Hypothesis and Prototype

HYPOTHESIS

Generating a hypothesis is the first step in developing a trading strategy. A hypothesis is simply an idea, or concept, that may yield an investment strategy or trading edge.

Hypotheses are either rooted in existing financial theory, or are empirical i.e. based on observation. The diagram below shows various sources of trading hypotheses.

Creating a trading hypothesis involves identifying a concept and defining it clearly so that it can be tested objectively using quantitative analysis. Below we look at two examples:

Example # 1
Source : Empirical
Timeframe : Short-term
Concept : “Tuesday seem to be a good day for the stock market to bounce off lows”
Hypothesis : “Systematically going long on Monday evenings after a significant market drop might yield a trading edge”

Example # 2
Source : Theory
Timeframe : Long-term
Concept : The “sell in May and go away” aphorism
Hypothesis :  “Exiting the stock market during the May-October period may represent a valid long-term trading strategy”

PROTOTYPE

The next step involves undertaking some preliminary quantitative research to validate or reject the hypothesis. This is also a good time to select the instrument to be used in the analysis, as well as the in-sample and out-of-sample periods. So using our two examples above:

Example # 1
Hypothesis : “Systematically going long on Monday evenings after a significant market drop might yield a trading edge”
Prototype : Go long at a 10 day low close and exit the position 24 hours later
Instrument : SPY (S&P500 ETF)
In-sample period : 1996-2005 (10 years)
Results:

The table above shows that, over the 10 year period of 1996-2005, Tuesday was the day of the week that generated the strongest one day “pops” off 10 day lows, as measured by the profit factor data. The hypothesis is therefore validated and deemed to merit further investigation.

Example # 2
Hypothesis :  “Exiting the stock market during the May-October period may represent a valid long-term trading strategy”
Prototype : Compare the historical results of holding a long position during the 6 month periods of May-October and November-April
Instrument : SPX (S&P500 Index)
In-sample period : 1966-1995 (30 years)
Results:

The above results show that over the chosen 30 year in-sample period, the stock market performed considerably better during the six months of Nov-Apr than during the 6 months of May-Oct. The initial hypothesis is therefore validated and warrants further analysis.

The prototypes in our two examples above both appear to justify further investigation. It should be noted, however, that this is not usually the case. Many, possibly most, trading and investment concepts found in financial literature do not actually withstand any form of quantitative scrutiny. The same applies to “in-house” empirical concepts. So probably 75% of hypotheses get rejected at this phase of the strategy development process.