The Gap Fading system

Discussion in 'Journals' started by thetrendfollowe, Sep 15, 2009.

  1. Hi everyone.

    Some background:

    I've been trading the ASX (local bourse) for the last 4 years, initially pure punting (FA lol) then discretionary TA and the last two years through mechanical systems, where I can say I've found my niche.

    My usual method is longterm trend following (LTTF), average hold 8 months (yes I got slaughtered last year! lol).

    Recently I've been looking at short-term strategies for two reasons.
    1/ Some income.
    2/ To compliment my LTTF system in producing more consistent, non-correlated performance.

    So far I have found that mean reversion and gap fading works well.


    The purpose of this journal is to track the realtime progress of a new system I have designed that will fade opening gaps; more specifically, it will go long gap downs.

    The market traded is US equities and there are 50 tickers that will be traded. These are all either part of the Nasdaq-100 or S&P100, so I am looking for liquidity here to minimise slippage.

    **Note, I have not included slippage in the testing at all, so still unsure at all how this will affect the system. Shouldn't be too bad since most of my tickers maintain 1-tick spreads and I enter with limit orders. We'll see. In testing, I have doubled IBs usual commission to at least partly account for this**

    In my testing, I have found fading opening gap downs is a robust strategy that works well with ANY sort of magnitide of gap.

    The entry will be at the open, using limit-on-open orders. The exit is just at the end of day (market-on-close order) OR there is a maximum stop loss that rarely gets hit. No overnight holds.

    Backtesting has shown about 50-100 trades a month with average nett profit of about $20k. Commissions in Amibroker were set to 1c/share.

    This will be traded via IBs PaperTrader with initial capital of US$1million (which is what it will be traded with in real-life in terms of BP, if the trial goes well).

    Position sizing is approx 100k per trade. Parcels will be rounded to 100 share lots. Profits are not re-invested.

    In real-life, profits would be withdrawn monthly. In order to simulate this as much as possible, no more than $1million worth of trades would be placed on market on any given day.

    After 6-12 months, and several hundred trades, i will make a decision whether or not to trade this system in with real money.

    More info and results from backtesting to follow soon.

  2. The system has been tested on in-sample (2004-2007) and out-of-sample data (2008-present), the results of which are in the file attached.

    There is a "Black swan" that occurred in one of the months last year. But I'm accepting that as a rare event. 2008 broke many records for volatility. While I had a 15% drawdown (and 20% for the year), many others traders and funds blew up. I still came out ahead for 2008 (nothing to brag about), which isn't bad for a long-only equities system considering the conditions.

    Though I should say that before this the max.DD of the system was closer to 2%!

    I may not start trading until next week as I'm still waiting for NYSE to approve my data subscription.

    Any comments are most welcome.
  3. I wish you well. Where did you get the historical data you used for backtesting with Amibroker?
  4. Hi giggollo;

    I get all my data from these guys:

    Cant recommend them enough.
  5. Thats an interesting data source.. and what did you do to verify that the opening and closing prices from this data provider are accurate and are actually what you would get in real life?
  6. And thats an interesting question.

    Iv been using these guys for my ASX data for years.
    Never a discrepancy between them and the exchange.

    I have checked some of the recent trades with IB's prices and it matches up okay.

    Have you had a bad experience with data providers in the past?
  7. Giggollo.

    Just to add to my response.

    I find it interesting because nobodys asked me that. Maybe its because premium data don't have a global reputation. Here in Australia, they are pretty much as good as it gets. Alot of big players that I know use them.

    But even if you do have a point, thats the whole purpose of this journal. To see if the results in backtesting can be replicated in real life.
  8. Beware that a number of popular data providers dont have the correct daily opening and closing prices for US stocks. This means your simulation may not match your reality if you are assuming the data is correct. If that is the case, you would eventually realize this once you start trading and find your actual fills are not the same as the data source prices, but im just giving you a heads up
  9. yobo


    Sounds interesting. Good luck. You may want to take a look at the Opening Orders Journal thread. Same type of strategy but a little different. Basically with the Openings, calculate fair value based on the futures, give your self a a little buffer, place a limit buy or sell opg order an scalp 5-10 cents. The folks at Bright Trading are famous for this. Works about 65% of the time.

    Another way to play the gap is to look at a 3 day chart and the 20 minute 20 day moving average line. Buy and fade based on where it opens in relation to the line. Works great for the mean reversion and its simple. I use the 20 minute time frame becuase I want to be out in the first 20 minutes.

    Placing open orders and then closing it out on the close sounds a bit crazy. I just don't see the logic. Also only buy gap downs works great when the market is moving up, but lousy when the market is moving down.

    Just my 2 cents.

  10. Hi yobo,

    Thanks for your comments.

    Yeah I only look at strategies that can be backtested and eventually automated. I was going to look at scalping, etc, but these are usually techniques where discretion is required. Not my cup of tea. Also execution must be real quick.

    The logic in the system is simply the upwards bullish bias in equities markets. I'm not going to argue with the results. You must understand what I came up with here is not opinion based, but rather evidence based.

    In down years, bearmarkets, etc, my maximum stop loss keeps me out of too much trouble. In fact 2008 was a profitable year if you can see on the attached file.

    Hit rate on this system is around 60%.
    #10     Sep 15, 2009