Time, price and volume frames: The two critical factors

December 23, 2015 01:00 PM

A common question for traders looking to build a trading system is what time frame to use. 
The question itself assumes that there is a single, universal and unambiguous answer. There is not. 
The choice of a time frame—as with the choice of any other parameter of a trading strategy—cannot be made without the context of the various other parameters and market specifics of a particular trading strategy.

Any time frame is a particular case of data compression. Time frame is only a particular case of such a compression: We can aggregate ticks into time-based bars, as well as in price-based or volume-based bars, and in each case we have a new data representation which hides a number of ticks inside a single data point (bar). While time-based bars are the most well-known and widely used, other methods of data compressions can be useful for identification of certain market processes.

When choosing the data resolution, or its granularity (be it time frame, “price frame” or “volume frame”), the key criteria are the two factors: the duration of the market process that your strategy is trying to exploit, and the complexity of the setup which triggers the trade. The former determines the upper limit of the suggested data compression and the latter determines the lower limit. This means that if the exploited market process takes days to evolve then most likely we don’t need data as granular as one minute. However, if at the same time the trade setup (a sequence of events in the price-time-volume space which may initiate or close a trade) requires even tick data to be correctly recognized. Then we have to face this fact and be prepared to wait for a long time while doing backtests.

The ambiguity here mostly comes from misunderstanding the very essence of a trading strategy: It should always reflect something that actually is going on in the market (something that we call a market process), and not just technical analysis formations and/or other geometric or numeric constructions. Typical examples of market processes are, of course, various patterns of accumulation and distribution, with different scales, different volumes and made by different market participants. Therefore we can observe their traces in the form of various patterns, easy to distinguish visually on charts or not. 

As the result we can observe similar patterns at different data resolutions, or using different data compressions, and this observation is very provocative. It tempts you to think about a “universal” strategy which should work with any data compression. However, even visually similar, these patterns are caused by different market processes in which different market participants take part, and their importance is also different. For example, a double top on a daily time frame is mostly caused by behavioral patterns of long-term high volume market participants. Therefore it will have far greater importance than a similar pattern on a one-minute time frame, which in many cases is not caused by any reasonable process and is purely random.

This doesn’t mean, though, that a pattern found in higher resolution should always be preferred to those found in lower resolution. If we have strong reasons to believe that a particular pattern is not random--that it is caused by a certain market process--and as a proof of concept we can see adequate statistics, then it is possible to trade it. The key to success is understanding that the probability of a positive outcome of a trade, or its mathematical expectancy, decreases in time after the moment the pattern has formed, and this time is proportional to the data resolution.

Finally, note that there are quite a few strategies which are independent of any particular time frame; they may use certain reference prices or other data sampled at predefined times, and the very analysis of bars or any other kind of time series is not required. Strategies of this kind are the most illustrative examples of why the choice of the data compression should always follow the initial trading idea, and not forerun it.

About the Author

Alex Krishtop is an expert in systematic and automated trading. He is the director of education at Algorithmic Traders Association and runs an exclusive educational course in systematic trading. @atassn