Designing A Legitimate Automated Scalping Program .. Pipedream?

Discussion in 'Index Futures' started by Algo_Design_Kid, Dec 29, 2010.

  1. I am having a really hard time trying to not only conceptualize but implement a scalping program. The inherent limitations in the backtesting engine are one thing. But I really do not believe that is the biggest hurdle.

    Without being able to recreate the entire order book I am finding it almost impossible to believe it can be done correctly. Not only that but just for good measure I usually add 1 tick slippage / side to all orders. So even if I cut this in half (half tick / side) in reality I am most likely only fooling myself.

    If anyone has some insight into these subjects or personally has invested the time and money into being able to acquire the entire book (with cancellations) I would be more than interested to hear about it.

    Thanks for your time
  2. Can you scalp manually? Have you ever had success trading intraday? Are you familiar with any scalping strategies used at prop firms?

    If you don't know how to manually scalp and/or never did, I'd say designing an automated scalping indeed a pipedream.
  3. Never have. I know myself well enough to know I don't have the mental game to manually trade something like this.

    At the same time this will not be my own personal account, but of course directly effects my livelihood. It is just an additional project I have always been interested in but only recently have started to put work toward. Regardless, it needs to have a historical demonstratable edge and needs to be automated. I have been collecting bid/ask data, but I am still not 100% sold on the idea that these orders are going to get filled.

    This also goes off on the tangent regarding being able to recreate the entire book with cancellations and if anyone / group of people have been satisfied with their investment.

    Without outright detailing my idea I am basing it off the DOM, creating "quadrants" and running this through a NN to look for opportunities. Of course, I am probably not the first person to think about something like this so it isn't like it really matters.

    More or less I was interested in how people have overcame historical backtesting inefficiencies (if at all), what kind of slippage was accounted for, data used etc.

  4. +1
  5. joe4422


    I think you would want to automate something you already do that is successful. Finding out exactly what happened in the past is not going to tell you what will happen in the future.
  6. It will be a complete waste of time backtesting a scalping strategy, because slippage is not necessarily one tick, especially in 6E.

    Attempting to do it manually highlights the immediate deficiencies that would then warrant it being automated,provided the strategy is sound, and that is speed of price change .

    I am in the process of having a guy write code for a scalping system (scalping =profit 20 ticks average loss 3 ticks...flipping both ways with a size allocation change at given points)...current average session 5...

    Will let you know how it turns out.

  7. The "quadrant" idea won't work. You are using data that is 99% noise. If you don't have familiarity with real world scalping strategies your chance of success is 0.
  8. rosy2


    you want to scalp or makemarkets? it might be easier to just place buys and sells around the BBO. I am curious why you cannot recreate the orderbook if you have the data
  9. So you are looking at a longer holding period than I am I believe. Mainly I want to trade very large size, bounded at last resort with time restrictions.

    Are you just going to run it on live simulation then go for execution?
  10. I don't exactly have the data, I have been recording some. Also, the time I want to trade is very heavy in volume and I am sure I am missing pieces of the data. That is why I was wondering if anyone has invested money into these areas. Ideally it would also include cancellations.
    #10     Dec 30, 2010