How to judge a stock score before you trust it

A score is one organisation's answer to one specific question. It isn't a verdict on the stock. Before you read the number, find out the question: is the price high or low against an estimate of worth? How does the stock rank against similar stocks? What do analysts think?

Used well, a score is one input into your own research. It isn't an instruction or a prediction.

A 2025 National Bureau of Economic Research working paper (not peer-reviewed) found that "the median individual investor spends approximately six minutes on research per trade on traded tickers, mostly just before the trade; the mean spends around half an hour" (NBER).

A practical sorting, not an official taxonomy.

You don't need a provider's full formulas to judge a score. You need to know what the number is for, what goes into it, and how far to trust it.

1. What does it measure? Look for a plain sentence finishing "this score measures…". Vague words like "strength" or "potential" are a warning sign. An invented example: after a year of gains, a score based on price movement might rate a stock highly while a valuation rating rates it poorly. Both can be right about what they measure.

2. What time horizon does it cover? Morningstar's star definitions refer to a "multiyear time frame". Many scores don't say: Seeking Alpha's help page for its Quant Ratings explains what is compared, but not what period the rating covers. A score built for a few months and a plan built for ten years can disagree.

3. What is it being compared against? Peers, the market, the stock's own history, or nothing. Seeking Alpha compares each stock with others in its sector, so a high score there says nothing about the whole market. A provider should say, in broad terms, what its number is a position on.

4. What inputs does it use, and what is missing? Seeking Alpha says its ratings draw on "the company's financial statements, the stock's price performance, and analysts' estimates". What happens when an input doesn't exist? A currency has no earnings. A good methodology says whether a gap is left out, treated as neutral, or counted as zero, because that changes the result.

5. How current is the underlying data? Separate how often a score refreshes from how recent each input is. A daily score can rest on a months-old company report, because a refresh can't make an input newer than its source. US regulators warn that AI-generated information "might rely on data that is inaccurate, incomplete, or misleading".

6. What does AI actually do? "AI-powered" may mean a model produces the score, writes the explanation of a score produced another way, or gathers information. Ask whether AI text can change the number, and whether the figures are calculated separately from the prose.

7. Can you see why, and what would change it? A score without reasons is a claim, not research. Look for the supporting and conflicting evidence, and what would move the score. Seeking Alpha says that on every stock page "you can see the underlying metrics that lead up to each of the Factor Grades".

8. What verifiable evidence supports it? Is there a record of how past scores worked out, over what period, against what benchmark, with rules fixed beforehand? Is any result a backtest or a live record? No published record isn't automatically wrong, but then there is nothing to weigh. Treat any claim of predictive ability with caution.

• What happens next. Moneysmart's forex guidance puts it bluntly: "Currency markets are extremely difficult to predict."
• Whether it suits you. A score doesn't know your goals, timeframe or risk tolerance. Moneysmart says the same of AI: it "is unlikely to understand your full financial situation, goals or risk tolerance".
• How much to invest, or when to exit. A score isn't a plan, or personal advice. Moneysmart's AI guidance says: "if you need personal advice, seek it from a licensed adviser".
• Whether one source is enough. The SEC's bulletin says investors "should not rely solely on an analyst's recommendation".

What there is to analyse changes with the asset class, so a score built for one can't simply be pointed at another.

Our reading, not a regulator's: because the inputs differ, a score for one asset class shouldn't be compared directly with a score for another.

The full explanation is on the methodology page.

What the Conviction Score represents. It is Olbera's own 0–100 summary of how strong and how consistent the research evidence is for a research case. It draws on five areas of market evidence, where each applies: technical, news and sentiment, macro, fundamental, and market flow and positioning. Each day Olbera researches stocks, ETFs, forex and crypto, and selects the strongest research case in each asset class.

How to read it.

• As a summary of one research case on the day it was produced, on Olbera's own scale. It isn't a percentage, a probability, a forecast or a return. Its colour and any Conviction Level label shown with it are separate cues, not ratings or recommendations.
• Alongside the rest of the case: the direction, the research horizon, the supporting and conflicting evidence, and the key risks.
• Not across asset classes, or against other providers' scores. Each asset class draws on different evidence, so the scores shown for different asset classes aren't designed for readers to compare directly.
• Not as a ranking. The Conviction of the Day is selected through Olbera's internal research process, not simply by picking the card with the highest displayed score, so it may not carry the highest number on screen.

Olbera's Terms of Service describe its content as general research and analysis for informational and educational purposes, not personal or personalised financial advice. That is how Olbera presents its service, not a statement by any regulator about how it is classified.

Australia. In March 2026 ASIC announced that its consumer site Moneysmart had published guidance for Australians using publicly available AI tools for financial questions (ASIC). The guidance covers general-purpose tools such as ChatGPT, Claude and Gemini, not specialist research products. It says such tools "can help with learning and research on general topics" but have "important limitations", and that output "should always be checked against trusted, independent sources to verify claims before acting on any information it gives you".

United States. In January 2024 the SEC's investor education office, FINRA and NASAA jointly issued an investor article on AI and investment fraud. It cautions against relying on AI-generated information generally and doesn't assess specialist research products. Its advice: "Be wary of claims—even from registered firms and professionals—that AI can guarantee amazing investment returns", and "Confirm the authenticity of underlying sources and review multiple sources of information before making investment decisions".

Both treat AI output as something to verify and warn against guarantees. Neither assesses Olbera. This guide is Olbera's own: it has not been reviewed or endorsed by ASIC, the SEC, FINRA, NASAA or any other regulator.

What does a high score tell me? Only what the provider's method found, in answer to the question that method asks. It isn't an instruction.

What is Olbera's Conviction of the Day, and is it the card with the highest score? Each day Olbera highlights one research case from across the asset classes as the Conviction of the Day. It is selected through Olbera's internal research process, not simply by picking the card with the highest displayed score. The displayed scores are on Olbera's own scale and aren't designed for readers to compare directly across asset classes, so the Conviction of the Day may not be the card with the highest number.

Is Olbera's Conviction Score a probability? No. It describes how strong the research case is today, based on the evidence available. Read more on the methodology page.