Chapter 21 · Why Crypto Keeps Repeating Cycles
Why does every cycle seem to replay the same script?
- Capabilities practiced
- System ThinkingFinancial LiteracyResearch
- Hands on
- Find turning points in stablecoin supply, open interest, and major-coin prices over the last two cycles; see which moves first.
- AI Lab
- Have AI summarize the shared structure of three market cycles, then identify where it mistakes correlation for causation.
A Real-World Problem
People who have spent three years here share an experience: newcomers' excitement creates a strange sense of déjà vu.
It is the familiar language, beyond another price rise.
During rallies: this time is different; institutions are coming; retail has not arrived; this sector changes everything; now is the best time.
During declines: it is finally over; regulation will erase the industry; only fundamentals survive; those were always bubbles.
Speakers change each cycle. Sentences do not.
Even the order of events repeats: a new direction, rising prices, more leverage, derivative variations, major incidents, a long quiet period, and the next cycle.
Once might be coincidence. Three or four repetitions suggest a recurring structure.
What is that structure?
Thought Experiment
Shrink the world to a closed market.
Starting point: 100 people each hold 100 in cash, totaling 10,000. An asset has 100 units, trading steadily at 100 each.
Cash is fixed at 10,000. Price reflects the relationship between money and units.
Step 1: outside money arrives.
An additional 5,000 enters. Cash becomes 15,000; units remain 100.
Price reaches 150. Nobody necessarily became smarter or made a better judgment. More money alone explains this simplified rise.
Step 2: borrowing becomes possible.
A lender advances cash against assets at 50% of market value.
The 100 units worth 15,000 support 7,500 of borrowing. Most borrowed cash buys the same asset, increasing purchasing power to 22,500.
Prices rise, collateral becomes more valuable, and borrowing capacity increases again.
Notice: borrowing capacity depends on collateral price, while borrowed funds push that price higher. The loop feeds itself.
Step 3: a story develops.
An essay explains scarcity, inevitable adoption, and the supposedly low current price.
It produces nothing directly but attracts outside money and encourages more borrowing.
Capital and leverage increase, prices rise, and the rise appears to validate the story.
Step 4: inflows stop.
No bad news is necessary. Without fresh cash, prices stop rising.
Some leveraged short-term holders exit. Their selling lowers price a little.
Step 5: the same mechanism reverses.
A 10% decline shrinks collateral and triggers forced sales.
Selling lowers price and triggers further liquidation. Lenders tighten standards and ratios, reducing purchasing power.
The story reverses too: its author may now explain why the sector was flawed from the start.
Price falls from 150 to 60, below the start, even though total cash may have declined by only 2,000.
The whole script uses three ingredients: cash flows, borrowed money, and a story.
Your Decision
Prices in this market have risen from 100 to 300. How do you assess its current phase?
What Happens
The choices map to four layers:
| Layer | Role | Observable proxies | Question |
|---|---|---|---|
| Story | Ignition and steering | Narrative concentration and project directions | Why outside money enters |
| Available money | Fuel | Stablecoin supply and changes | Available purchasing power |
| Borrowed money | Amplifier | Open interest, utilization, collateral totals | Forced selling after declines |
| Price | Output | Market quotes | Combined result of the first three |
Most watch the final row and use it to explain the others.
The main lesson: rallies and declines are opposite directions of the same mechanism.
| Link | Rising market | Falling market |
|---|---|---|
| Collateral price | Higher borrowing capacity | Lower borrowing capacity |
| Leverage | Expands and lifts prices | Forced closure lowers prices |
| Liquidity providers | Returns attract capital and deepen markets | Exits reduce depth and increase slippage |
| Story | Gains validate it and attract money | Losses discredit it and accelerate exits |
| Each step | Reinforces the next | Also reinforces the next |
The last row is crucial.
You have seen it twice:
- Chapter 13: algorithmic depegs trigger support-token issuance, lowering its price and weakening the peg further.
- Chapter 18: declines trigger liquidation, market selling, and further liquidation.
These are one structure appearing at three scales. Inside a protocol, liquidation cascades; across a market, cycles.
Build a Model
First, the rising loop:
- New narrative
- Outside inflows
- Rising prices
- More valuable collateral
- More borrowing capacity
- More purchases
- Rising prices
Then the falling loop, using the same links:
- Inflows stop
- Prices stop rising
- Leverage exits
- Prices fall
- Collateral shrinks
- Forced selling
- Prices fall
These explain three puzzles.
First, tops often lack bad news. The rising loop ends when new money cannot sustain it, not necessarily when something bad occurs.
Second, declines are faster. Rising markets require voluntary purchases; later declines rely on automatic liquidation, faster than human decisions.
Third, protagonists change but scripts persist. Stories must be new to attract capital; the underlying structure can repeat unnoticed.
Read this limitation carefully:
The model describes structure, not timing.
It measures current fragility, not when the system breaks. High leverage can persist; low leverage can still face outside shocks.
Timing claims demand something the model cannot deliver. Assign each observed number to its layer, rather than infer trading dates.
In summary:
Cycles are leverage phenomena, beyond fluctuations in psychology.
Emotion does not automatically sell a position at three in the afternoon. Liquidation contracts do.
What It Is Called
Here this means available purchasing power, rather than one pool's depth.
Distinguish market-wide money, often proxied by stablecoins, from tradability of one asset, covered in Chapter 15.
Both can deteriorate together and reinforce each other.
Obtaining purchasing power now against a promise.
On-chain credit is mainly collateralized: total credit moves with collateral prices. Rising prices expand it; falling prices contract it.
Traditional systems have banks and central banks that can act against the cycle. Their absence contributes to steeper on-chain cycles.
Borrowed funds enlarge positions.
It is a symmetric amplifier of gains and losses, but outcomes are asymmetric: gains require realization decisions; losses beyond a threshold are realized automatically.
Chapters 16 and 18 explain individual liquidation; this chapter covers aggregation.
A story explaining why something should have value.
It is a real variable, not an insult. Narratives steer external capital.
Beware circular validation: gains confirm belief, which creates more gains, without an independent test.
Beliefs change reality, and the changed reality reinforces beliefs.
Ordinary goods often see demand decline as prices rise. Reflexive markets can see rising prices create demand, including more collateral-supported purchasing power.
Price becomes an internal driver rather than a passive signal, letting the system propel itself upward and then reverse.
Hands-on
Produce an overlay and answer: which moved first?
Choose three curves.
Stablecoin supply is fuel; perpetual open interest or major-protocol borrowing is amplification; a major asset price is output.
Use the same sufficiently long interval containing at least two clear rises and declines.
Write definitions, more important than plotting.
Which stablecoins, chain balances, exchange accounting, and cross-chain deduplication are included?
Which venues and coin or dollar denomination define open interest? These can diverge during volatility.
Record source, definition, and retrieval date for each curve.
Normalize each series relative to its starting value, allowing different scales to share a chart.
Mark turning points using one written rule, such as retreating X% from a 30-day high.
Changing the rule may change ordering. Recognize that sensitivity.
Complete this table.
| Interval | Stablecoin turn | Open-interest turn | Price turn | Order |
|---|---|---|---|---|
| First rise | ||||
| First decline | ||||
| Second rise | ||||
| Second decline |
Write observations and three uncertainties.
Definitions, turning-point rules, and intervals may change the apparent sequence. This is the Lab's key discovery.
Two cycles cannot establish a timing system. Patterns from two samples lack statistical force.
The outputs are defined, verifiable curves replacing a vague word, and firsthand evidence that definitions affect conclusions. Chapter 23 deepens this.
AI Lab
Constrain the output:
Compare the past three Crypto cycles in a structural table:
1. Main narrative and emergence date
2. External inflow forms, observable indicators, and sources
3. Leverage accumulation products and indicators
4. Turning-point trigger
5. Step-by-step decline transmission
6. Biggest difference from the other cycles
Rules:
- Separate contemporaneously observable facts from hindsight-only knowledge.
- Identify evidence for every causal claim.
- Add a section on where the structure failed or fit least well.Then mark every causal word and challenge its evidence.
Three recurring model failures:
- Correlation presented as causation, ignoring third factors or reciprocal effects.
- Hindsight contamination, putting later-known facts into contemporary signals.
- Selective examples, excluding cycles that contradict the requested pattern.
Count each occurrence. Human cycle reviews make these same mistakes frequently too.
Verify these points yourself after using AI
- Can every date and figure be verified publicly?
- Does causal wording merely describe simultaneous events?
- Do three samples support its conclusions?
- Does it mistake hindsight-only information for observable contemporary signals?
- Do every-cycle claims include all cycles or only supporting examples?
- Does it provide counterexamples where the structure fails?
- Does it insert timing advice beyond the evidence?
Real-World Case
Chapter 13's algorithmic collapse, Chapter 18's cascades, and market cycles share one feedback structure.
Protocol: collateral falls, liquidation sells, and price falls further. Asset: depegs trigger issuance, weakening support assets and the peg. Market: prices contract credit, forcing sales and further declines.
Whenever each step strengthens the next, investigate the feedback loop immediately.
Several historical turns lacked remarkable negative events during the turning weeks.
Later explanations attach an event, while participants remember prices simply ceasing to rise.
Slower fresh inflows can stop a loop requiring continual funding. Leverage then completes the reversal.
Comparable declines often take far less time than rallies.
The execution difference is voluntary buying decisions versus automatic liquidation sales.
People decide one; code executes the other. Chapter 18's mechanism scales across markets.
New leading themes attract fresh capital; repeating an old story rarely does.
Different stories can tempt people to conclude the entire cycle differs.
Narratives change while structures recur. Distinguish those layers.
New themes can contain durable value and accumulated infrastructure. Recognizing cycles does not deny progress.
Change One Variable
Prices still respond to external flows, with less amplification and no liquidation-driven acceleration.
Declines no longer force sales, breaking the link from a 50% fall to another forced decline.
But market makers and arbitrageurs lose capital efficiency. Leverage supports both depth and fragility, echoing Chapter 15.
Credit no longer directly follows collateral prices, reducing automatic contraction.
But identity, credit records, and legal recovery introduce risks and requirements that conflict with anonymity and irreversibility.
These missing prerequisites explain why most on-chain credit stays overcollateralized, as Chapter 16 showed.
Acceleration and sudden stops disappear from inflows, moderating price curves.
Collateral borrowing still expands with prices, so fragility still accumulates, more slowly.
External-flow variation amplifies cycles rather than solely causing them. Separate it from leverage.
This chapter offers no such tool and doubts its existence, but consider the scenario.
Users act early, changing prices and triggering indicators sooner, which causes still earlier action.
Reflexivity changes the indicator's target. Widely used prediction methods alter the system they predict.
The model promises a clearer description of present fragility and its location, nothing beyond that.
Takeaway Questions
During rallies everyone seems to profit without a payer.
Over time, later holders fund a substantial part of earlier gains. This is accounting in the market's zero-sum component, not a conspiracy. Identify your position in that transfer.
Ask why liquidity providers remain.
Token-funded incentives shrink with prices, encouraging departure precisely when depth is needed. Chapter 15's slippage becomes especially consequential then.
High-leverage holders lose first through automatic liquidation, followed by LPs facing poor exits as depth shrinks.
Knowing your place in that sequence is more useful than predicting price.
This question helps assess survival through cycles.
Subsidies and volume make many projects look useful in rallies. With tokens down 90% and subsidies gone, does anyone still use the protocol? Chapters 26 and 30 make this test practical.
Chapter Check
Emotion does not automatically execute orders at thresholds.
Collateralized credit expands and contracts with prices and forces sales below liquidation levels.
No fresh human decision is needed along that automatic chain, making it observable and parameterized.
Emotion influences the narrative layer and outside flows. Both matter through different mechanisms.
All involve positive feedback at different scales.
Depeg creates issuance and weaker backing. Decline creates liquidation and further selling. Market decline contracts credit and forces selling.
Locate where output becomes input again. That enables acceleration in either direction.
Incoming funds commonly become stablecoins, roughly tracking purchasing-power flows.
It measures fuel independently of asset price.
Limits include inconsistent definitions, delayed market entry, payment and savings use unrelated to trading, and few complete cycle samples.
It is a coarse reading, not a clock, suitable for broad descriptions rather than precise timing.
Rallies need continuing fresh inflows, not merely the absence of bad news.
Slower inflows stop gains; holding costs encourage leveraged exits; selling triggers initial liquidations and automatic propagation.
The later-named trigger may simply be contemporaneous, the causal error examined in the AI Lab.
Check whether you can:
- Translate vague cycle labels into three verifiable indicator layers.
- Describe fragility rather than predict rupture.
- Identify output-to-input loops in new mechanisms.
- Avoid packaging the model as a timing system.
The fourth is essential. The model improves description of current conditions. Using descriptive structure as a forecasting tool is a costly mistake.
Key idea
Cycles are feedback loops in which liquidity, leverage, and narratives reinforce one another, not mysticism.