The 9 Indicators Behind the Cycle Thermometer: Logic, Weights & Track Record

Quick answer: The nine indicators in the BTC Cycle Thermometer fall into four information groups — independent sources 35% (mining cost 20% + Fear & Greed 15%), momentum 23% (weekly MACD 15% + weekly RSI 8%), trend 26% (Mayer 10% + 200W SMA multiple 10% + power law 6%), position 16% (ATH drawdown 8% + Pi Cycle 8%). Correlated indicators share group weight instead of each taking a full slice, so no signal family gets double-counted.

Trend group: distance from long-term value

Mayer (price ÷ 200D SMA) under 0.8 marked every bear bottom, above 2.4 is the classic overheat line; the 200-week SMA is the slower anchor every bear low touched or briefly broke; power-law regression fits the multi-year trajectory (price = c × days^5.82) and reads current deviation. All three answer the same question, hence 26% combined rather than three full slices.

Momentum group: weekly force and turns

Weekly RSI is the early-warning type — it hit 90+ in April 2021 and printed the classic divergence before the top. Weekly MACD is the confirmation type — normalized as DIF as a % of price (comparable across cycles), its high cross-downs and deep cross-ups confirmed every major turn: roughly +15~25% at the 2021 top zone, -15~-20% at the 2022 bottom. One warns early, one confirms reliably; MACD as the confirmation signal takes 15%, RSI 8% — 23% together.

Independent sources: the differentiators

Mining cost is the only non-price input: estimated from the latest-gen miner's marginal cost. Price below it means even the most efficient miners lose money (seen at Mar-2020 and Nov-2022) — the hardest bottom anchor, hence the largest single weight at 20%. Fear & Greed blends volatility, momentum and social data — an independent sentiment axis.

Position group: half-strong helpers

ATH drawdown is powerful at bottoms (historic lows sat at -77%~-84%) but says little at tops; Pi Cycle nailed the 2013/2017/2021 tops but has low resolution at bottoms. Both are one-sided, hence moderate weights.

Each raw value maps to a 0-100 sub-score via piecewise-linear anchors set from historical behaviour, then weight-averaged; failed sources degrade gracefully with re-normalization. The full anchor table is public in the tool's Methodology section. Any single indicator misfires — but the moments when nine signal families hit extremes together have been, historically, the cycle's tops and bottoms.