe the situation is different now.

Market orders work fine, there are no problems with them whatsoever, even the slippage is usually negligible. So we did not have to seek more sophisticated options for executions yet.

Of course MKT orders are for retail traders and liquid markets only, if you are a big customer/fund, you have to execute orders in much smarter way to avoid moving the market. We don't provide these solutions yet (but we have a partner providing execution algos for bigger clients working closely with us).

But even for big clients, I still think MKT orders will be fine, as long you divide your orders to multiple smaller orders and execute them in right time looking at level 2 market data.

- Have you tried setting up stop loss orders in IB? Noted one post on the forum that you rather sit out losses and rely on stop bars, and smoothen hits running many pairs… but still think this hurts if pairs drift?

The problem is that you can't use stop loss orders effectively - because you are interested in the cumulative profit loss of the whole pair position, not only of its legs. So adding stop orders to particular stocks don't help at all.

You could set a stop loss order for the whole pair position if you used a combo order, or you can execute stop loss for the whole position in algorithmic way. But even if you do that, it won't help you much. And it will never help you to predict the worst risk anyway. Why? Because you trade stock market, which is closed during nights, weekends and holidays. You can react when the market opens again, but usually this is too late. Big moves in stock market usually happens during weekends, and a large gap occurs. Then when you react Monday morning, it is too late.

Once I have faced a situation where I was in a pair position and then during the weekend, the long stock dropped 40% and the short stock increased 15%. Both stocks gapped against my position. I can admit that I have no idea how to control risk in scenarios like this one apart from diversification over different markets and strategies.

Yes, stop loss may help you if pair position gradually drifts away to red numbers without significant gaps. But usually time-based stop loss has almost the same effect.

Maybe I will add the stop loss option again for the backtester so users can see for themselves, but is is not a high prio atm. Pair traders I know don't use loss-triggered stops.


- Very much like your “slot-based” allocation as it allows optimum exposure / weighting between pairs. But still, how to you determine priority for pairs to be entered if several signals are triggered or over time? As they cross thresholds, by profit potential?

In PTL Trader this actually does not happen, because positions are triggered by intraday (real-time) price events. So scenarios where multiple signals come at the same millisecond are quite rare. The fastest signal wins.

In backtester, the order is determined by the strategy position in the portfolio, I think. There is no possibility to set priorities at the moment (because it will be useless for trading based on real-time prices anyway).

About the slot occupation settings - at this moment, users just set the slot occupation per strategy manually if they want. In future, I will implement some basic meta-strategies (successful strategies can automatically get some boost in slot allocation, while slot occupation for less-performing strategies can be tuned down until the strategy recovers).

admin

  • Administrator
  • Sr. Member
  • *****
  • Posts: 315
  • Karma: +12/-0
    • Pair Trading Lab