Some general questions i need answered about automatic trading

• 10/18/2017 at 8:06 AM #49766
  jebus89

  Hi there! Thank you so much for your time to read my post and answering!

  First question: I’ve been getting some decent (not great) results on a 10 m strategy that I’m working on, but I’m unsure of how much time i should put into it tho, if the short timeframe is going to suck anyways or not.

  Are >1h strategies worth building/making/spending time on? Anyone got one or more running? Im thinking they would be less suitable because of the small datasample for your backtesting. Any other reason why >1h strats suck/are good?

   

  Question 2: When do you know if a strategy is bad or not? Let’s say you get a 60% winrate and 1,5 gain in backtest, maximum drawdown $500 and max cons. losses: 5. Then you backtest for a month and u get OK results, then you run it live. Then you start loosing money.

  When do you call it off and say “geez this didnt go as i planned”? is it when you exceed the maximum amounts of losses? Drawdown? Any real life experience is good to know! Not knowing about this has been an issue for me since day 1 algotrading/coding 🙂

   

  Question 3: Lets say you have a working strategy, makes good entries, and exit is close by, just 20 pips away (example). Your run it live and seem to get good results. Should you/could you then use the same strategy for the same entry but just another exit? (Lets say 100 pips away)

  Sure you’d loose some of that winrate% I’d assume, but maybe up the gain alot. Is that a good idea or is it a rookie mistake? As far as i can think of, the risk is being able to fuck up 2 trades from the same entry, so double your losses when a loosing trade occurs, then again if a trade goes your way you get the 20 + maybe 100 pips… Of course the numbers are just examples but the question remains: Should you run the same strategy, on the same market & timeframe, but with different exits?

   

  Question 4: How long should you paper-trade your strategy before you run it live? I realise this might be a very personal thing, some might just run strategies live from day 1 while others wait X amount of time. I’m currently running stuff on a demo account just because I’m afraid I’d loose money if the backtest proves to be curve-fitted to much.

  I’m new to the game so any real life experience is very welcome! 🙂

  

  jebus89

  Participant

  Topics: 118

  Replies: 586

  Been thanked: 238 times
  10/18/2017 at 1:13 PM #49827
  Nicolas

  1. Yes, definitely, whatever timeframe, even if you don’t have enough long history. The best you can do is to make Out Of Sample analysis, with paper trading or even by live simulation in ProBacktest. Develop your ideas on 70% of the whole history and test it over the complete one.
  2. Same as 1/, use OOS validation.
  3. Depends of the strategy itself, but if it has been designed for a 20 points SL, do not change it. Take care of your Risk Reward ratio.
  4. Same as 1/, use OOS validation. If your running strategy in paper trading is doing well, technically (same as backtest behavior) and with expected performance (DD%, consecutive wins/losses), you are about to validate it. If not, go back to your blue print.

  In fact, there is no good or definitive answers to your questions, we got all the same problems of validation 🙂

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20830

  Been thanked: 4743 times
  10/18/2017 at 3:38 PM #49845
  jebus89

  Nicolas, thank you for taking time to reply. You have given me some insight. I think that I wasn’t clear enough on question 3, let me rephrase: If you have an algo that enters the market with good entries, is it smart or dumb to duplicate the same algos with the same entry, timeframe and market, but with different exits?

   

  For example if you make an algo that follows a trend, you might want an algo that takes quick profits/secure profit along the way, but at the same time you might want one that runs longer and tries to capture the whole trend. Is it dumb to have the same entry but different exit on the same market and timeframe?

  I hope I was more clear 🙂 OOS ftw!

   

  Ps thanks the for the guide on walk forward testing!

  

  jebus89

  Participant

  Topics: 118

  Replies: 586

  Been thanked: 238 times
  10/18/2017 at 4:27 PM #49850
  Nicolas

  That’s not a bad idea (sorry for misunderstanding) you are thinking of a partial exit (which is not possible with ProOrder), so that’s the appropriate technical solution to do it.

  BUT: bear in mind that you will double your loss (in points) and not your profit this way.

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20830

  Been thanked: 4743 times
  10/18/2017 at 7:24 PM #49876
  jebus89

  Right you are Nicolas, about the loss. I need to calculate in the fact that i might face a maximum drawdown on both algos at the same time. I’m glad to hear that your positive to the idea tho.

  

  jebus89

  Participant

  Topics: 118

  Replies: 586

  Been thanked: 238 times
  10/19/2017 at 8:56 AM #49905
  Nicolas

  it has already been discussed on forums many times, but think of diversification by making a portfolio of different kind of strategies and if possible without correlation between them.

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20830

  Been thanked: 4743 times
  10/19/2017 at 11:35 AM #49929
  jebus89

  Im sure it has Nicolas! Im trying to read and absorb everything in here, but its a lot to take in! 🙂

  

  jebus89

  Participant

  Topics: 118

  Replies: 586

  Been thanked: 238 times
  10/19/2017 at 11:42 AM #49930
  Despair

  Diversification is the holy grail. Everybody I know or heard of who has some success with automated trading is running a portfolio of many strategies.

  Have them as uncorrelated as possible and watch your overall risk exposure (size of your positions). Begin with small order sizes and when the strategy proofs itself by producing profit increase the position size gradually. Do not take too big positions (IMO one of the most common mistakes ) even if it seems “boring” in the beginning.

  

  Despair

  Blocked

  Topics: 97

  Replies: 1034

  Been thanked: 217 times
• Author
  Posts