Not logged in
Viewing 15 posts - 16 through 30 (of 36 total)
  • #130235|
    Customer
    477 Posts

    The feature won´t hurt as long as it can be turned on and off, just saying, I don´t think it´s good to add all kind of filters in every corner of SQ4 since this will make curve matching absolutely perfect since you can bascically filter out each bad trade from the backtest with a little filter here and there:)

    #130237
    Customer
    654 Posts

    Threshold, I could be mistaken but I think Geek’ is referring to your comment above.

     

     

    #fanning the flames of discussion.

    And my comment was in direct reference to it, hence “this is (even) less of a filter than allowing an EA to trade only @ certain time range.”

    #130238
    Customer
    654 Posts

    The feature won´t hurt as long as it can be turned on and off, just saying, I don´t think it´s good to add all kind of filters in every corner of SQ4 since this will make curve matching absolutely perfect since you can bascically filter out each bad trade from the backtest with a little filter here and there:)

    Yes, there are dangers in SQ! Optimization and using too short data and too few data sets for examples! Not everyone can be saved ;)

    If such a feature existed I might actually experiment with lower time frames. 1 of the main reasons I don’t is because they will get blow out of the water be such events as Cash Rates and NFP.

    In backtesting these will not be filtered out as “bad”. More than likely they will be big winners, or normal losers-
    Backtest doesn’t include spread widening and mass slippage.

    So the strategy will essentially be looking for “normal” volatility patterns only, no outsized winners (or losers), or minimizing it as much as possible.

    Anyway this is all theoretical and debate. I say the feature could be useful and be applied to robust strategies if done correctly. Worse filters already exist.

    #130240
    Customer
    477 Posts

    I get you, what I´ve done though, apart from the common 14 years backtest, is to simply backtests a few of these (returning) news event days and used 20 to 30 pips slippage in the data window settings. If that was overall profitable, then all is good:)

    #130446
    Customer
    143 Posts

    Good or not, its very easy to code switching EA on/off based on news releases, also for backtesting. I presume we can curvefit our strategies even more with this feature so im voting yes ;-)

    #130464
    Customer
    477 Posts

    Not so easy. How do you want to run a backtest with news avoidance? Where do you get the history news release dates from and especially reliable?

    #130466
    Customer
    654 Posts

    The federal reserve website.

    But this is some really advanced stuff. I don’t think its possible to program EAs for MT4 to link to calendar events.

    #130467
    Customer
    143 Posts
    #131785
    Customer
    745 Posts

    This is the next area I’m looking at.

     

    It’s pretty obvious I guess that opening a position right before a big news event (like US to announce start/end of QE or interest rate) is going to cause massive volatility and maybe a 500 pip spike both ways in a matter of minutes.

     

    What worries me is curve fitting to avoid major news events.  ie. we give SQ just price data and it finds a strategy. Perhaps the ones that look profitable were curve fitted to open positions that fitted with big economic news releases or avoided them by chance.

     

    Since it is these big news announcements (inc NFP) that drive the markets, but in SQ and MT4 we have no visibility of them.

     

    Blind trading is asking for trouble. No professional trading would even be contemplated without looking at upcoming and recent news events to guide trading decisions.

     

    Can we add an economic news feed to SQ4 to either trade news or avoid it?

     

    Cheers,

     

    Mike

    #131791
    Customer
    477 Posts

    Yea, you can add more and more filters to the strategy, but I say if it worked well in a 14 years backtest without any news filter, then you should just let it do the same going forward. Anything else is just increasing curve matching. If the strategy is profitable for 14 years, it´s rather important that it wins long term. And news can and have triggered some really good trades too.

    #131792
    Customer
    745 Posts

    Yea, you can add more and more filters to the strategy, but I say if it worked well in a 14 years backtest without any news filter, then you should just let it do the same going forward. Anything else is just increasing curve matching. If the strategy is profitable for 14 years, it´s rather important that it wins long term. And news can and have triggered some really good trades too.

     

    But there are not that many “major” news events in 14 years.  How do you know your strategy isn’t curve fitted to miss or randomly choose what happened to be good decisions at the time?

     

    If you knew the USA was going to announce bigger QE in an hours time or maybe they might annouce an interest rate rise (first since 2008) would you expect traders to blindly open positions in front of it?

     

    Having working at a trading desk I can tell you no.1 thing that drives order flow is news and economic data.  Look at volatility just before news and it goes very low, no big traders are blindly opening positions when the price might whipsaw and take out stop losses over and over again until market settles (which might take hours or even days to adjust).

     

    What do people think here?  That a simple strategy can survive news events when the strategy cannot see them looming up ahead?

    #131794
    Customer
    654 Posts

    If your strategies are not M5 with 10 pip SL you’ll be fine. This is the known problem with automated scalping and why a lot of us don’t use it.
    (hint: higher timeframes with wider stops)
    The best you can do for lower timeframe tight stops is use real tick data, it has the spread widening build into it and there have been many NFPs in the past 14 years (168).

    More often than not major moves go in the direction of the long term trend.

    #131795
    Customer
    654 Posts

    This is also leaving the realm of technicians and quantification. By definition we are using charts and market structure because we believe outside events are incorporated into that structure and current pricing.
    This is discretionary and the best way to handle it is take to EA offline during these events and times if they’ve proven poor performance in the past in my opinion or alternatively cut the risk in half or more during them.

    #131797
    Customer
    745 Posts

    Have a look at EURUSD on 10th and 11th of July 2013 for example.

     

    http://www.forexfactory.com/calendar.php?day=jul10.2013

     

    Those two days reversed the trend completely and wiped out 2 weeks of downtrend.

     

    I’m not convinced large stop losses and higher timeframes work when these sort of severe and fast reversals are commonplace.

    #131798
    Customer
    654 Posts

    Yes but the initial illiquidity event only lasts minutes, not 2 days.
    Your concern is illiquidity and slippage obviously. I’m on a 1 minute chart during that event and the slippage was  less than 100 pips (and it was an especially out-sized event, 30-60 pip gaps are more common). Most of my stop losses are in the hundreds of pips(outside of the gap, in areas of excess liquidity). This certainly is a concern for scalping, which is why I stay away. There is much that goes into that style of trading, internet speed, broker, execution platform. The best way to beat HFTs is to not compete with them on their timeframe. The huge trading firms with servers close to exchanges rule the lower timeframes.

    The majority of events prior to the reversal drove the trend down, and the majority of moves after that reversal drove it up. They’re drivers of trend, and on rare occasion, reverse it. This should not wipe out a robust system. Moves like these happen in every market, in stocks its probably the worst.

Viewing 15 posts - 16 through 30 (of 36 total)

You must be logged in to reply to this topic.