Can I be confident with this strategy?

Discussion in 'Automated Trading' started by Outlander, Feb 4, 2017.

  1. Hi all,

    I have been struggling with a question, hope someone can help here.

    I have a strategy (let's say strategy A) that worked very well on first half year of 2016, yield a sharpe ratio of about 4.0+. It's an intra-day strategy trading futures. But after Aug, 2016, it suddenly stopped working, I lost half of my profit by the end of 2016.

    I am working on another similar strategy (B), it works well through out the data of whole year 2016 (I guess I am not over-fitting), but it's not working for data of 2015, big difference.

    The question is, should I be confident enough to continue to trade A or trade B with real money in 2017?

    Thanks.
     
    pauljherrera and SimpleMeLike like this.
  2. algofy

    algofy

    I'd trade B small size live. Assuming you didn't overfit it is profitable in the most recent market scheme. Although when market conditions change again, it may not work.
     
    Outlander likes this.
  3. Jarym

    Jarym

    I had this 'problem' with my strategies - they'd work for various lengths of time (3-8 months). It was frustrating at first but then I learnt how to figure out when they stopped working early - before I lost all my gains and then moved onto creating a new strategy.

    All strategies work until they don't. That doesn't meant to not bother using them, just means keep an eye on them.
     
  4. algofy

    algofy

    Yep, different strategies work in different market conditions.
     
    Jarym likes this.
  5. Handle123

    Handle123

    Well, 2015 was very much more choppy except for drop in August. 2016 Indexes were pretty much Bull market after January drop. And 2016 was bullish most of the year till mid August then dropped till election day. So quicker you can identify chop from trend, you can switch or adapt, check out weekly charts on www.barchart.com

    I have a scalping method that stops putting out signals when trend method starts giving signals. So perhaps what you need are some rules when volatility increases, volatility can mean larger stops and more patience to wait for smaller risk trades or lower timeframes.
     
    Outlander likes this.
  6. Thanks, that makes sense.
     
    Jarym likes this.
  7. That number (3-8 months) is exactly what I encountered. :D How do you tell if the strategy stops working or it's just a normal drawdown? I am thinking that, if the backtest performs well in last month, I'll trade next month. Is it a good estimate?
     
  8. algofy

    algofy

    This is the art part of Algo trading. It's very difficult but I'd say you need enough of a trade sample set and expectations of likely worst case scenarios and then judge your live results in accordance with a predetermined plan of when to shut it off and reevaluate.
     
    Outlander and Jarym like this.
  9. Jarym

    Jarym

    Actually, with intraday strategies its a bit simpler - I expect a certain minimum win rate (e.g. on average a strategy might have a win rate of 60% but my judgement will tell me if over a 2-3 day period it shows a 40% win rate then I immediately start investigating. I'll either stop it altogether or run it with reduced size.

    Looking at a backtest is not so useful for me - just because it worked last month doesn't mean it'll work next month. It might, or it might just work for a couple of weeks or a few days. That's why for me, monitoring and reacting quickly is important.

    Overall, I doubt I do as well as most of the good discretionary traders. However, I am proficient with coding and value not having to be at my terminal all the time.
     
    SimpleMeLike and Outlander like this.
  10. Hello,

    Good questions.

    Hopefully I can shed some help as I am learning how to properly develop a system, back test the system, and employ the system live.

    I belive you should be very confident in your system if you have the metrics from testing your system that shows profitability. All we have to support our system is past historical performance.

    Your system could be experiencing drawdown. December 2016 was a choppy market from my experience.

    Your question is a good one. Hopefully more system developers can respond.

    I have never run an Algo system and often I hear or read that no Algo system works forever. Call me naive, but I don't belive this. Yet, i have no proof of why I don't belive this, but common sense just tells me, if a trading system works, it just works. Their will be drawdowns of course.
     
    Last edited: Feb 5, 2017
    #10     Feb 5, 2017