Market Sentiment Analysis Spots Tops
Key facts
- Stock Companion's news sentiment score is a whole number from 0 to 100, where 50 is neutral, produced by state-of-the-art AI models trained specifically on financial data that read the news flow for each ticker.
- A Nature study published on 21 June 2026 combined sentiment from a finance-tuned language model with technical indicators and historical price data in an attention-based hybrid deep learning model to improve stock price prediction.
- A Nature study published on 26 February 2026 used a multilayer heterogeneous graph network to examine how sentiment relates to stock price volatility in the new energy vehicle market.
Most traders use stock market sentiment analysis the wrong way around. They treat a high reading as a reason to buy and a low reading as a reason to sell, which means they buy the most crowded names at the moment the good news has been fully priced. This article addresses a narrower problem: how to read a sentiment score so that it warns you a run is ending rather than pulling you into it. The answer is to watch for gaps between what the news says and what price does, not the level of the score by itself.
Why tops show up in news sentiment before they show up in price
A rally runs on new information. Early in a move, each headline (a contract win, a trial readout, an analyst upgrade) changes what buyers think the stock is worth, and price reprices quickly. Later, the headlines keep coming, but they repeat what everyone already knows. The sentiment score stays high because the language is still positive. The price response shrinks because there is no one left to surprise.
That gap is the core of the method. A sentiment score measures the tone of the information. Price measures whether that information is still moving money. When tone stays bullish but price stops responding, the marginal buyer is already in.
The academic work points the same way. Research published in Nature in June 2026 did not use sentiment as a standalone predictor. It fed sentiment into a model alongside technical indicators and historical prices. The February 2026 Nature study linked sentiment to volatility rather than to direction alone. Both suggest sentiment carries the most information when it is read against price behaviour, which is how you should read it too.
Three top patterns to look for in the sentiment score
Bearish divergence
This is the classic setup, borrowed from momentum oscillators. Price prints a higher high. The sentiment score, at the matching point, prints a lower high. The news around the second peak is less enthusiastic than the news around the first, even though the stock is higher. If you already watch RSI divergence, the logic is identical. You are just using news tone in place of price momentum.
Exhaustion plateau
The score sits well above neutral, in the 70s or higher, for an extended stretch while price chops sideways. This is a rule of thumb rather than a backtested threshold. The point is persistence without progress. Bulls have control of the narrative and cannot push price further.
Good news, bad reaction
This is the most actionable pattern, and it is often the earliest. A clearly positive headline lifts the score sharply, and the stock closes flat or down that day. One instance can be noise. Two or three in a short window usually means distribution: holders are using good news as liquidity to sell.
| Pattern | What the score does | What price does | Confirmation to wait for | Typical reliability |
|---|---|---|---|---|
| Bearish divergence | Lower high | Higher high | Close below 20-day moving average | Moderate; can persist |
| Exhaustion plateau | Stays well above 50 | Flat range | Range breakdown on volume | Moderate |
| Good news, bad reaction | Spikes up | Flat or down on the day | Repeat within a few weeks | Higher when repeated |
| Plain high reading | High | Rising | None; not a top signal | Low as a sell trigger |
The last row is there on purpose. A high score with rising price is a trend, not a top. Selling it is the most common error this framework is meant to prevent.
How this plays out in small caps: CRMD, PGEN and RIG
Sentiment behaves differently depending on what drives the stock. The three names below show the range.
CorMedix and Precigen are small-cap biotechs. Their news flow is dominated by regulatory, clinical and commercial-launch headlines, which produce sharp sentiment spikes. These stocks are prone to the "good news, bad reaction" pattern, because traders often position ahead of a known catalyst and sell on the announcement. On the CorMedix (CRMD) signal page or the Precigen (PGEN) signal page, the thing to check after a major headline is whether price confirmed the jump in the score over the following sessions, not just the score itself.
Transocean is an offshore drilling contractor. Its sentiment is tied to oil prices, rig contract announcements and day-rate commentary, and much of its news flow is sector-wide rather than company-specific. That makes the exhaustion plateau more common. Sector optimism can keep the score elevated long after the stock has stopped responding. On the Transocean (RIG) signal page, compare the score's persistence against whether price is still making progress.
The general lesson: in catalyst-driven names, watch single-day reactions. In macro-driven names, watch persistence.
Where sentiment-based top calls fail
Thin news coverage
A small cap with two headlines a month does not generate a meaningful sentiment series. One article can swing the score, so divergences built on sparse coverage are unreliable. Before trusting a pattern, check how many news items sit behind the reading.
Strong trends overrun divergence
Bearish divergence can repeat three or four times in a powerful uptrend before price turns. Shorting the first divergence is how traders lose money while being "right." Treat divergence as permission to tighten stops, not as a reason to reverse.
Headlines the model cannot price
A sentiment model reads tone. It cannot know that a "positive" financing announcement means dilution, or that an "in line" trial result was below whisper expectations. Biotechs and heavily shorted names generate exactly these cases. This is why the price reaction matters more than the score for those stocks.
Insider selling has innocent explanations
Pairing sentiment with SEC Form 4 data is useful. Insiders must report most transactions within two business days, so the data is timely. Many sales are pre-scheduled under 10b5-1 plans or made to cover taxes on vesting. Open-market sales by several insiders into a stretch of strong sentiment carry more weight than a single planned sale.
What to do differently on Monday morning
1. Separate trend names from top candidates. For each holding, ask whether the score is high and price is still responding to news. If yes, leave it alone.
2. Scan for the three patterns. Flag any stock showing a lower sentiment high against a higher price high, a long plateau above 50 with no price progress, or a positive headline met with a flat or negative close.
3. Require one technical confirmation. A close below the 20-day moving average, a failed breakout, or a range breakdown on rising volume. Without it, a sentiment pattern is a watchlist entry, not a trade.
4. Check Form 4 filings. Note clustered open-market sales by officers and directors over recent weeks, and discount planned or tax-related sales.
5. Act proportionally. A top signal should usually mean trimming, tightening stops or skipping new entries, not an all-or-nothing exit.
Stock Companion's multi-factor buy/sell signals already blend sentiment with technical and insider inputs. A falling composite signal on a stock with a still-high sentiment score is itself a version of the divergence described above.
Can a high sentiment score mean a stock is about to fall?
Not by itself. A high score with rising price describes a healthy trend. The warning comes when the score stays high or rises while price stalls, makes a weaker high, or falls on good news. Read the level of the score as context and the gap between score and price as the signal.
How long does sentiment divergence take to play out?
There is no fixed window, and anyone quoting one precisely is guessing. Divergence can persist through several price highs in a strong trend. Use it to tighten risk, then wait for a technical break such as a close below a key moving average before treating the top as confirmed.
Is news sentiment better than social media sentiment for spotting tops?
They measure different things. News sentiment reflects reported information and tends to move with fundamentals and catalysts. Social sentiment reflects crowd positioning and can turn faster and noisier. For top detection, news tone that stops lifting price is cleaner evidence, because it shows real information losing its effect.
Start with your own watchlist
Pick the three or four positions that have run furthest, and check each one for the three patterns above before the week starts. If you want the sentiment score, composite signal and Form 4 activity on one page per ticker, create a free Stock Companion account and add those names to your watchlist.
Sources and method: Research figures and dates come from the Nature studies published 26 February 2026 and 21 June 2026 cited above, and descriptions of the score come from Stock Companion's own signal methodology.
Stock Companion's signals are generated by an automated model and are not financial advice.