High Frequency Trading
HFT is a method of trading that uses powerful computer programs to transact a large number of orders in fractions of a second. (Investopedia)
Investopedia gives excellent background of the value of HFT, comparing two otherwise similar indexes. Upon first glance, two indexes look almost identical. However, when zooming into milliseconds, the two deviate from each other significantly. Hence, new opportunities open up, originally invisible at the macro level.
Chicago Booth further explains two reference indexes ES and SPY, with a great background discussion. Think, for a second, about physics theories of relativism and quantum physics, the former predicting macro physics occurrences, and the latter predicting the micro-world. Entirely different properties apply depending on the scale. Similarly, LFT (low frequency trading) and HFT follow entirely different patterns.
Low Frequency Trading
As opposed to high frequency trading, low frequency trades mean that very few trades taken over a monthly cycle. (Alpari)
- As the time period goes from long intervals to very short intervals, the correlation between the indexes goes from approximately 1 (strong correlation) to approximately .008 or less (very weak correlation).
- The usual HFT approach is to trade many times a day, with each trade having a tiny gain margin.
Meanwhile, on the internet, some entities speculate that HFT trading times will become even smaller than hundreds of milliseconds.
Surely, we can ask the question: is it technologically possible to go smaller than milliseconds? And, given the above two points, does zooming in make sense?
LFT + HFT Strategies Combined
For example, what if the traders could pick a judicious entry point, and then deliberately increase their holding time? Perhaps from 5 to 40 minutes. Longer holding times could do two things. First, the lower frequency energies could be exploited (small swings). Second, you could aim to trade a few times a day, while seeking opportunities with larger gain margins (higher efficiency).
The compounding of the [larger] daily gains could theoretically result in really large annual gains.
In conclusion, for this what-if case, an adaptive robust predictor is required.