Predictive power means that an indicator, signal, or complete system has some predictive ability, predictive power, or predictive value. That is, predictive ability is an edge in the indicator, signal, or system that can result in an asymmetric reward-to-risk and/or a positive asymmetry® ratio. When an indicator, signal, or system has an asymmetric reward-to-risk and/or a positive asymmetry® ratio it has an edge, a positive probability or a positive mathematical expectation. For example, Trend following is a trading strategy according to which one should buy an asset when its price trend goes up, and sell when its trend goes down, expecting price movements to continue. So, a tactical trader applying a simple trend following signal like a cross below the 200-day moving average does so assumes the price may continue the downward trend. On the other hand, a trend follower supposes when a price trend that crosses above a simple 200-day moving average the trend is likely to continue up. That is necessarily a predictive statement about the future. To determine if it has predictive “ability”, we have to test it then operate it in a real-time walk-forward to discover if the signal leads to a greater magnitude of profit when it wins than the magnitude of loss when it’s wrong. That is, to demonstrate predictive ability, the indicator, signal, or system must demonstrate positive asymmetry ratio (a positive mathematical expectation).
Predictive power can be quantified from a market indicator, trade signal, or quantitative trading system through testing or the expectation of actual historical signals, trades, or performance history.
Predictive power can be applied to charting, technical analysis, or a systematic trading program. For charting, technical analysis, or a systematic trading program to be profitable, it must necessarily have some predictive ability and predictive value.
Predictive power shouldn’t be confused with just the probability of winning percentage. Many trading signals may have a winning percentage less than 50%. That is, after the average trade is entered and exited, they may have less than a 50% winning percentage, but that doesn’t necessarily mean it has less “predictive power”. The winning percentage of the signal alone leaves out the expectation, which is how much profit it earns when it wins relative to how much it loses when it loses. Predictive ability is a function of mathematical expectation, not just winning %, which is the probability of winning.
The key difference between those who operate quantitative systematic trading programs vs. chartist and technical analysts is quantitative systematic trading programs should be applying the entry, exit, and sizing system and testing to determine if it has predictive ability before the system is complete – a more scientific process. Properly developed quantitative systematic trading programs apply this scientific method to testing and development. Developed correctly, quantitative trading systems are created from objective observation and statistical inference.
Statistical inference is the process of drawing conclusions about data or scientific truths from data.
Inference is the process of drawing conclusions about a parameter one is seeking to measure or estimate. There are many methods of performing inference including statistical modeling.
Charting and technical analysis are often more visual and less quantified through testing. To discover predictive ability from charting and technical analysis involves either backtesting the signals or studying the past signals or trades to determine predictive ability.
Predictive ability is what separates the tactical trader who operates a system based on positive asymmetry and positive expectation from one who uses chart patterns and indicators naively without knowing if their indicator, signal, or system has predictive ability.
This definition was written by Mike Shell and may not concur with other definitions applied in other fields. This definition is written specifically to define how “predictive ability” describes trading signal or system expectation and probability. We have also shared a similar definition using the term “predictive power” and predictive value, which Mike Shell considers essentially similar, if not the same.