Let’s be honest, finding a real edge in the markets is tricky. So I thought it would be fun and potentially beneficial for everyone to start a discussion about the types of strategies we are all either researching or trading on Alpaca right now. Let’s come together as a community and find out where all the alpha is hiding!
I know some of you may be a bit reluctant to give away any hard-earned secrets, and that’s quite understandable. So feel free to be as detailed or vague as you’d like.
Currently, I have been researching and backtesting strategies based on SPX gamma exposure from the options market. This is very watered down, but the general premise is to try and estimate the direction of dealer hedging activity based on the assumption that market makers want to stay delta neutral.
Here is a white paper that discusses the theory in more detail. I have been trying to build a long/short day trading system around this based on the previous days gamma exposure and the SPX spot price at open.
So… What kind of strategies are you all researching or trading right now??
Just now coming across this thread but I really like this idea.
While I find a lot of “bug-fixing” discussions on these forums, I’d really find a lot more value in strategy discussion.
While I don’t have much to offer as far as strategies are concerned, I can offer up plenty of findings around tooling that others may find helpful.
Here they are:
Strategy signaling: Strategies are written / edited in TradingView’s pinescript as studies. When a buy or sell alert triggers, a TradingView notification sends a payload on to AWS lambda
Order execution: - My own code that talks directly to the alpaca API, resides on AWS lambda, and only gets executed when a strategy triggers a “buy” or “sell” order. This approach gives me cloud-reliability at minimal cost.
Strategies - Quite new here (hence the discussion) but have so far found a stochastic RSI strategy with positive results. Still on the hunt for other strategies.
Backtesting - Right now I am primarily using TradingView’s backtesting platform to decide whether or not a strategy is worth pursuing. I’ve fallen into using TradingView due to the ease of use of Pinescript and the fact that using TradingView’s platform means I have to setup that much less infrastructure myself. Generally, if a strategy meets enough criteria there, I decide to move forward with more extensive backtesting to see if other symbols (only trading equities at the moment) have similar success. Since TradingView doesn’t have anything like a user-API to automate testing with its backtesting engine, I have to resort to some sort of webscraping tool. For this, I’m using UI Vision to run macros for thousands of symbols for a single strategy. While this is way better than manual testing, it is a fairly slow and buggy automated way to test strategies in general. Next steps here are to transition strategies into a better backtesting engine such as backtrader, where backtesting could be purely programmatic.
I have been digging into this for a bit now and here is what I’ve found:
I don’t have the nerve to watch a trade for a long time.
This puts me in the scalper(ish) category. I want to be in and out relatively quickly and in the same trading session
Simple strategies work best IMO
Try different platforms so you can find one that suits you.
Backtesting is not always as clear-cut as it may sound.
If you’re using technical indicators, VERIFY THE FORMULAS
There is an overwhelming amount of free information. It is very easy to end up with analysis paralysis
About knowing what kind of trader you are:
Strategies will vary depending on your “nerve” when trading. I found out (the hard way) that, once I place a trade, I tend to watch/obsess over it and it negatively affects the trading strategy because I immediately want to “tweak” something that almost always ends up making things worse.
With that in mind, scalping and a completely automated way of trading is right for me
Trading Platforms
Try as many as you can. You’ll get a feel for what works for you.
TradingView
This is a pretty good platform to get started
A bit expensive if you want to get more than just basic functionality
A good way to start with strategy building
Pinescript is a good starting language
QuantConnect
More powerful but also more complicated
The Python programming language is all the rage in the financial world so it’s good to know it
Backtesting in QuantConnect is more thorough and their portfolio analyzer (Pyfolio??) is pretty detailed
Blueshift
Think QuantConnect with live trading attached to it.
Uses the same backtesting engine (zipline) than QuantConnect
Python
Backtrader
If you’re comfortable programming in Python, you should check this one out
It works well with Alpaca
It’s very complete
Flexible
Directly connecting to Alpaca
This is what I’ve been doing lately
You have the most flexibility at the expense of “you have to code it all”
I can choose the programming language (more on this later)
Trading Strategies
Following the “simple is better” opinion, I’m using a variation of this
I’m using 1-minute bars instead of the recommended 2-minute ones
I’m completely ignoring Bollinger bands
I confirm with an adaptaion of the “Choppiness” indicator by TradingView (I adapted the thing to Python and Golang) to avoid “choppy” markets
Methodology
When I started with this, I started with QuantConnect until I found Backtrader. I did my entire set of strategies and indicators in Backtrader.
I then started looking at the examples in the alpaca github page and started collecting ideas from their example strategies
Once I was happy(ish) with the screeners, indicators and strategy, I ported the entire thing to Golang (less memory/resource utilization, smaller sizes and faster than Python). I gained size/speed at the expense of a lot of coding.
Now I’m paper-trading my strategy in a completely automated way:
The screeners selects the symbols to trade during that day:
Screener1 is based on the Momentum-Trading-Screener example in the alpaca github pages
I instantiate a “worker” (asyncio when in python and go-routines when in golang) per symbol
The strategy sells everything 15 minutes before market close
The strategy only evaluates new trades if the market is open
I let it ride and (other than killing the Docker container where it’s all running), I can’t “tweak” the strategy thus, eliminating my biggest obstacle (myself)
Verify the formulas
While working with Backtrader, I found out that, some of the technical indicators are not always programmed in the same way and don’t always follow the instructions of the original author. This made me verify the formulas for each indicator I’m using just to make sure that I am getting the right numbers
My name is Jason and you might have seen me around the forum/slack as I am currently an Intern at Alpaca. After reading through this post and the replies I would like to propose an idea. Each week I will post a weekly edition of “Where is the Alpha Hiding” discussion. Here Alpaca Community Members can discuss and share some interesting trading strategies they have been working on. Each post can range from only discussing the idea all the way to sharing the entire code for the strategy.
Throughout the week, other community members can vote on interesting ideas/topics by liking the post(s) that is the most interesting to them and at the end of each week (Sunday) we will have a winner, based on the number of likes. Then we perhaps can have the winner featured in an article on Alpaca Resources or in the weekly newsletter.
Right now, each community member would be limited to posting one strategy per week. There would not be any limits obviously on posting comments, suggestions, feedback, on other people’s strategies, however.
Are people interested in this? Please let me know your thoughts.