TIDE NOTEThe water is cool (heat 28), but the moon isn't in place (timing 0) — on the last three cycles' rhythm, the top window opens in about 29 months.
Heat × timing. The tide must be high and the calendar right before it crosses the line. Next top window 2029-08 – 2029-12
TIDE NOTEThe ebb-tide floor is on trial: price has floated 26% off this cycle's $58,525 low; hold for 38 more days and the “shallower bears in the ETF era” case gains another vote.
Triggered earlier this cycle from 2026-06-09, peaking at 59 (28 days ≥ 50). Coldness × bottom timing. Is this cycle's low $58,525 the cycle bottom? Verdict on 2026-12-19
15 years oftidesEveryebb
Left axis: scores (0–100); right axis: BTC price (log). The scales differ, so line crossings mean nothing. Dashed lines mark the 50 and 70 thresholds. Tap the legend to toggle lines; switching top/bottom swaps in the matching default lines.
Heat bycategory
Holder behavior has no free, republishable full history and isn’t scored; its 20% weight is spread proportionally across the other four, so the weights here are higher than the original 35%, 20%, 15% and 10%.
Cycleposition
About 35 months to the center of the halving-based window. If price makes a new cycle low, the low and its window shift later.
Coldness bygroup
Cycle structuretest
Are bear markets shallower in the ETF era? Is this cycle's low (2026-06-30, $58,525) the cycle bottom?
Testing: if price holds above the candidate low of $58,525 (2026-06-30) through 2026-12-19, that supports “this low is the cycle bottom and ETF-era bears are shallower”. A break below it fails the hypothesis. The candidate low was fixed when the hypothesis was set and does not move with new lows; once price exceeds the previous top, the result is locked.
| Bottom | From top | Drop | MVRV | Peak | ≥50 days |
|---|---|---|---|---|---|
| 2015-01 | 406 | -84% | 0.56 | 74 | 71 |
| 2018-12 | 364 | -84% | 0.69 | 93 | 129 |
| 2022-11 | 366 | -77% | 0.75 | 84 | 213 |
| 2026-06This cycle | 267 | -53% | 1.10 | 59 | 28 |
From top: days since the previous top; Peak: highest bottom signal that cycle; ≥50 days: days with bottom signal at or above 50.
Top timing 0
Two clocks, each 0–100, averaged. Purple dots are the four past tops; the band is the window (dark = full score, light = partial).
Days since last halving
○ 100 days past this window (next window dates below)
From the 2028-04-15 (est.) halving: full-score window 2029-08-03 ～ 2029-12-01; scoring range 2028-12-06 ～ 2030-07-29.
Days since cycle low
○ 869 days until the full-score window
From the current low ($58,525): full-score window 2029-03-29 ～ 2029-07-27. A new cycle low would shift the low and the window later.
Day counts at past tops
| Top | Since halving | Since cycle low |
|---|---|---|
| 2013-12 | 371 | 747 |
| 2017-12 | 525 | 1067 |
| 2021-11 | 546 | 1059 |
| 2025-10 | 534 | 1062 |
Window centers are the average of the last three tops (2017, 2021, 2025); full score within ±60 days, then linearly down to 0 over another 240 days. The shorter 2013 cycle is excluded.
Bottom timing 97
Two clocks, each 0–100, averaged. Magenta dots are the three past bottoms; the band is the window (dark = full score, light = partial).
Days since last top
● Inside the full-score window; leaves in 38 days
From the 2025-10-06 top: full-score window 2026-08-21 ～ 2026-12-19; scoring range 2025-12-24 ～ 2027-08-16.
Days since last halving
◐ 16 days past the full-score window; score fading
From the 2024-04-20 halving: full-score window 2026-06-28 ～ 2026-10-26; scoring range 2025-10-31 ～ 2027-06-23.
Day counts at past bottoms
| Bottom | Since prior top | Since halving |
|---|---|---|
| 2015-01 | 406 | 777 |
| 2018-12 | 364 | 889 |
| 2022-11 | 366 | 912 |
Window centers are the average of the three past bottoms (2015, 2018, 2022); full score within ±60 days, then linearly down to 0 over another 240 days. Mid-cycle crashes (2019-12, 2020-03, 2021-07, 2024-08) all had timing 0, so the bottom signal did not trigger.
Heat 28
Weighted average of 4 categories: On-chain valuation, Capital flows, Leverage & derivatives, Sentiment. The cycle method projects this cycle's expected top from past tops; percentile indicators only score above 35%; each indicator is capped at 120. Tap any indicator for details and history.
On-chain valuationWeight 35% · effective 44% 41
Capital flowsWeight 20% · effective 25% 0
Leverage & derivativesWeight 15% · effective 19% 53
SentimentWeight 10% · effective 12% 0
No free data: Google Trends “bitcoin” (no official free API; skipped in v1)
Holder behavior (long- and short-term holders’ cost basis, profit/loss and supply) has no free, republishable full history, so it isn’t scored. For these charts, see Market state › External references.
Coldness 33
Weighted average of three groups: valuation, miners and price structure. 100 means this cycle's projected bottom level; 0 means neutral or above. Tap any indicator for details and history.
ValuationWeight 50% 0
MinersWeight 30% 48
Price structureWeight 20% 95
Holder behavior (long- and short-term holders’ cost basis, profit/loss and supply) has no free, republishable full history, so it isn’t scored. For these charts, see Market state › External references.
Data
Daily values for the last 30 days, plus full downloads.
Top signals
| Date | Price | Signal | Heat | Timing | On-chain | Flows | Leverage | Sentiment |
|---|---|---|---|---|---|---|---|---|
| 11-11 | — | 0 | 28 | 0 | 41 | 0 | 53 | 0 |
| 11-10 | 73,661 | 0 | 28 | 0 | 41 | 0 | 53 | 0 |
| 11-09 | 75,161 | 0 | 29 | 0 | 41 | 0 | 56 | 0 |
| 11-08 | 74,968 | 0 | 29 | 0 | 41 | 0 | 61 | 0 |
| 11-07 | 75,724 | 0 | 30 | 0 | 43 | 0 | 61 | 0 |
| 11-06 | 75,827 | 0 | 31 | 0 | 44 | 0 | 62 | 0 |
| 11-05 | 75,803 | 0 | 31 | 0 | 43 | 0 | 66 | 0 |
| 11-04 | 80,488 | 0 | 35 | 0 | 54 | 0 | 59 | 0 |
| 11-03 | 77,656 | 0 | 34 | 0 | 48 | 0 | 69 | 0 |
| 11-02 | 76,681 | 0 | 31 | 0 | 43 | 0 | 67 | 0 |
| 11-01 | 78,135 | 0 | 34 | 0 | 51 | 0 | 62 | 0 |
| 10-31 | 77,401 | 0 | 34 | 0 | 49 | 0 | 62 | 4 |
| 10-30 | 77,110 | 0 | 34 | 0 | 48 | 0 | 61 | 11 |
| 10-29 | 78,193 | 0 | 35 | 0 | 48 | 0 | 60 | 20 |
| 10-28 | 81,237 | 0 | 35 | 0 | 55 | 1 | 38 | 27 |
| 10-27 | 80,801 | 0 | 35 | 0 | 59 | 1 | 29 | 31 |
| 10-26 | 79,950 | 0 | 36 | 0 | 59 | 1 | 33 | 33 |
| 10-25 | 80,039 | 0 | 35 | 0 | 57 | 0 | 32 | 35 |
| 10-24 | 80,699 | 0 | 32 | 0 | 55 | 0 | 19 | 34 |
| 10-23 | 80,924 | 0 | 36 | 0 | 64 | 1 | 15 | 36 |
| 10-22 | 81,257 | 0 | 32 | 0 | 51 | 1 | 20 | 41 |
| 10-21 | 84,176 | 0 | 37 | 0 | 63 | 9 | 14 | 36 |
| 10-20 | 82,899 | 0 | 41 | 0 | 65 | 16 | 21 | 35 |
| 10-19 | 81,918 | 0 | 36 | 0 | 59 | 17 | 8 | 36 |
| 10-18 | 81,935 | 0 | 35 | 0 | 61 | 9 | 7 | 40 |
| 10-17 | 81,269 | 0 | 33 | 0 | 53 | 13 | 7 | 43 |
| 10-16 | 81,538 | 0 | 35 | 0 | 57 | 15 | 6 | 47 |
| 10-15 | 82,116 | 0 | 36 | 0 | 60 | 11 | 6 | 45 |
| 10-14 | 81,499 | 0 | 37 | 0 | 61 | 9 | 7 | 50 |
| 10-13 | 83,141 | 0 | 38 | 0 | 65 | 10 | 5 | 53 |
The top row is today: the price appears after the daily close, so it shows “—” for now; scores already use whatever data is available today (sentiment and derivatives usually update a day ahead of price).
Bottom signals
| Date | Price | Bottom signal | Coldness | Bottom timing | Valuation | Miners | Price structure |
|---|---|---|---|---|---|---|---|
| 11-11 | — | 30 | 33 | 97 | 0 | 48 | 95 |
| 11-10 | 73,661 | 28 | 34 | 97 | 0 | 50 | 95 |
| 11-09 | 75,161 | 27 | 33 | 97 | 0 | 50 | 93 |
| 11-08 | 74,968 | 27 | 34 | 97 | 0 | 50 | 93 |
| 11-07 | 75,724 | 25 | 27 | 98 | 0 | 30 | 92 |
| 11-06 | 75,827 | 25 | 27 | 98 | 0 | 28 | 92 |
| 11-05 | 75,803 | 25 | 28 | 98 | 0 | 31 | 92 |
| 11-04 | 80,488 | 24 | 23 | 98 | 0 | 20 | 85 |
| 11-03 | 77,656 | 24 | 24 | 98 | 0 | 22 | 89 |
| 11-02 | 76,681 | 24 | 29 | 99 | 0 | 35 | 90 |
| 11-01 | 78,135 | 24 | 24 | 99 | 0 | 20 | 88 |
| 10-31 | 77,401 | 24 | 24 | 99 | 0 | 21 | 89 |
| 10-30 | 77,110 | 24 | 24 | 99 | 0 | 21 | 90 |
| 10-29 | 78,193 | 23 | 25 | 99 | 0 | 26 | 88 |
| 10-28 | 81,237 | 24 | 23 | 100 | 0 | 21 | 84 |
| 10-27 | 80,801 | 23 | 23 | 100 | 0 | 21 | 84 |
| 10-26 | 79,950 | 23 | 23 | 100 | 0 | 21 | 85 |
| 10-25 | 80,039 | 23 | 23 | 100 | 0 | 21 | 85 |
| 10-24 | 80,699 | 23 | 23 | 100 | 0 | 21 | 84 |
| 10-23 | 80,924 | 23 | 23 | 100 | 0 | 21 | 84 |
| 10-22 | 81,257 | 23 | 26 | 100 | 0 | 30 | 83 |
| 10-21 | 84,176 | 23 | 22 | 100 | 0 | 21 | 79 |
| 10-20 | 82,899 | 23 | 23 | 100 | 0 | 21 | 81 |
| 10-19 | 81,918 | 23 | 23 | 100 | 0 | 21 | 82 |
| 10-18 | 81,935 | 23 | 23 | 100 | 0 | 21 | 82 |
| 10-17 | 81,269 | 23 | 23 | 100 | 0 | 22 | 83 |
| 10-16 | 81,538 | 23 | 23 | 100 | 0 | 21 | 82 |
| 10-15 | 82,116 | 23 | 23 | 100 | 0 | 21 | 81 |
| 10-14 | 81,499 | 23 | 23 | 100 | 0 | 21 | 82 |
| 10-13 | 83,141 | 22 | 22 | 100 | 0 | 21 | 80 |
The top row is today: the price appears after the daily close, so it shows “—” for now; scores already use whatever data is available today (sentiment and derivatives usually update a day ahead of price).
Market state
Trend, bull-market dips and short-term overheating. Independent of the top/bottom signals and unaffected by the mode switch; the evidence is weaker than for those signals — reference only.
Price is above both the 200-day and 20-week moving averages.
Bull-market dip watch
Triggered 11 times since 2014: median 90-day return +13.8% (73% higher) vs +12.4% on ordinary bull-market days; price typically fell another −5.8% within 30 days. Recent: 2021-05, 2021-12, 2024-05, 2025-02
Short-term overheating
Overheating is not a reliable sell signal: what followed was mixed, and a big drop was only slightly more likely than on an ordinary day (see the stats below). Top timing is 0; only when timing is also inside the top window can overheating mark a secondary top.
Triggered 16 times since 2014: median 90-day return +4.5% (ordinary bull days +12.4%); a 15%+ drop within 45 days followed 38% of the time (ordinary 32%). Recent: 2021-10, 2023-01, 2024-02, 2024-11
Price & trend lines
Built from an October 2026 indicator study: 48 indicators, observed on 2014–2019 and tested on 2020–2026. Single overheating readings (RSI, distance from moving averages, greed) were on average followed by further gains, and combined overheating had mixed outcomes — neither is a reliable sell signal; the bull-market dip watch shows no reliable difference from ordinary bull-market days (after 2020 its triggers did worse; the earlier edge came mostly from two 2017 cases) — a state reference only. The weekly Supertrend (ATR 10 × 3, from Bitstamp daily candles, completed weeks only) flips about once a year with few whipsaws, but it confirms rather than predicts — by the time it turns down, price is usually weeks past the high. Thresholds are conventional values, not fitted to history. For reference only — not investment advice.
External references: worth a look, not scored
These datasets have no free, republishable full history, so we don’t score them and can’t validate them against past tops and bottoms like our other indicators. Links open the original sites in a new tab; the notes are ours, unaffiliated with those sites, and we don’t reproduce their charts or figures.
Holder behavior
Cost basis and profit/loss of long- and short-term holders. There is no free, republishable full history (free APIs cover about 4 years at most), so they aren’t scored.
STH cost basis · STH-MVRV
The average purchase price of short-term holders (coins held under 155 days). Above it, recent buyers are in profit and bull-market pullbacks often stall nearby; below it, they are underwater. STH-MVRV = price ÷ this cost basis, and STH-NUPL = 1 − 1 ÷ STH-MVRV (below 0 means below cost).
Caveat“Below cost in a bull market = a good entry” only works if you already know it is a bull market; in bear markets price stays below this line for months.
STH cost basis vs True Market Mean (cross)
The True Market Mean (TMMP) is the average cost of coins still in active circulation. When the short-term holder cost basis crosses above it, this is often read as “bull market confirmed”: recent buyers are paying above the market average; a cross below is read as a confirmed bear market.
CaveatWe tested it in October 2026 with CheckOnChain’s history: the last three crosses above came 190–490 days after the cycle bottom, with price already 135–239% above it — always later than the weekly Supertrend turning up; the ratio correlates 0.85 with MVRV Z-score; it whipsawed in May 2016 and April 2020; there are only 5 events. A confirmation signal, not an entry signal.
NUPL by cohort
Splits the market-wide NUPL used in our coldness score into long- and short-term holders. Long-term holders’ paper profit peaks near cycle tops and shrinks to its smallest near bear-market bottoms.
CaveatHighly correlated with the NUPL and MVRV we already score — not independent evidence.
Long- / short-term holder supply
Coins held by long-term holders fall in the late bull market (old coins sold to new buyers) and rise during bear markets (re-accumulation).
CaveatTurns usually become clear only months later, and the 155-day cutoff shapes the curve.
LTH-SOPR
The profit multiple at which long-term holders sell. It jumps when they take heavy profits near tops; dropping below 1 late in a bear market (long-term holders selling at a loss) has tended to come near bottoms.
CaveatWe tried it with about 4 years of free data; adding it made the 2025 top signal worse, so it was not adopted.
Long-cycle valuation
They need the full coin-age distribution: free sites show the charts, but the data cannot be republished.
Reserve Risk
Price relative to how unwilling long-term holders are to sell. Low readings mean a low price with confident holders (historically early in the cycle); high readings have come near cycle tops.
CaveatPeak readings fall each cycle, so — like our classic indicators — a fixed red line doesn’t work.
RHODL Ratio
Compares the realized value of coins moved within a week with coins held 1–2 years. It rises when new money floods in, near cycle tops.
CaveatAlso declines cycle to cycle and correlates strongly with MVRV-type indicators.
Capital momentum
We can compute this from data we already collect, and we tested it.
SSR Oscillator
BTC market cap ÷ total stablecoin supply, expressed in standard deviations from its 200-day mean. Often read as how much stablecoin “dry powder” is available.
CaveatWe tested the same formula in October 2026 (from 2019): it correlates 0.77 with price’s own 200-day deviation, so it mostly reflects momentum; the 2021 and 2025 cycle tops read only 0.9 and 0.2 while the March 2024 mid-cycle high read 2.8; after readings ≥ 2 the median 90-day return was still +26%. Not a top or bottom signal.
Short-term structure
Reflects leverage and hedging over days to weeks — unrelated to the cycle position.
Liquidation heatmap
Estimates the price levels where leveraged positions would be force-closed. Price often swings hard around dense clusters, sometimes sweeping through them before reversing.
CaveatA model estimate, not actual positions published by exchanges.
Options open interest & put/call
Shows where the market has bought protection (puts) or upside bets (calls), and when large expiries land.
CaveatPositions may be hedges rather than directional bets, so they don’t map directly to sentiment.
Cycle strategy
“Alert + trend confirmation”: the top/bottom signals locate the cycle and the weekly Supertrend confirms the trend. Walk-forward backtest since 2014: 736× vs 97.9× for holding.
Next: enter alert (top signal ≥ 50 or top timing ≥ 50). By the cycle clocks the earliest is about 2029-02-01 (assuming this cycle's low holds). Once on alert, it sells only when the weekly Supertrend turns down.
Early-top version (comparison): from 650 days after the cycle low (2028-04-10), any new all-time high within 180 days puts it on alert without waiting for timing. Now 133 days since the low — not active yet.
Rules
- AlertWhile holding, top signal ≥ 50 or top timing ≥ 50 → on alert (stays until a sale).
- SellOn alert, the weekly Supertrend (ATR 10 × 3) turns down → sell everything.
- Buy backIn cash, bottom signal ≥ 50 or the weekly Supertrend turns up → buy everything back.
Comparison: early-top version (added 2026-10-15; does not replace the main strategy): one extra alert condition — at least 650 days since the cycle low and a new all-time high within the last 180 days (no heat, no clocks). It covers the blind spot of a top arriving a year or more before the clocks expect. In walk-forward backtests its trades are identical to the main strategy; with the clocks shifted a year late to simulate early tops, the main strategy rides the 2014 and 2022 bears while this version sells after each top (plus one extra mistaken sale in Nov 2019). The 650 days were not picked from data: any threshold from 300 to 800 days gives identical historical results (the new-high condition does the work), so it is a judgment call — shorter than the shortest low-to-top ever (747 days in 2013) to cover a short cycle, longer than the 490 days at the pre-halving new high of March 2024 to exclude too-early highs; for this cycle it falls on 2028-04-10, around the 2028 halving.
The top signal and timing light up early, so they only put the strategy on alert; the actual sale waits for the weekly trend to confirm. Top signal = heat × timing ÷ 100 never exceeds timing, so the alert condition is effectively “top timing ≥ 50” and heat itself does not affect trades. Thresholds are fixed at 50, not fitted to history; all 27 combinations of 40/50/60 beat holding, but they produce only 3 distinct position paths — not 27 independent tests.
Since finalization on 2026-10-06 (day 35)
The rules were finalized on 2026-10-06. Everything before is a backtest (the rules were chosen with history in view); everything after is unseen data — the real test. The scenario paths were frozen at the same time to compare with what actually happens.
No trades since finalization; scenario paths frozen on 2026-10-05
| Scenario | Expected today | Actual | Gap | Top |
|---|---|---|---|---|
| Conservative Closest | $86,905 | $73,661 | −15.2% | $110k |
| Base | $89,138 | $73,661 | −17.4% | $184k |
| Optimistic | $91,721 | $73,661 | −19.7% | $327k |
Simulated equity
Log scale; start = 1. Purple bands mark alert periods (alert → sale); the faint dashed line is the batch version and the blue dots are the early-top version: every historical trade matched the strategy, so the dots sit on the strategy line; they will separate only if the two diverge.
Three scenarios for the next top
Price vs scenarios: today's price is 15.2% below the conservative path (37 days running) (vs the paths frozen on 2026-10-05). A warning appears only after 30 straight days more than 20% outside the range — 20% is roughly one month of typical bitcoin volatility.
How the estimates moved: since 2026-10-05, Conservative +0.0%, Base −0.5%, Optimistic −1.0%。No notable change since the snapshot (counted when any top estimate drifts ≥ 3% or the cycle low or top date changes). Each day's estimates go into an append-only record (data/scenario_history.csv), so changes can't be quietly rewritten.
- ConservativeThe top-to-prior-low multiple keeps shrinking (111.82 → 21.21 → 7.92); extrapolated to about 1.88× × this cycle's $58,525 low.
- BaseTop MVRV (4.72 → 4.43 → 2.85 → 2.29) extrapolated to about 1.78 × the realized price (now $53,926, grown at the last 4 years' 27% a year to the top date).
- OptimisticThe power-law trend price (fitted on all history, top multiple 1.0). If the top multiple keeps falling along 11 → 6.6 → 2.9 → 1.2, the result lands near “Base”.
- PullbacksThe paths are smoothed to only rise, but on the way from low to top the last three cycles saw maximum pullbacks of -36%, -61%, -28% (2017-06 → 2017-07, 2019-06 → 2020-03, 2024-12 → 2025-04), and none broke below the starting low. The shaded band under the base path marks that range: even on track, price could first fall 30–60%.
- Not drawnTop ÷ previous top (17.31 → 3.44 → 1.85) extrapolates to about $64k — below the last top, i.e. no new high next cycle; listed for reference.
Dashed lines are the paths frozen on 2026-10-05; they are not redrawn from each day's price, so you can see how far actual prices drift. The dots at the right are each day's recomputed top estimates, which move only when structural inputs such as the cycle low or realized price change. Path shape is the average of the last three cycles' low-to-top paths (smoothed and forced to only rise — real prices will have big pullbacks); the top date is the midpoint of the two timing clocks' centers. Each method rests on only 3–4 data points and assumes diminishing returns continue; high and low differ by about 3× — that gap is the uncertainty. Recomputed daily; not a forecast or a price target.
Performance by start year
| Start | Strategy | Hold | Strat / hold | Batch / hold | Early-top / hold | Strat CAGR | Hold CAGR | Strat max DD | Hold max DD |
|---|---|---|---|---|---|---|---|---|---|
| 2014 | 736× | 97.9× | 7.5× | 5.0× | 7.5× | 67% | 43% | −74% | −84% |
| 2016 | 552× | 169× | 3.3× | 2.1× | 3.3× | 79% | 60% | −74% | −84% |
| 2018 | 11.3× | 5.5× | 2.1× | 2.5× | 2.1× | 31% | 21% | −81% | −81% |
| 2020 | 21.1× | 10.3× | 2.1× | 2.5× | 2.1× | 56% | 40% | −66% | −77% |
| 2022 | 2.3× | 1.6× | 1.5× | 1.6× | 1.5× | 19% | 9% | −67% | −67% |
Growth of 1 unit invested on January 1 of each year to today. Each start begins fresh — fully invested and not on alert (no alert state carried over from before the start) — so 2018 and 2022 come out lower than the same stretch of the continuous run from 2014. 0.1% fee per trade; cash earns nothing. Batch version: sells 1/3 on alert and the rest when the weekly Supertrend turns down; otherwise identical (for comparison). Early-top version: see the rules (for comparison).
Past trades
| Date · trigger | Action | Price | Outcome |
|---|---|---|---|
| 2014-03-03Weekly ST down | Sell | $677 | rebought 57% lower |
| 2015-07-13Weekly ST up | Buy | $293 | held +2,241% |
| 2018-02-05Weekly ST down | Sell | $6,850 | rebought 37% lower |
| 2018-11-20Bottom signal ≥ 50 | Buy | $4,318 | held +795% |
| 2021-05-24Weekly ST down | Sell | $38,630 | rebought 48% higher |
| 2021-10-11Weekly ST up | Buy | $57,289 | held −27% |
| 2022-01-10Weekly ST down | Sell | $41,817 | rebought 51% lower |
| 2022-06-16Bottom signal ≥ 50 | Buy | $20,310 | held +353% |
| 2025-11-17Weekly ST down | Sell | $91,911 | rebought 33% lower |
| 2026-06-09 LatestBottom signal ≥ 50 | Buy | $61,640 | holding |
ST = Supertrend. The highlighted row is a bad sale (the 2021-05 mid-cycle dip was misread; it bought back higher).
Limitations
- It does not sell the top: it waits for trend confirmation, and past sales came 26–65% below the peak. The edge comes from sitting out bear markets and buying back near bottoms.
- It makes mistakes (a bad sale in 2021-05). There is no stop-loss, so drawdowns of −70% to −80% can still happen while holding (from a 2018 start, the max drawdown equals holding’s).
- Small sample: only four bear markets to test; if the four-year cycle breaks, the strategy may fail.
- The edge is concentrated early: 7.5× holding from 2014, but only 2.1× from 2018, 2.1× from 2020 and 1.5× from 2022.
- An early top in 2028 is a blind spot: the alert effectively depends only on top timing, which the clocks put no earlier than 2029-02-01. If the top comes in 2028, the strategy won’t go on alert and will ride the whole decline. Historically, “heat ≥ 75, timing < 30, 150–450 days after the halving, ≥ 650 days after the low” occurred in only 4 stretches: Nov 2013 was the cycle top, but after Jan 2021, Feb–Mar 2021 and Nov–Dec 2024 price rose a further ~99%, ~80% and ~38% (Nov 2024 was ~200 days after the halving, right after the election, with heat at 87) — so heat alone can’t flag an early top (tested Oct 2026).
- So an “early-top version” (see rules) is tracked separately from 2026-10-15 in its own append-only record (data/strategy_ledger_early.csv). Historically it never differed from the main strategy; differences can only appear after 650 days from this cycle’s low (Apr 2028) while the clocks have not arrived — and a sharp mid-bull drop in that window could shake it out.
- Walk-forward: each cycle uses only the tops and bottoms known at the time; the latest cycle uses the site's current signals. But top dates are assigned in hindsight (no one could confirm a top on the day), and data revisions aren’t versioned, so “no look-ahead at all” is not yet proven.
- How tops are confirmed matters (tested Oct 2026): if a top only counts after a 35% drop or a weekly Supertrend turn, the trades from the continuous 2014 run are identical, while the 2022 start goes from about 1.5× to about 3.1× holding (max drawdown −67% → −35%). But in real time, May 2021 also fell more than 50% from its high, and the same rule would have mistaken it for the cycle top — so this still relies on knowing afterwards which peak was the top.
- From the finalization date (2026-10-06), each day’s decision is appended to a record that is never rewritten (data/strategy_ledger.csv); live tracking uses that record.
- Simulated results, not actual trading. For reference only — not investment advice.