What time does the trading day begin?
Before answering, ask which market, which exchange, which time zone and which definition of a trading day. Suddenly a very simple question has become complicated. This is what happens when we confuse the time displayed on a chart with the time experienced by the market.
A clock is a convention. The movement of time is not.
The exchange can open at a particular hour, close at another and change its schedule for holidays. A charting platform can display New York time, London time or the time on your computer. None of these choices changes the sequence in which transactions occurred. They change how we label that sequence.
And labels matter when we analyze data.
Suppose you build a daily model using closing prices. One vendor defines the daily close using the exchange settlement. Another uses the last traded price before a session break. A third creates a continuous twenty-four-hour bar ending at midnight. All three may be describing the same market, yet their daily returns will differ.
Now imagine testing a relationship across ten years of data without noticing that your vendor changed its session definition halfway through. Your research may uncover an astonishing new pattern. Congratulations. You have discovered a software setting.
The distinction between clock time and market time becomes especially interesting when activity varies throughout the day. Five minutes during a major announcement can contain more transactions than an hour during a quiet session. The clock says five minutes. The market may have experienced an enormous amount of activity.
That does not make clock time irrelevant. It makes it one possible way of measuring duration.
We can sample by elapsed seconds, by number of trades, by volume, or by changes in price. Each choice emphasizes different behavior. None should be treated as automatically superior in every circumstance.
The practical lesson is remarkably unexciting: define your time conventions before testing anything. Record the exchange time zone. Record daylight-saving adjustments. Record the session boundaries and the treatment of holidays. If you combine datasets, make sure they agree on what a timestamp means.
It's not glamorous work. No one is going to invite you onto financial television to discuss the heroic act of checking time zones.
But there is little point searching for repeating patterns if the calendar itself keeps changing underneath the data.
The market may be difficult to understand. At the very least, we should know what time we think it is.
Services
