Context and references.
The five combinations on this page β funding plus RSI plus MVRV, open interest plus funding plus liquidation clusters, and three others β are grouped so that each one reads three dimensions that do not move together. What follows for each is the same three things: the conditions that have to line up before the signal counts, the window over which it is worth watching, and the situations in which it is known to fail. There are no scores and no percentages here, deliberately. Any hit rate you could put in this table depends on the test window, the asset, the parameters and the fee assumptions; change one and the number changes.
The mechanism a combination actually provides is a veto. One indicator reading an overheated market can be overruled by a second indicator reading a dimension the first one cannot see β and that veto is the whole point. What it costs is signal frequency: three conditions lining up at once is far rarer than any one of them firing. That is not a defect. A combination is a filter, not an amplifier.
A practical caveat: combinations behave differently in trending and ranging markets. Through the long chop from May 2022 into January 2023, three-indicator setups fired rarely and resolved ambiguously, which is the honest way to describe it β nobody publishes a denominator for how often they were right. Applying a regime filter (realized volatility, or a simple trend/range classification) before acting on any combination signal is the standard mitigation.
SEC and CFTC compliance considerations: US-domiciled readers should verify that any leveraged strategy these combinations might suggest is consistent with their broker's margin rules. CME futures provide a regulated alternative for the same exposure with cleaner reporting.
The five combinations explained.
Combination 1, overheating: funding sustained well above its quiet-market level, daily RSI above 75, and a stretched valuation metric such as MVRV Z-Score. Signal condition: all three at once, not two out of three. Observation window: days to a couple of weeks. Checked against Binance data, the configuration was clearly in place before the March 2024 top (daily RSI(14) above 80 from 27 February to 4 March, BTCUSDT funding peaking at 0.0881% on 5 March). Around the November 2021 top it was only partial: daily RSI's last print above 75 was 78.2 on 20 October, and it was 59.3 on the day of the 10 November high, although funding averaged about 0.035% from mid-October. It did not line up in April 2021, when daily RSI peaked around 67.5 and only funding ran hot, and December 2017 predates Binance perpetual funding data, so it is not checked here. Note also what this does not say β these are occasions someone chose to look at, and nobody publishes the count of times it lined up and nothing happened. Known limitation: it describes how stretched the market is, never when that resolves; stretched can get more stretched for weeks. Rather than a fixed percentile threshold, compare the current funding reading against the trailing range of the contract you actually trade.
Combination 2, deleveraging pressure: open interest at the top of its own range, funding sustained high, and a liquidation cluster within a few percent of spot. Signal condition: all three, with the cluster close enough that a normal daily range can reach it. Observation window: hours to days. Before the May 2021 flush, BTCUSDT funding averaged about 0.039% over 1-18 May. Before the August 2024 yen-carry unwind the setup was not there: funding sat at +0.004% to +0.01%, far from "sustained high". Known limitation: a dense cluster explains where the fuel sits, not whether it gets lit β open interest can keep climbing, and the cluster can sit untouched for a week.
Combination 3, accumulation window: bullish RSI divergence, a MACD cross, and funding that has been negative for at least a full day. Worth being precise about that third leg, because the direction gets remembered backwards more often than not: negative funding means shorts are paying longs, so a sustained negative print is shorts paying to stay short β crowding on the bearish side, not the bullish one. Signal condition: all three, with funding negative across several consecutive settlements rather than one print. Observation window: one to four weeks. Pieces of it appeared around the November 2022 post-FTX low β daily RSI made a higher low into the 21 November price low (about 31.5 against 24.5 on 9 November), MACD crossed up on 24 November, and BTCUSDT funding had a run of five straight negative settlements from 14 November β though not all inside one 48-hour window. Around the August 2024 low, funding only turned negative for a full day on 10 August, five days after the bottom. Known limitation: funding can stay negative for weeks while price grinds lower, so this marks a zone, not an entry price.
Why a second dimension beats a second indicator.
A single indicator in crypto is noisy for a structural reason: it measures one thing. RSI measures price momentum, and price momentum comes from leverage, from spot demand, from macro β none of which RSI can see. So an RSI of 75 is consistent with retail chasing a top and equally consistent with sustained institutional buying, and the number alone cannot separate them. Adding a second momentum indicator does not help, because it is reading the same input. Adding a dimension the first one is blind to does, because it can contradict the first one. That is the entire argument, and it is worth noticing that the argument is mechanical rather than statistical. Stacking a fourth and fifth layer keeps adding veto power in theory, but in practice the conditions stop lining up at all, which is why nothing on this page uses more than three.
Implementation discipline.
The practical implementation rule: all three indicators must fire within the same 24-48 hour window. Stale signals β one fired three days ago, two fired yesterday β are not the same configuration as three firing together, because by the time the third arrives the first may already have resolved. Regime filtering helps as well. ADX above 25 (trending regime) and ADX below 20 (ranging regime) flip the effectiveness of momentum-versus-mean-reversion combinations. In trending regimes, momentum-confirming combinations beat mean-reverting ones; in ranging regimes the reverse holds. A simple trend/range classification you compute yourself is enough for this filter.
Common failure modes.
Three failure modes recur. First, single-asset application (a combination that works on BTC may not work on SOL due to liquidity and leverage profile differences). Second, regime drift (a combination calibrated for 2021 conditions may not work in 2024 ETF-era conditions). Third, look-ahead bias in backtesting (using closed-bar data without intrabar slippage assumptions). The mitigation pattern is to test each combination on a held-out validation window of at least 12 months, run forward-paper-trade for 4-8 weeks before committing capital, and recalibrate quarterly as market structure evolves. SoSoValue and Farside Investors provide the ETF-era inputs that make modern recalibration possible.
Combination 4: BTC.D rotation + ETH/BTC breakout + stablecoin issuance.
Combination 4 is the altseason-confirmation stack. Each layer reads a different market structure dimension. BTC.D rotating down from a 60%+ base confirms capital is leaving BTC; ETH/BTC pushing above 0.060 with momentum confirms ETH is leading the rotation; stablecoin supply expanding meaningfully over the trailing 30 days confirms dry powder is being deployed. Signal condition: all three, sustained over weeks rather than a single day's reading. Observation window: one to three months β this is a position-trade signal, not a scalp. The clearest example is the spring 2021 rotation, when BTC dominance fell sharply between February and May while ETH/BTC (Binance) went from a 0.030-0.046 range in February to a high of 0.0824 on 15 May, and USDT plus USDC supply grew from about $32 billion on 1 February to about $82 billion on 31 May (DefiLlama). Dominance figures differ between dashboards, so none is quoted here. Known limitation: one clean example is an illustration, not a base rate. Dominance can fall because BTC is falling faster than everything else, which is a very different market from the one this stack is meant to catch.
Combination 5: Coinbase Premium + ETF flow + Glassnode LTH supply.
Combination 5 is the institutional-positioning stack. Coinbase Premium tracks US spot demand; ETF flow data from Farside Investors tracks aggregate institutional flow; Glassnode's long-term holder (LTH) supply share tracks structural conviction. When the premium holds positive, ETF flow holds positive, and long-term holder supply is not being distributed, the institutional layer is leaning towards accumulation. Signal condition: all three sustained over weeks; single-day prints on any of them mean nothing. Observation window: months. The first half of 2024 is the usual reference period, and it is less clean than it is usually remembered: Farside's daily totals were net positive on 77 of 122 trading days (about $14.5 billion net over the half-year, with outflow stretches in April-May and June), while the Coinbase premium measured against Binance BTCUSDT at the daily close was positive on only 93 of 182 days, with a median near zero (that comparison also carries any USDT/USD basis). Known limitation, and it is a big one: ETF creations and redemptions now route around the Coinbase spot book that the premium measures, so the premium reads a smaller share of institutional demand than it did before 2024.
Position sizing across the five combinations.
Each combination points in a direction, and the direction is all it gives you. Combination 1 (overheating) is a reason to stop adding and to start trimming, not a reason to flip short. Combination 2 (deleveraging pressure) argues for a sharper reduction than the first, because the fuel and the ignition point are both identifiable. Combination 3 (accumulation window) is a low-conviction entry β small, staged, in a zone rather than at a price. Combination 4 is a rotation argument, shifting part of an existing crypto allocation rather than adding new capital. Combination 5 is structural confirmation, which is a reason to hold through noise rather than a reason to add. What none of them tell you is how much. That part depends on your own estimate of odds and payoff, on the size of your capital base and on what a bad month does to the rest of your life β and it is why this page does not attach a percentage to any of the five. Sizing that comes from a number we made up is worse than sizing you thought about yourself.
Testing a combination on your own data.
Since this page gives you conditions rather than scores, the obvious next question is how the conditions behave on the market you actually trade. The method is unglamorous: collect daily indicator readings from public sources (Coinglass for derivatives, a free on-chain tier, alternative.me for sentiment); write each combination's trigger conditions down precisely enough that you cannot fudge them afterwards; step through the historical sample; record what happened over the following N days. Whatever you get out of that belongs to your window, your asset and your fee tier, which is exactly why it is worth more to you than a number from us. Two things distort the result if you let them. Look-ahead bias is the common one β using closing-bar data as if you could have acted at that price flatters every result. The fix is to assume realistic slippage per side and a volume-weighted fill rather than a mid-price fill.
A note on validation against ETF-era data.
All five combinations described here read thresholds that were set on pre-2024 market structure. The ETF era introduces changes that prior cycle data does not capture, and the change shows up as a downward shift in the reading required to mark an extreme: valuation metrics that only flagged cycle tops at high readings in 2017-2021 were already rolling over at visibly lower readings in 2024. We are deliberately not quoting a replacement threshold, because one cycle is one observation and the next one will move it again. Combinations that include ETF flow as a layer (Combination 5) track the new regime more closely than combinations built purely on legacy on-chain metrics. For combinations that rely on legacy metrics (Combinations 1-3), the recommended adjustment is to use percentile-based thresholds rather than absolute thresholds. A "MVRV in top decile of trailing 365 days" rule generalizes better across regime changes than a "MVRV above 3.5" rule. The same principle applies to RSI, funding rate and on-chain metrics across the indicator stack. One final implementation note: combinations should be evaluated on a rolling forward-looking basis rather than once and frozen. Market regime drift is real β a combination calibrated against 2021-2022 data may underperform meaningfully against 2024-2025 data because the ETF era introduced new flow patterns. Recalibrating periodically on the trailing 12 months of data keeps the thresholds from drifting away from current conditions.
Crypto assets are volatile and not suitable for every investor. This page is editorial analysis, not financial advice.