Some thoughts on tick data for long to medium-timeframe trading

1/n

First, let's get the facts straight.
Other timeframe traders (investors) can be visible on micro timeframe data but micro timeframe traders(locals) cannot be visible on other timeframe data(>daily)
So obviously if you have the access then the logical choice is using micro-timeframe data (tick).
Theoretically sounds great but practically if you're hunting for a pattern that goes beyond days or weeks in tick data then it requires huge data to evaluate such patterns.
The workaround is to compress the tick data based on information gain rather than transforming the data based on a fixed timeframe like 1 m 15 minute and then do the evaluation.
Remember the timeframe is just for our convenience so we can read the market induced information. There is no magic timeframe and if it exists then it most likely a curve-fitted one. Ofc that's my own observation and not a ground truth
What kind of information gain? that is a huge subject and you have to be creative using your own observation. Machine learning can be very useful here.
You can also use research in signal processing such as Fourier transform. https://t.co/wmXNtJu7wg
There is an old code in my Github repo to calculate Peak and Turf which I later used in my Market profile code to get composite High-volume nodes and low volume nodes (cHVN & cLVN)
https://t.co/2PjqIfGpoH
Doing everything right still there is a limit and you will be missing some important information that is beyond the scope of tick data such as fundamental data, interest rate cycle, Commitment of trades / open interest, commodities etc.
Luckily there is no rule to exclusively use one type of data, you can always fuse and make a hybrid model n/n

More from Trading

You May Also Like