Sharing automated strategy tips
- This topic has 11 replies, 7 voices, and was last updated 1 week ago by
.
-
Making this post to share tips for creating automated trading strategies and my thoughts. My approach to automated trading is it shouldnt be trying to copy a human traded strategy. Translating human trading to automated systems is challenging due to factors like intuition that can’t be easily coded. I think automated trading requires a completely different mindset. It should maximize its strengths, particularly in short timeframe trading. I prefer not to code swing trading strategies over multiple days because I can do that better manually. Instead, automated trading should find trades that we might miss due to mental exhaustion or the inability to process multiple pieces of information simultaneously.
What I and probably most people currently lack are creative ideas. Rather than sharing complete strategies, which people wont do, esp with the marketplace (I wouldnt either if i had a winner). I think we should share concepts and experiences. Let’s discuss what you think works and what doesn’t, and see if we can collaborate. Heres some of my inital thoughts and tips-
Entries
Most indicators I’ve found are useless. If an indicator works on one market but not another, it will eventually fail. I’ve stopped using traditional indicators for entry signals since they’ve been tried in all combinations. It may work it may not, it may work for a week or a year and fail utterly. I now code entries based on price actions, such as simple high/low or fractals. Price action entries work across most stocks and are fundamental to trading and have been consistent since the beginning of markets.
Noise
A major issue we face is market noise. How do we code what humans can easily see? We can visually distinguish trending from ranging markets dynamically, but coding this is difficult because it’s always fixed. I started experimenting with entries on very close timeframes, like 1 minute and 55 seconds, to confirm signals. If it works on 1 minute, it should work on 55 seconds. I find it helps to filter out some bad or sporadic signals.
Exit
I find exit strategies to be as important, if not more important, than entries. Fixed exits, like 20 points in a dynamic market, are not sustainable long-term. I’ve been using 2 Supertrend with different values to combat noise as my standard exit strategy. I think most effort goes into creating a exit strategie/money management.
Backtesting
Optimising values always leads to failure (in my experience). Even walkforward, monte carlo (different software) etc. Optimising values of an indicator (back to my first point) and deploying the strategy just fails. I have just been using the optimisastion system to find strategies on different markets. Eg if option1=0, if option2=1 to test different indicators or snippets of code. Mainly to quickly filter out code that dosent work. To compare different snippets eg direrent exit strategies etc.
These are just a few thoughts from my personal experience. What is your experience from automated trading?
2 users thanked author for this post.
To add a few humble thoughts… I used to experiment a lot with all kinds of indicators. Often with SMAs… often in time units such as 1 minute… I often tried to incorporate my preferences, such as exiting the trade daily, or something similar into my systems. Yes… it works for a while, until the market structure changes a bit… then you notice how over-optimized these systems were. Today I mainly only use range breakout systems… often without any indicators at all. I either orient myself on daily marks such as Dhigh/dlow or highest[x](high). I also no longer set take profits, but rather work with trailing and/or time exits within a day. This way a good trade runs as far as possible and is not cut off by a take profit. As a second category of systems, I use mean reverses of all kinds in high time units such as H4. In my opinion, the most important thing is the choice of market. For me personally, DAX, for example, is quite difficult to trade. I only use SP500 and NASDAQ.
2 users thanked author for this post.
Go with the big money… ideally SP500 only long. That way you’re on the safe side. No matter whether scalp, swing or mean reverse.
1 user thanked author for this post.
Im starting to look more into a semiautomatic/manual way off using algos coz off the fact that no market stays the same forever.
So using screeners to look for stocks and then have long only or short only algos that will run untill conditions shange or dont work out as expected.
I also feel that exits are way more important then entrys.
1 user thanked author for this post.
I am not Phoentzs but one of my favourites trailing stop loss code is from you SnorreDK, from this thread: https://www.prorealcode.com/topic/nas-3m-rsih-dmh-v0-1-john-ehlers/
Do you have another favourite?
I have noticed the same thing about indicators. I havent found a robust strategy based on indicators. Usually my strategies are built with a price action as base (candle stick pattern/high-lows etc) and then with some extra filters, like trend filters (SMA,LinearRegression), volatilityfilters, trendstrength or volume. Add that with a nice trailing stop and it seems to usually work nice and robust for me.
One thing that really helped me was that when I first was developing strategies, was that I only had access to 200k units. When I later got access to 1M bars, I could try all of my 30-40 developed strategies with 2-8 years OOS. Most of my strategies was curvefitted. A few of them showed the same nice equity curve, as in 200k units. Those I trade today! A nice eye opener how valuable OOS data really is, in order to avoid curve fitting.
1 user thanked author for this post.
Find exclusive trading pro-tools on 
