Guides › How we rate stocks
How we rate stocks
A plain-English walkthrough of the composite verdict, confidence score and trade plan.
What a rating is (and is not)
Every stock page on this site ends with a short verdict such as BUY, WEAK BUY, SKIP, WEAK SELL or SELL, plus a confidence percentage. That verdict is the output of an automated research system. It is a structured summary of several signals — not a recommendation tailored to you, and not a promise that the stock will move that way.
Think of it as a study note: “here is what these indicators looked like on this date.” Use it to learn how signals fit together. Do not use it as the only reason to place a trade.
The five inputs
The model blends five kinds of information:
- Technicals — momentum and trend from price and volume (for example RSI, MACD, and moving averages).
- Trend structure — breaks of structure, order blocks and fair-value gaps sometimes called “smart money concepts.”
- Fundamentals — valuation, growth, profitability and earnings context when the data is available.
- News tone — recent headlines scored by a financial language model (FinBERT) as positive, negative or neutral.
- Language-model research — a short written read of the recent story around the name.
None of these alone is enough. A glowing headline on a stock that is already overbought can still earn a SKIP. A dull day of news on a stock with improving structure can still lean bullish.
How confidence is calculated
Confidence is the strength of the combined score, shown as a percentage. A BUY at 70% means the model’s composite leans clearly positive. A BUY at 40% means it leans positive, but the agreement across signals is weaker. Low confidence is useful information: it often means the picture is mixed and patience is wiser than forcing a view.
Entry, target and stop
When the model can compute them, it also proposes a reference entry, a target and a stop-loss, plus a date to reassess. These levels come from the same analysis that produced the verdict. They are not live orders and they are not personalised. If a page shows SKIP, treat any levels as illustration only.
Reward-to-risk is simply how far the target is above the entry compared with how far the stop is below it (for a long-side plan). A plan that risks more than it hopes to gain is usually less attractive than one with a healthier ratio — but even a good ratio fails often in real markets.
How to use a stock page
Read the page in this order: (1) the verdict and date, (2) the plain-English summary, (3) technicals and structure, (4) fundamentals, (5) news, (6) the plan. Ask yourself whether you understand why the model leaned that way. If you cannot explain it in a sentence, the page has done its job as a learning tool — and you should not treat the rating as settled.
For a deeper look at each signal family, see How the analysis works and the individual guides on RSI, stop-losses and reading news with the model.
For information and education only. Not investment advice. Ratings are produced by an automated model, can be wrong, and may be out of date. Do your own research; you can lose money investing.