Shared Strategies/Algorithms & Successful or Failed Autotrading Bots
- This topic has 8 replies, 5 voices, and was last updated 6 hours ago by
.
-
Hi all, I have just joined this forum in the hope of finding strategies that work for autotrading. I previously backtested hundreds of strategies involving MACDs, EMAs, OBV & more. (Although at the time I didn’t realise I could change the timeframe, I now do it with 200k units). I found quite a few successful backtests, out about 40 bots live on various markets from S&P500 to Oil and FX. I even made a video on it: https://www.youtube.com/watch?v=VPINK8mA0Uc
After about 3 months all had failed.
I understand that generally people don’t want to share algos or they can be bought, but perhaps this thread could be to share just the combinations of indicators etc? Or maybe there is already a post on this?
I’m currently remodelling my strategies from before to include Fibonacci pivots, the momentum indicator and using trailing stops instead of clear cut exits, which seem to be having a positive effect.
Any ideas, much appreciated 😊
1 user thanked author for this post.
I think you’re overcomplicating things. Combining and optimizing numerous indicators rarely improves a strategy. If anything, it’s one indicator that makes logical sense. Find a significant advantage in the market that has persisted for many years and try to leverage and improve upon it. Just as an example… we know that the S&P 500 always has the tendency to rise. That’s the advantage. And you can use it. So, there’s a high probability.
3 users thanked author for this post.
Agree with Phoentzs. One thing you could try to fresh your mind a bit, is to throw all indicators in the bin, and just really play it super simple. Play with breakouts from dclose or dopen. Try to exit EOD (as that also saves your from lots of overnight fees) or with some basic ATR trailing stop you can find on this forum. No optimization needed, just pure simple price action can get your further than you think!
3 users thanked author for this post.
Yep, Phoenz and Snal are on the money.
If your going to use these known calculations keep it simple. My best (long term robust) bots that have run for years have less than 20 lines of code and pass the pub test, meaning they make logistical sense like Phoentz described.
Moreover, this is largely about expectations. If you’re a manual trader that has edge it can be very difficult to find a bot to perform as you may expect. I think few get that far but can be done for sure. But that is not the starting place. Unfortunately, the starting place algorithmically is to find a a 42%ish WR with a gain/loss of 1.3+. This just shows you it is possible and your process has evolved to “I now know something”. Once you find that SIMPLY (not easily) then look around and raise your expectations.
May I suggest running through your known indicators and testing them into oblivion and record that data as your own data base. Maybe this is what you have done I don’t know? Then you will have a short list of calcs you may be able to use to write a set of better than nothing bots. You might even get something decent, but won’t quite be able to live of it.
For me, once I’d done all that it was then possible to completely melt my brain by mostly throwing it all out and start to think outside the black box to where the real creativity comes in.
I’ll go one further and save you years of pain. Forget entries and exits, they come after structure and logic. Any indicator will get you acceptable entries and exits. But what is the Regime? How is the market structured for this trade? I just spent a year of looking to find a 49% D grade edge just needed the appropriate regime classification to take it to 81.5% and PF of 3 occuring twice a week. No joke.
3 users thanked author for this post.
Thanks everyone , really appreciate all your feedback and it’s given me a lot to work with. A few points…
“Unfortunately, the starting place algorithmically is to find a a 42%ish WR with a gain/loss of 1.3+.“ thanks for this, yes I’ve found quite a few over that, although sometimes number of trades is only like 10 or less, so only counting the ones over say 30 trades in the backtest, is this reasonable amount?
Thanks for the simple dclose idea, just did a quick backtest it looks promising. Although the gain/loss ratio hasn’t been great so far. But will keep trying.
“I just spent a year of looking to find a 49% D grade edge just needed the appropriate regime classification to take it to 81.5% and PF of 3 occuring twice a week. “ Apologies @ can you explain this in more basic talk, what’s a d grade edge?
The Regime side is really interesting, thanks , have a lot to learn here.
yeah I’m keeping a reasonable record of systems and going back and trying to improve them. so might publish some one day.
1 user thanked author for this post.
Buy the SP500 blindly on Monday and sell it on a specific day of the week—whichever one is up to you. Then compare the result with one of your existing systems and decide for yourself which is better. That’s the rough draft.
1 user thanked author for this post.
Find exclusive trading pro-tools on 
