How is a stock scoring system made?
First there’s the issue of deciding which metrics are going to be part of the total score. Then, how much any of those metrics will weight on the final result. There could be, say five metrics equally averaging on the result, or there could be a weighted average, as in one metric being more important than others when it comes to summarizing.
Remember what we’re looking for, is a way to rank companies from the most attractive as an investment to the worst, even if it brings its own challenges.
Challenges of stock scoring systems
One of said main challenges are tradeoffs. Suppose you started by screening ideas and came up with two possible companies for investment. One being Tesla (TSLA), and the other Amazon (AMZN). So, you need to compare them, to decide which one promises better return on your investment at the current market prices.
Many a financial advisor would say: “Ah, but they’re from different industries, so you can’t really compare them. It’s like comparting pears to apples”. Ok, there’s some truth to that. Their economic drivers aren’t the same, and their future development most likely won’t be the same. But the need remains, you want to decide one or the other, so you must compare them somehow on an investment basis (for there’s one that will return better than the other on a specific timeframe).
In more general terms, suppose you’re comparing any two stocks. Say one of them, is a vegetative high-margin big company. The other a low margin, high growth small firm. “Which one is better?” One may wonder. But a more specific question could be: how much growth I’m willing to trade off for better margins? Some people may say,“I would always prefer higher growth than better margins”, but that may prove unwise. Say, wouldn’t you trade a tenth of a percentage point in expected growth to get double the profit margins? I know I would accept the deal. So, there must be a trading ratio of how much high margins are worth, and how much high growth is worth, and how much investors are willing to pay for them. Probably many metrics should be included, like financial strength, cash flow generation ability, among others.
That’s why scoring systems come to place. Scoring has implicit or explicit trading off ratios of how much of something they’re willing to give to get some other thing in some measure. Scoring helps make a universal metric on which to rank its members, trying to make comparable things that are difficult to compare, like companies from different industries or with very different characteristics.
But scoring isn’t a panacea by itself, for it has its own shortcomings:
Component metrics relevancy
First there’s the issue of which ratios or metrics are going to be part of the score. Significant metrics may be excluded. On the other hand, almost-irrelevant metrics may be included. That may depend on the sole criteria of the proponent of the score.
Secondly, the scoring system method may be undisclosed. You may get a magical final number, which you either believe or you don’t. Of course, scoring could be a trade secret, and therefore its construction not disclosed.
Maybe you don’t know which key metrics are included, and more probably than not, you don’t know the weighting of said individual metrics on the final result either. It may, or may not, be a simple average.
Thirdly, there could be a problem of a weighting bias in the metrics. Say for example, a scoring system is built around two key metrics: Price Earnings and Return on Equity. Problem is, both use as input the Net Profit of the company, so a one-time high earnings company will have a significant but unsustainable boost in the scoring system. The whole scoring system could be overweighted in the same inputs, making the scoring more of an illusion than an integral measure.
Another example of wrong weighting could be if the score is built around too many metrics of the same thing (say, alternative measures of growth) while very few others of other (say, profitability, or financial position), with no weighting adjustment whatsoever.
There also is no guarantee that the implicit or explicit ratio of trade off different characteristics are fair in the score. Back to our example, how much growth must be trade off to improve profit margins, or vice versa.
Implicit time horizon
On another point, scoring may have implicit time horizons. Scoring could be myopic, like rewarding higher companies that cut research and development expenses, improving short term performance at the expense of sacrificing of long-term possibilities. There probably is a time horizon implicit in the scoring system, which could easily be unknown even for its proponent. For example, rewards could maximize on average one year after the scoring high is reached (or any other timeframe). Statistical tests may be conducted to learn of the implicit time frame, mostly through backtesting.
Finally, a scoring system relies on more than one metric, requiring a larger dataset than any of the individual metrics included.
A lot of sensible criteria is needed when building a stock scoring system.
Stock scoring summary
Stock scoring pros
- Scoring tries to make comparable things that are inherently difficult to compare (eg: stocks from different industries or with very different characteristics)
- Scoring enables us to make an integral ranking, sorting on that metric.
- Scoring may measure different characteristics of the firm, providing a more holistic measure than any individual metric on its own.
Stock scoring cons
- Important key metrics may be left out of the score, while including others irrelevant or of minor importance.
- The weighting of the different metrics that comprise the score aren’t always disclosed. Said weighting could be unfair, redundant, o heavily biased.
- One single metric shouldn’t be sufficient to decide on an investment decision, even if it comprises more than one in itself. E.g.: you should still do your own independent research to get the big picture.
- Credibility issues may arise, especially if the scoring system method is undisclosed.
- A time horizon may be implicit in the scoring system, but probably unknown unless statistically backtested.
- More inputs are needed than on any of the individual metrics that are included.