MKTSTK recently came across an excellent article by trading psychologist Brett Steenbarger in which he describes how traders can expand their capabilities by embracing new perspectives. This resonated deeply with MKTSTK, as the problem of perspective can lead to numerous mistakes in trading and life.

Dr. Steenbarger made this concept tangible by talking about the different ways we can look at *time* in trading. Most financial charting platforms only present graphs based on clock time; we can access a myriad charts on a scale of 1-minute, 5-minute, 15-minute and so on… But when you really step back and think about it, this is an arbitrary convention.

We can define time in any number of ways: what would happen if we defined time based on volume, not on the oscillations from a cesium atom? Alternatively we could use the number of trades, regardless of size.

How does that change our perspective of asset prices? To illustrate our exploration, let’s look at the same stock plotted over the same interval using three different methods:

**Clock time (normal, everyday line/bar/candlestick charts)**- E.g. the blue line is the 10-minute closing price for SPY on 9/25/2015

**Volume time (advance time based on the volume of shares/contracts traded)**- E.g. the red line counts a “tick” of its clock’s time using accumulated volume
- In this case you calculate the average amount of volume that trades every 10 minutes (approx 3 million shares)
- Then count each trade’s qty until you hit the average
- One you reach the average, record the last trade price
- Reset the count and start your next time tick

**Trade time (advance time based on the number of trades, regardless of size)**- E.g. the green line counts a “tick” of its clock’s time using the accumulated number of trades
- Similarly to Volume Time, here you can calculate the average number of trades that occur every 10 minutes (about 10k trades)
- Then increment a counter for each trade until you hit the average
- Once the average is reached, record the last trade price
- Reset your count and start the next time tick

MKTSTK first encountered these concepts when trying to make a consistent Volume Weighted Average Price (VWAP) measure that made sense throughout the trading day. The problem is, if you go ahead and calculate the VWAP using an arbitrary time interval (say 10 minutes), some time periods will have a ton of volume and some periods will have none.

**Do you really think a time period with 100 million shares has the same informational value as one with 100 shares?** Probably not, so a change of thinking is in order. The solution I arrived upon in this case was to use trade time similar to how its defined above: e.g. calculate the VWAP everytime 100 trades occur.

While the chart encompasses only one day, it has some characteristics worth investigating further. In this case, clock time ticks the fastest (i.e. it accumulates the most ticks over the trading day). This perhaps implies that clock time includes long periods of inactivity which are missed by alternative times. Trade time ticks the slowest and Volume time in between the two.

These shifts in perspective might seem strange to our sensibilities, but your trading algorithms can look at time as generally as you want. **Give it the proper logical equipment and your machine can see in volume time as readily as clock time** or anything else you can dream up.

Looking at these charts makes me wonder how plotting multiple assets in volume/trade-time against one another would look…

## SliceMatrix Update

We’ve opened the signup for SliceMatrix real time, a groundbreaking platform for real-time and historical analysis of the financial market.

SliceMatrix provides tools for real-time visualization of the financial market in addition to historical analysis of asset prices and correlation over time.

Please contact us using this form if you are interested in attaining a real-time or historical license, for more information, or to schedule a demo.

Follow @MKTSTKBLG

Lead image licensed under CC BY 3.0 US from Andrey Kravtsov and Anatoly Klypin

Categories: Quantitative Trading

There is some work on “equivolume” bars here: https://www.quantopian.com/posts/equivolume-bars-slash-hurst-exponents

Similar idea, trying to down-sample trade data based something closer to volume-clock than wall-clock.

LikeLiked by 1 person

great share!

I’ve always wondered whether the U-shaped volume to time of day distribution could be modelled as a https://en.wikipedia.org/wiki/Beta_distribution

LikeLike

also, I would bet that the 10am spike in volume in Mr. Thornington’s graph is caused by the cumulative impact of economic releases such as ISM, philly fed, and umich consumer sentiment, all of which come out around then. although this effect might have diminished over the years.

LikeLike