Predictive Frameworks (and Their Failures)

A critical look at the price-prediction models popular in crypto: Stock-to-Flow, Pi Cycle Top, MVRV Z-Score, Mayer Multiple, Wyckoff, log regression bands. What worked, what failed, and what survives honest scrutiny.

22 min · advanced · part of Understanding Crypto Markets

The Allure and Danger of Prediction

Crypto attracts predictions. Every cycle produces models claiming to forecast the next price level, the timing of the next top, the depth of the next bottom. Some of these models build large followings, sell newsletters and courses, and become embedded in the way traders talk about markets. Most of them fail. Not by a little — by a lot. The ones that worked once usually fail the next time. The ones that survive multiple cycles often have so many parameters that they explain the past beautifully but predict the future poorly. And the human tendency to remember the hits and forget the misses creates a survivorship bias that makes prediction look more reliable than it is. This lesson is not a takedown of every framework. Some indicators have genuine signal value as cohort-behavior tools. The MVRV Z-Score, when interpreted carefully, has tracked tops and bottoms reasonably well across cycles. The Mayer Multiple has historical statistical regularities. But none of these are forecasting tools in any reliable sense. They are descriptive — telling you where the market is relative to historical norms — not predictive. The honest framing is: predictions about specific prices and dates do not work. Understanding context — where we are in a cycle, what historical analogs exist, what the cohort behavior suggests — sometimes does. Lesson m1 covered halving cycle theory; lesson m2 covered on-chain analytics. This lesson is about the popular price-prediction models, what they actually got right and wrong, and how to think about predictive frameworks generally.

Also in this lesson

Stock-to-Flow: The Most Famous Failure
Pi Cycle Top, Mayer Multiple, MVRV Z-Score
Realized Cap, Realized Price, and Cohort Models
Why Technical Analysis Often Fails in Crypto
What Survives Honest Scrutiny
For Deeper Reading

Key terms

Stock-to-Flow (S2F): Price model published by PlanB March 2019 using stock/flow ratio. Predicted $98K Nov 2021 (actual: $57K) and $100K+ throughout 2022 (actual: $20K bottom). Effectively discredited by 2022. Textbook example of overfitting.
Pi Cycle Top Indicator: Created by Philip Swift 2019. Triggers when 111-day MA crosses 2x the 350-day MA. Accurate at the 2017 top (within 3 days), accurate at April 2021 local top, but did NOT trigger at the actual November 2021 cycle top.
MVRV Z-Score: MVRV normalized by historical standard deviation. Readings below 0 = bear-market lows; readings above 7 = cycle tops historically. The 2025 cycle peaked at ~4.5, below the historical 7-threshold (ETF participants changed the holder base).
Mayer Multiple: Created by Trace Mayer ~2017. Bitcoin price / 200-day moving average. <1.0 = buy zone, 1.0-2.4 = fair value, >2.4 = overheated. Descriptive (where market is) rather than predictive.
Realized Cap: Sum of every UTXO valued at the price when it last moved. Reflects holder cost basis. Less volatile than market cap. As of early 2026: ~$700-800B for Bitcoin vs ~$1.2T market cap. Introduced by Coinmetrics 2018.
Realized Price: Realized cap divided by total supply. Average cost basis of all holders. ~$35,000 for Bitcoin in early 2026. Long-term-holder realized price: ~$26K; short-term-holder realized price: ~$77K.
Wyckoff schematic: Early-20th-century framework by Richard Wyckoff identifying Phase A-E patterns of accumulation/distribution. Frequently overlaid on Bitcoin charts. Useful as descriptive heuristic; over-applied as predictive tool with poor track record.
Logarithmic regression bands: Bitcoin price plotted on log scale with regression curve and standard-deviation bands. Tops and bottoms historically near upper/lower bands. Descriptive rather than predictive.
Survivorship bias: Tendency to remember winning trades and forget losing ones. In TA case studies, examples where the technique worked are featured; counter-examples omitted. Across the population of TA practitioners, average return after fees is ~0 or negative.
Reflexivity: Self-referential dynamic where actions based on a prediction cause the prediction to materialize. When a TA indicator becomes popular, its triggers become self-fulfilling — and then self-defeating as everyone front-runs.
This time is different: Phrase warned against in finance, but sometimes genuinely true. Each Bitcoin cycle has had different participant composition; structural changes (e.g., spot ETFs in 2024) genuinely change market dynamics, reducing reliability of historical-fit patterns.
Long-term holder (LTH) accumulation: Rising LTH-cohort supply during a bear market. Has been a leading indicator of cycle bottoms. The signal is noisy and lagged but has been more reliable than price-based TA across multiple cycles.

Continue this lesson — 6 more sections in the CryptoBipto app.

Open lesson

Educational only — not financial advice.