Take one apart
“Bitcoin rose 4% today, driven mainly by improving market sentiment.”
One question: how does the author know sentiment improved?
Almost certainly because the price went up. There is no other observation channel — nobody surveyed global holders at the moment of the move. So unfolded, the sentence reads:
“Bitcoin rose today because people wanted to buy more. How do we know people wanted to buy more? Because it rose.”
That's a circular argument: the conclusion has been packed into the premise. Its distinguishing feature is that it is always right and therefore carries no information. Price up is improving sentiment, price down is cautious sentiment, price flat is a wait-and-see mood. No price path could falsify it.
What does a falsifiable explanation look like? Roughly this shape: “X happened at time Y, X affects price through mechanism Z, and we can see X in dataset W, which is independent of price.” Note the last clause. Evidence independent of price is the dividing line between an explanation and a restatement.
Six sentence patterns worth recognising
Recognising the pattern is far more useful than memorising any particular conclusion. These six cover the overwhelming majority you'll meet.
1. Sentiment: the result, renamed
“Sentiment improved.” “Risk appetite recovered.” “The market came under pressure.” “Bulls lacked conviction.” The test: swap the phrase for “the price went up/down” and see whether the sentence still reads normally. If it does, it was a restatement.
2. Flows: a direction with no data
“Money is leaving crypto” sounds concrete, but to stand up it has to answer three questions: which measure of money, over what window, sourced from whom? Without all three it's the same sentence as “sentiment weakened”. Worse, the intuition behind it is usually wrong anyway: every trade has a buyer and a seller, and the money you got for selling went to a buyer — it didn't evaporate.
3. Both-ways cover
“If support holds, the rebound may extend; if it breaks, further downside is possible.” This covers every outcome and therefore contains nothing. Its function is to make the reader feel that somebody is doing analysis.
4. Borrowed authority
“Analysts believe…” “Market participants noted…” “Institutions broadly expect…” These dress an unevidenced judgement as a sourced one, except the source is anonymous, uncheckable and therefore unaccountable. When you see that subject, weight the sentence at zero.
5. Correlation wearing causation's coat
“Country X released data that day; bitcoin fell 3%.” Two things happening on the same day is an extremely low bar — a great many things happen every day. Establishing cause needs at least: timestamps that match to the minute rather than the day, a plausible transmission path, and some repeatability in comparable situations. Missing any of those, this is coincidence with good typesetting.
6. Selective evidence
“Historically, every time this indicator appeared, price did X.” Claims like this almost always omit the cases where it didn't. And in crypto, many supposed regularities have single-digit sample sizes — the bitcoin halving has happened four times. Four samples cannot support a rule. That's not pedantry, it's the minimum requirement. See does price always rise after a halving.
“Every halving has been followed by a rally, so this time…” — the first half is a (selective) statement of fact, the second half is a prediction, and a “so” stitches them together. The reader believes they are consuming an explanation while actually receiving an unsupported directional call. This site treats that construction exactly like shouting a trade, and never publishes it.
Why this content persists
Not because the writers are stupid. Because both sides of the market demand it.
Supply side: there has to be a story every day. Financial content publishes daily, but genuinely explicable market events don't occur daily. On a day when nothing happened and price merely wobbled inside its normal range, “there is no identifiable driver today” is the honest answer — and it cannot be a published article. Circular reasoning becomes the standard part that fills the gap.
Demand side: readers want reassurance, not explanation. The number in your account is falling, which is uncomfortable; discomfort comes from uncertainty; and any statement at all — even an empty one — temporarily relieves uncertainty. “Because sentiment” delivers psychological closure, not knowledge. That's also why this material leaves nothing behind and yet gets read again tomorrow.
Worth sitting with: what you want and what you're searching for may not be the same thing. If you want reassurance, this site can't help. If you want to be able to read it yourself next time, keep going.
A thirty-second falsifiability test
Apply this to any explanation you meet:
-
Flip the outcome
Suppose the price had moved the other way today. Could the same author write an equally coherent piece from the same material? If yes, the material has no explanatory power.
-
Look for evidence independent of price
Does the explanation rest on anything other than the price itself? Funding rates, liquidation totals, ETF flows, on-chain transfers, official filings are all independent observations. “Market performance” is self-reference.
-
Check the time scale
Explaining a ten-minute collapse with month-scale factors (miner output, long-term holder behaviour) is a scale mismatch. Each layer's time scale is in the four layers.
The replacement: four steps you can run
None of this needs paid data or professional software. It won't hand you a definitive answer for “why it fell today”, but it will narrow the answer to one or two layers and tell you what the narrowing was based on.
Step one: is the whole market moving together?
Open any market overview and compare the majors over the same window. All moving together by similar amounts means systemic pressure — you can ignore single-asset news. Only one moving means it's that asset's own story — go find its specific event. This step eliminates most wrong directions and takes thirty seconds. The market heatmap computes it for you.
Step two: is there a matching flow or verifiable event?
To chase a systemic move further, look for a primary record you can match to a timestamp: a regulator's filing, a platform's official notice, a large on-chain transfer, an ETF creation and redemption disclosure. The standard is “attributable, timed, and I can open it myself”, not “sources say”.
Step three: check leverage and liquidations
If the move was a vertical minute-scale drop, layer 3 is the prime suspect. Look at open interest before and after, and at liquidation totals and the long/short split over the same window. A crash accompanied by enormous forced selling and a slow grind lower with none are different events, even if they look similar on a daily chart. Current readings are in funding rates and leverage; the mechanism is in liquidation cascades.
Step four: sentiment last — and only with a measurable proxy
If the first three come up empty, what's left may be called sentiment. Even then, use a proxy that is independent of price rather than a feeling. Two common ones: funding rates (how crowded derivatives positioning is) and published sentiment indices.
This is why sentiment sits last in our framework as a residual: it is the name for whatever the first three layers couldn't account for, not an independent cause. Naming a residual is not explaining it.
| Filler version | Checkable version |
|---|---|
| Sentiment weakened and bitcoin came under pressure | Majors fell together by similar amounts in the same window, equity indices fell alongside, and no primary crypto-industry event was found |
| Profit-taking triggered selling | Liquidation data for the window shows long liquidations far exceeding shorts, with the decline concentrated in a few minutes |
| Positive news landed and capital flowed in | An approval document was published on date X and can be found on the regulator's site; most of the price move occurred before publication |
| A whale dumped and caused the drop | A large transfer to an exchange is visible on-chain, but whether it was subsequently sold cannot be confirmed, and neither can any causal link to this move |
Three documented moves to practise on
All three are extensively documented. They're used here only to demonstrate what step one looks like — no extrapolation, and no suggestion that any of it repeats.
Case one: the market-wide collapse of mid-March 2020
Verifiable facts: during the global pandemic shock, many asset classes fell hard within the same few days, US equities triggered circuit breakers repeatedly, gold — normally considered a haven — also fell for a stretch, and bitcoin lost an extreme amount over two days.
Filler version: “Panic spread; bitcoin's safe-haven thesis was disproved.”
Checkable version: step one already shows this as market-wide, and not just within crypto — across nearly everything sellable. That's the signature of a liquidity shock: when institutions need cash they sell what they can sell, not what they ought to sell. The driver sits in layer 2 (macro), with layer 3 (leverage) supplying the magnitude.
Case two: the collapse of a major trading platform, November 2022
Verifiable facts: a then-leading platform experienced a run, halted withdrawals and entered bankruptcy proceedings within days; the court filings are public. Crypto fell sharply over the same period while equities showed no comparable synchronised drop.
Filler version: “Confidence collapsed and risk appetite deteriorated sharply.”
Checkable version: step one returns crypto alone, which eliminates layer 2 and points directly at layer 4. And the event has primary sources: platform notices, court documents, visible on-chain movements. This is the easiest category in the framework to reach a firm conclusion about.
Case three: the cross-market drop of early August 2024
Verifiable facts: around 5 August 2024 Japanese equities suffered an unusually large single-day fall, risk assets weakened across several markets, and crypto declined in the same window.
Filler version: “Crypto suffered a black Monday.”
Checkable version: market-wide again, with the origin outside crypto. Hunting inside the crypto industry for a cause (which project blew up, which whale sold) would have been wasted effort — step one already told you the answer isn't there.
Every conclusion came from the thirty seconds of step one: is the whole market moving, and how wide is “whole”. It requires almost no expertise and eliminates most of the wrong directions. Nearly everyone skips it, searches “why is it down today”, and accepts whatever the first result says.
Three questions for any number you're shown
The difference between filler and hard material usually shows up in whether a citation survives interrogation. For any figure, ask:
- What exactly is being measured? “Net inflows of $500 million” — into what? Exchange spot accounts, ETF share creations, or an on-chain address cluster? Different measures can mean opposite things. ETF net-flow figures are the most frequently misread of all; see how ETFs affect price.
- Over what window? The same number over one hour, twenty-four hours or thirty days implies completely different magnitudes. A number without a window is not a number.
- Who compiled it, and can I open it myself? Primary sources (filings, official notices, block explorers, exchange reference pages) beat secondary aggregates, which beat “it was reported”. Aggregated data isn't unusable, but you need to know it was stitched together from various endpoints with methodologies that differ by site.
None of that requires financial training. It only requires refusing to let vagueness pass. A number that fails those three questions carries the same information as “sentiment improved”.
When you finish checking and still can't say
This is the most common outcome, and the one this article most wants you to accept.
Markets produce a large volume of moves with no particular cause — the aggregate of countless independent decisions. In a market with no daily limits, no closing bell and wildly varying liquidity across the day, a move of a few percent doesn't need a reason.
So when all four layers come up empty, the correct conclusion is: “this move has no identifiable single driver; it falls within normal variation.” Unsatisfying, but true — and more useful than a fabricated cause, because it doesn't leave you with a false causal model to apply to the next decision.
No daily recaps, no directional calls, no price targets, no timing advice. Every “why” page is written as method, because method still works in six months and same-day attribution is dead in three days. Those rules, and how the content is produced, are set out on the about page.
You write filler too: three self-checks
Circular reasoning doesn't only happen in articles. It happens in your head. The three most common versions share one feature: they all occur after the event.
One: retrofitted narrative. The price finishes falling, you remember a headline from two days ago, and “ah, that was it”. But before the fall, that headline meant nothing to you — you selected an input that fits, with the outcome already known. The test is simple: did you write it down or tell anyone at the time? If not, it was assembled afterwards.
Two: selective memory. You will clearly remember the time you thought “this feels like it's going to drop” and were right. You will barely remember the equally numerous times you weren't. That's not a character flaw, it's how memory works: confirmed judgements come with an emotional spike, disconfirmed ones quietly disappear. The only defence is recording predictions before the outcome.
Three: mistaking “I understand” for “I can anticipate”. This is the one readers of this site most need to watch, and the easiest trap to fall into right after the mechanics click. Understanding the four layers genuinely improves your explanatory power, and people readily mistake explanatory power for predictive power. They are different things: explanation attributes causes to what already happened, prediction judges what hasn't. Progress in the first does not automatically transfer to the second.
Common questions
Are financial media lying?
Mostly not deliberately. It's a structural consequence of format and output requirements: daily publication demands a story every day, and explicable events don't arrive daily. Once you see the mechanism you don't need to be angry about it — just adjust your expectations and treat that content as an index of what happened, not as an answer to why.
Does sentiment really have no effect on price?
It does, and considerably so over short horizons. What this article rejects isn't “sentiment matters” but using sentiment as an explanation — because in the absence of independent observation, the word is just a synonym for the price. To talk about sentiment you need proxy data that doesn't derive from price, and you have to admit those proxies have severe limitations of their own.
Will my conclusions differ from the media's if I run the four steps?
Often, and usually yours will be more conservative — for instance the media says “because of announcement X” and you find the bulk of the move happened before X was published. That situation is extremely common; the reason is in why the price already moved before you saw the news.
Is there a faster way to judge whether a piece of news matters?
Run it through the news impact classifier first to place it in a layer and see the mechanism it usually works through — plus what it commonly does not mean. It won't tell you where the price goes; it only tells you what category you're holding.
Will this method improve my returns?
We don't promise that and have no basis to. What it can do is stop you building a few false causal models, and mean that during the next sell-off you know what to check rather than hunting for a statement that makes you feel calmer. Whether that converts into returns depends on decisions we play no part in.