Sentiment, Narratives and an Evidence-Based Research Note

A persuasive market narrative can combine real facts with unsupported conclusions. Your research note should make that boundary visible. This lesson brings charts, protocol information, on-chain data and sentiment together in a short explanation that a reviewer can question and reproduce.

Education only. Crypto assets can lose substantial value or become worthless. No lesson requires buying an asset, depositing money or sharing real passwords, private keys or recovery words.

What sentiment measures

A restaurant owner in New York can sense a busy night by the noise from the street. Noise is a real signal, but it can mislead: a parade is loud, too. Sentiment data tries to measure the market's mood from prices, social media and search behaviour.

The best known is the Crypto Fear & Greed Index published by Alternative.me. It runs from 0 ("extreme fear") to 100 ("extreme greed") and currently covers Bitcoin only. Its methodology page lists the ingredients. The publisher marks surveys as paused. The table lists published components without assuming how a paused input is reweighted.

Notice that half the index comes from price behaviour itself. A big fall raises volatility and lowers momentum, which pushes the index towards "fear". So the index partly re-describes the chart you already saw.

Social volume counts how often a coin is mentioned on social media. It can be inflated by bots and coordinated posting. Google Trends shows search interest. Google explains that its figures are scaled from 0 to 100 relative to all searches in that place and period, where 100 is the peak for that search, not an absolute count. It uses a sample and contains some random noise. A score of 100 in one country and 100 in another can mean very different numbers of searches.

Sentiment is real information about mood. But mood can be manufactured, which brings you to the limits.

They do not directly measure future demand at a particular price. People who post are not necessarily representative of all holders or potential buyers. A search spike may reflect a security incident rather than positive interest.

Verify the central claim and promotional incentives

On 3 October 2022, the US SEC announced charges against Kim Kardashian for promoting a crypto token on Instagram without disclosing that she had been paid US$250,000 to do so. She agreed to pay US$1.26 million to settle the charges. SEC Chair Gary Gensler warned that a celebrity or influencer endorsement does not mean an investment is right for you.

That case shows a wider problem. Sentiment and "on-chain insight" posts can be paid for, coordinated or written by people who already hold the asset. Social volume can be bought. A chart can be cropped to the period that supports a story. When a post pairs a striking chart with a strong opinion, ask who benefits if you believe it.

The other trap is false cause. Ice-cream sales and sunburn rise together in a hot summer in Italy, but ice cream does not cause sunburn; the sun causes both. In crypto, "exchange outflows rose and the price rose" may share a cause, such as good news, or be coincidence. Correlation, which you met in the chart and context lessons in Level 7, describes moving together. It never proves that one thing caused the other.

With all these limits, how can on-chain and sentiment data still be useful?

If it claims a regulator approved something, inspect the official document and its scope. A project's official source is useful for its announcement, while product performance and legal implications may need other evidence. Do not attach an official link to a broader conclusion that it does not establish.

Separate facts, calculations, estimates, hypotheses and forecasts

Facts are documented claims about current or past conditions. Calculations follow stated inputs. Estimates use a methodology with uncertainty. Hypotheses propose an explanation or relationship. Forecasts describe future outcomes. A note can contain all five, but it should identify them rather than make every sentence sound equally certain.

For example, a dated unlock schedule is evidence; a percentage increase in available units is a calculation; likely seller behaviour is an estimate or hypothesis; a future price fall is a forecast. The steps do not automatically prove one another. A natural explanation says why one fact may matter and then names the assumptions that connect it to the possible outcome.

Test contrary evidence and invalidation

A useful note tests the strongest reasonable alternative. A token may face an unlock while demand or liquidity also increases. A rising address count may reflect bots. A price breakout may occur on a thin venue. Include the evidence that weakens the preferred interpretation rather than hiding it in a separate file.

Define an observable invalidation condition. If a thesis depends on a feature launching by a date, failure to launch changes the thesis. If it depends on broad activity, evidence that one bot creates most activity weakens it. Invalidation concerns the research claim; a trading stop is an execution instruction and may serve a different purpose. Both should be explicit if a later practice plan uses them.

A detective in Mexico City does not solve a case from one footprint. She looks for several independent clues and stays open to the possibility that she is wrong. Crypto data needs the same approach.

First, prefer evidence that comes from different sources. Price, the Fear & Greed Index and momentum indicators are partly the same information, so three of them agreeing is not three clues. An on-chain metric, a derivatives metric from the derivatives data lesson in Level 7 and a macro fact from the chart and context lessons in Level 7 are more independent.

Second, avoid overfitting: tuning a rule until it fits the past perfectly. With enough metrics and enough settings, you can always find a combination that "would have" called every top. That rule describes the past, not the future. Write your rule down before you look at the result, and keep the failures.

Third, write each observation with its source, metric definition, time and whether it was raw or adjusted. Then add what would prove your reading wrong, as in the chart and context lessons in Level 7. Now practise.

Write the bounded research note

A concise note contains the question, exact product, dated evidence, calculations, interpretation, alternatives, unresolved issues and next step. It can conclude that more information is needed. Avoid a price target if the evidence does not support one. A well-documented declined action can demonstrate stronger reasoning than a confident trade based on a popular story.

No single chart, flow metric or sentiment measure guarantees an outcome. Combining weak correlated signals does not make them independent proof. Review the underlying definitions and dependencies first. The next level translates a hypothetical idea into a risk budget and records, so even a plausible thesis is not allowed to override affordability or execution uncertainty.

Try the following questions aloud or on paper before opening the worked answers. Compare your reasoning as well as your final answer.

References and further reading