7 Things I've learned about trading from the industry's smartest people
If you want to be a successful (algo) trader, there's no better way than to follow the advice of some of the world's most savvy traders.
The first (annual) Algo Trading Summit was a huge success!
Over 2,500 registered for the event, with an average of 500 people watching the live stream at any given moment — and, so far, the video recordings have over 5,000 views. Not too shabby.
In this post, I want to share and summarize some of my takeaways from the event. I would love to know what you think about these takeaways.
Let me know by replying to this tweet (and follow us while you're there 😉).
* The complete video recordings, along with the slides and code sent to us by the speakers, are available on this GitHub repo.
Let's get started...
1. Beware of backtesting based on clean data
Clean data, such as the one you download from Yahoo! or buy from various vendors, is usually cleaned to remove deleted orders, outliers, and spikes. However live, real-time data isn't.
It's not uncommon to see outlier ticks throughout the day showing AAPL at $9,000 or at $20. If you're using stops, or measuring momentum based on sudden moves — and are not prepared for such outliers — your strategy will make the wrong decision and your returns will suffer.
Furthermore, orders with less than 100 stocks are not forced to be filled based on NBBO prices, and will most likely be directed to a dark pool or the brokers' ECN, where fill prices will be worse. Sometimes much worse.
Why dirty data is your friend (Chris Bartlett)
2. Integrating alternative data gives you an edge
"Alt" data (or non-traditional data) can add value to your strategies by providing you with additional and differentiated insights, compared to OHLC data alone.
It allows investors to use sentiment to augment fundamental analysis and to monitor changes over time. It helps investors better identify stock-specific risks, analyzing behavioral flows, use sentiment to augment fundamental analysis, etc.
What might come as a surprise is that you can scrape alt data yourself (and even sell it), and are not limited to buying it, which could be quite expensive. Cool!
Integrating alternative data into your strategies (Saeed Amen)
3. Machine learning is the future
We had several talks that took advantage of ML in their trading in innovative ways – from factor modeling and portfolio construction through good old pairs trading.
⚠️ If you need to brush up on your math before going that route (I did), check out Mathematics for Machine Learning which offers a free book
Feature selection is the new factor modeling (Ernest Chan)
Smart-Beta portfolios - a machine learning approach (Alexandr Proskurin)
Applying machine learning to pairs trading (Illya Barziy)
4. You stand no chance against HFT firms.
I'm not sure if I even had to write this one, as it is so glaringly obvious. What you can do, and probably should do, is try to lower your trading frequency so your strategy is less price sensitive.
Unless you plan to invest $100M+ in infrastructure, that's pretty much it.
I recommend you watch The Hummingbird Project movie about how far HFT firms do to shave a few nanoseconds. Fascinating.
A look inside the world of HFT: What lies beneath the surface? (Nitesh Khandelwal)
5. Predicting regimes is easier rather than predicting a "move"
In his talk, Brian Blandin suggested that trying to predict a regime (trending, reverting, zigzag) is actually the better way of doing things.
He adds that we pretty much know what is the best course of action in each market regime (trend-following when trending, mean reversion when reverting, and volatility selling/buying when volatility contracts/expands). For that reason, it's better to have a strategy that works best in each of the different market regimes — and turn them on and off, based on a market regime prediction model.
It makes sense, as trying to create a strategy that works well in all market weathers is doomed for failure.
A Scientific Approach to Market Prediction (Brian Blandin)
6. Use leverage correctly. The simple way
Leverage is a double-edged sword. We've all heard that before. That is why deciding on how much leverage to use is the single most important decision that any trader will have to make. The question is: How do we make that decision? ...and what are we targeting?
There are many fancy ways to decide on the amount of leverage one should use, like the famous Kelly formula (and the half-Kelly) and Optimal-F. But you want to keep things simple, you could simply use the strategy's drawdown as your north star.
Related talk: Down, Down, Deeper, and Down (Robert Carver)
7. Never pass on a great opportunity 😁
Tradologics is scheduled to launch the platform in the fall. However, we've opened up registration — for a limited time — in celebration of a successful Algo Trading Summit.
You now have an opportunity to create your Tradologics account and enjoy a 35%+ discount on our most popular plan!
Check out this special offer page for more details and a video walkthrough of the platform.
Don't miss this out! That page is going to be taken down by the end of the month, and when registrations are open again — the price will return to normal.