Gate Research Institute: ETF Outflows Suppress Risk Appetite, Enabling Bidirectional Systemic Traversal of Weak Markets

Null Summary

• In May, the crypto market shifted from a surge early in the month to a pullback in mid-month, followed by low-volatility consolidation near month-end. BTC, ETH, and SOL all formed phase highs in the first half of the month before entering a correction phase. Mainstream ETFs weakened in their ability to absorb flows, while the share of perpetual contract trading volume stayed high, resulting in a market structure characterized by weaker spot performance and leverage-driven dominance.

• The dual-moving-average cluster breakout strategy performed best. The equal-weight buy-and-hold return for the three assets was about -6.09%; the long-only strategy’s return was about -3.65%; and the dual-direction strategy’s return was about +2.11%. Returns mainly came from the short trend legs of ETH and SOL, validating that the May market was more suitable for dual-direction trend trading.

• Under low-volatility compression, disciplined trading outperformed subjective judgment. The EMA12 exit mechanism effectively controlled losses from false breakouts, and 3R take-profit preserved trend gains. The current market is still in a phase of directional selection. A trading framework that identifies regime states, controls risk, and executes dual-direction signals is superior to a subjective strategy of chasing rallies.

The main contradiction in the May crypto market was that after prices surged at the start of the month, the spot absorption and leverage-trading structure diverged. BTC, ETH, and SOL all formed phase highs in early May, and then moved into retracements and low-volatility consolidation. BTC fell from a 4H close of 77,117.4 USD at the start of the month to 73,684.0 USD at month-end, with a monthly return of -4.45%. ETH fell from 2,283.02 USD to 2,007.0 USD, with a monthly return of -12.09%. SOL fell from 83.90 USD to 82.44 USD, with a monthly return of -1.74%. SOL’s closing decline was relatively smaller; its intramonth high touched 98.40 USD, then retraced to around 80.00 USD. Actual trading volatility was significantly higher than what the monthly gain/loss percentages alone indicate.

The strategy backtest results for the month are clear. Equal-weight buy-and-hold for the three assets produced about -6.09%; the long-only moving-average cluster breakout strategy produced about -3.65%; and the dual-direction moving-average cluster breakout strategy produced +2.11%. The dual-direction strategy achieved about +8.2% in excess returns versus buy-and-hold. Gains mainly came from the short trend leg after mid-May, with ETH and SOL contributing most clearly.

The effective trading framework for May is: first identify moving-average cluster compression, then allow price to complete directional selection upward or downward; manage failed signals with EMA12; limit loss per trade with a fixed stop-loss of 2.5%; and preserve trend gains with 3R take-profit (3 * 2.5%). This framework fits May’s market structure: low win rate, high payoff-to-risk, and trend legs concentrated.

US stock factors strengthened this judgment. In May, AI-weighted stocks had support from earnings and industry sentiment. Nvidia reported strong quarterly results, with Q1 FY2027 revenue of about 81.6 billion USD, and the stock once again revisited the milestone of a roughly 5 trillion USD market capitalization. The correlation between BTC and the S&P 500 remained at a high level: public samples show that the 30-day correlation within 2026 reached about 0.74 at one point, and near the end of May it was still around 0.6. In May, crypto assets did not break away from the US stock risk-budget framework. Continuous outflows from BTC ETFs, weaker inflow absorption in ETH ETFs, and an increased share of perpetual trading volume jointly led to crypto performance lagging behind US tech leaders.

1. Market structure: surge early, weaken in mid-month, low volatility at month-end

The first phase in May occurred from May 1 to May 6. BTC rose from 77,117.4 USD to 82,828.2 USD, ETH rose from 2,283.02 USD to 2,423.99 USD, and SOL continued to May 11, reaching 98.40 USD. During this stage, short-term moving-average clusters shifted upward, volatility remained within a controllable range, and the market showed characteristics of a repair rally. SOL’s resilience led; at the start of the month, capital was willing to expose to higher risk.

The second phase started on May 7. BTC failed to stabilize above 82,000 USD; ETH could not maintain above 2,400 USD; and SOL formed a monthly high near 98 USD. Breakout signals began to fail repeatedly. After prices retreated toward EMA12, exits were triggered again and again. After May 14, long trades in BTC triggered stop-losses. Long trades in ETH after May 6 continued to fail. SOL entered a clearly downward segment after May 15.

The third phase was concentrated from May 22 to month-end. BTC moved down to around 73,000 USD, ETH approached 2,000 USD, and SOL returned to around 82 USD. Price volatility narrowed, the moving-average cluster width decreased, and the market entered a new compression state.

The intramonth pullbacks confirmed each asset’s role. BTC’s maximum intramonth drawdown was about 12.5%, ETH about 18.8%, and SOL about 18.7%. BTC acted as a risk anchor, while ETH and SOL amplified risk appetite. After BTC weakened, ETH and SOL retraced faster. From a strategy perspective, it became necessary to reduce long exposure in high-beta assets.

2. Capital structure: stablecoins remain, mainstream ETFs weaken

As of May 31, the total market capitalization of stablecoins across the market was about 32 billion USD, and DeFi TVL was about 251 billion USD. Underlying USD liquidity did not experience a systemic outflow. CEX spot trading volume was about 124.2 billion USD over 24 hours, and CEX perpetual trading volume was about 894.4 billion USD; perpetual volume was about 7.2 times spot volume. Price discovery was increasingly carried by the derivatives market.

ETF flows became an important source of pressure in the second half of May. Public news showed that the BTC spot ETF experienced net outflows for 9 consecutive trading days, totaling about 2.8 billion USD. There was even a single-day net outflow of about 649 million USD, including an outflow of about 448 million USD from BlackRock IBIT on that day. ETH ETFs faced similar pressure: in the last week of May, ETH ETF net outflows were about 241 million USD.

However, capital did not completely leave crypto assets. Small net inflows appeared in altcoin ETFs such as SOL and XRP, and new ETF narratives like HYPE also drew attention. Funds shifted from mainstream BTC and ETH ETFs toward thematic ETFs and high-elasticity products. This structure indicates that the core issue was a cooling of mainstream spot absorption, with capital participating in local rotation and short-term trading.

Derivatives data aligns with this. The active buy-sell ratios for BTC, ETH, and SOL were all below 1, and active sell orders were slightly stronger. Funding rates were about 0.01%, not reaching an extreme crowded level. The May market was in a typical state: leverage trading was active, spot follow-through was insufficient, active buy volume was not strong, and breakouts were prone to turning into false moves.

3. US stock linkage: AI leaders support the Nasdaq, while crypto ETFs create capital pressure

Crypto performance in May also needs to be observed within the US stock risk-appetite framework. The correlation between BTC and the S&P 500 remained high. The 30-day correlation had reached about 0.74 within 2026, and near the end of May it was still around the 0.6 range. On a broader scale, BTC in May behaved like a high-beta risk asset and did not have stable independent safe-haven characteristics.

The core support on the US stock side came from AI and large-cap technology stocks. Nvidia released strong quarterly results in May: Q1 FY2027 revenue was about 81.6 billion USD. The stock once again reached historic highs and revisited the milestone of a roughly 5 trillion USD market cap. AI-weighted stocks confirmed the Nasdaq’s risk appetite through earnings. Crypto assets lacked an equivalent earnings anchor; prices were influenced by ETF flows, derivatives leverage, and liquidity expectations.

This cross-asset divergence affected the strategy outcomes in May. US tech leaders had earnings-driven support, while BTC and ETH mainstream ETFs saw outflows. Capital reallocated within large risk assets. Traditional funds continued buying high-certainty AI leaders, reducing willingness to allocate to BTC ETFs. Crypto long breakouts that lacked spot absorption were also more likely to turn into blow-off rallies followed by pullbacks once volume expanded.

The May macro data window also compressed the risk budget. Key data during the month included NFP, CPI, PPI, the second estimate of GDP, and PCE. Employment, inflation, and growth data directly impacted US Treasury yields, the US dollar, and Nasdaq valuations, which then transmitted to BTC via ETF subscriptions/redemptions and perpetual funding rates. Near month-end, the market looked ahead early to June data such as employment, ISM, JOLTS, and ADP, as well as the FOMC and the options expiry window. Low-volatility compression in crypto appeared around these events, and shrinking positions and lowering risk budgets are a reasonable explanation.

4. Volatility: short-cycle compression forms, but price strength is insufficient

By the end of May, BTC’s 7-day 4H realized volatility was about 0.46%, and its 30-day 4H realized volatility was about 0.64%. ETH was about 0.7% and 0.81%, respectively. SOL was about 0.76% and 1%, respectively. The short-cycle volatilities of all three assets were below the medium-cycle volatilities, and the market entered a low-volatility compression phase.

Low-volatility compression implies the market is nearing directional selection, but it does not mean it will break upward. At the end of May, BTC closed at 73,684.0 USD, with EMA12 near 73,776.35 USD. ETH closed at 2,007.0 USD, with EMA12 near 2,016.34 USD. SOL closed at 82.44 USD, with EMA12 near 82.39 USD. BTC and ETH were still in weak ranges, and SOL had just returned to around EMA12. Price strength was insufficient; low vol was closer to consolidation after a decline.

The width of the moving-average clusters shows the same state. BTC’s moving-average cluster width at month-end was about 0.57%, ETH about 0.63%, and SOL about 0.58%, all below the 2.2% strategy threshold. This environment would frequently trigger breakout signals. May data shows that after moving-average cluster compression, dual-direction trading must be allowed. If you only trade upside breakouts, you would systematically miss downtrends.

5. Strategy backtest: 4H moving-average cluster compression breakout system

The strategy uses six moving averages to form the moving-average cluster: EMA6, EMA12, EMA24, SMA6, SMA12, and SMA24. The moving-average cluster width equals (the maximum value of these six moving averages minus the minimum value) divided by the current closing price. If the previous 4H candle’s moving-average cluster width was below 2.2%, and the current 4H close breaks above the upper boundary of the cluster, go long at the next 4H open. If the previous 4H candle’s moving-average cluster width was below 2.2%, and the current 4H close falls below the lower boundary of the cluster, go short at the next 4H open.

The exit rules are fixed. For long positions, exit when price breaks down below EMA12. For short positions, exit when price breaks up above EMA12. The stop-loss per trade is 2.5%, and take-profit is 3R, i.e., 7.5%. If the same candle simultaneously triggers both take-profit and stop-loss, handle it by prioritizing stop-loss. Transaction costs per round trip are deducted at 8 bp. If there is still an open position at month-end, close it at the last 4H closing price.

This report tested two strategy variants using the same framework. The long-only version trades only the upside breakout signal. The dual-direction version trades both the upside breakout and downside breakdown signals. May results show the dual-direction version matched the market state better.

5.1 Long-only strategy: upward breakout signal quality declines

The long-only strategy failed overall. BTC had 11 trades, with a return of -5.36%, a win rate of 18.2%, and a maximum drawdown of -10.08%. ETH had 10 trades, with a return of -6.49%, a win rate of 10.0%, and a maximum drawdown of -10.64%. SOL had 11 trades, with a return of +0.91%, a win rate of 18.2%, and a maximum drawdown of -7.11%.

BTC’s long-only gains were concentrated in the first two trades of the month. Entered on May 1 and exited on May 4, net gain +2.09%. Entered on May 4 and exited on May 7, net gain +0.92%. After that, signal quality deteriorated. A long entered on May 14 triggered the stop-loss, resulting in a net loss of -2.58%.

ETH’s long-only performance was the weakest. Entered on May 1 and exited on May 5, net gain +3.17%. After that, the subsequent 9 long trades were losses. Most of ETH’s upside breakouts were weak rebounds and should not be interpreted as trend expansion.

SOL’s long-only had small profits coming from two trades. Entered on May 5 and exited on May 8, net gain +3.95%. Entered on May 8 and hit the 3R take-profit on May 10, net gain +7.42%. Most of the other signals resulted in losses. SOL was the only asset with positive returns through a long-only approach, with gains highly concentrated.

5.2 Dual-direction strategy: short trend legs contributed most gains

The dual-direction strategy significantly improved results. BTC’s dual-direction strategy return was -2.83%, ETH +3.14%, and SOL +6.05%. The equal-weight dual-direction strategy for the three assets returned +2.11%, while the equal-weight buy-and-hold return for the three assets in the same period was about -6.09%.

BTC’s dual-direction strategy was still a loss, but the loss was smaller than the long-only approach. BTC traded 18 times, with a win rate of 22.2% and a maximum drawdown of -10.74%. Two short trades contributed the most: a short on May 15 exiting on May 20 delivered a net gain of +2.35%; and a short on May 26 exiting on May 30 delivered a net gain of +3.42%. In mid-May, BTC generated more false signals; repeated switching between long and short created costs.

ETH’s dual-direction strategy returned +3.14%, with 18 trades, a win rate of 38.9%, and a maximum drawdown of -8.26%. The key trades were: short on May 15, exiting when it hit 3R take-profit on May 17, net gain +8.03%; and short on May 26, exiting on May 29, net gain +2.68%. ETH’s long signals failed, and the short legs formed the main source of gains.

SOL’s dual-direction strategy returned +6.05%, with 22 trades, a win rate of 22.7%, and a maximum drawdown of -8.17%. SOL provided both long and short trend trades. A long on May 8 exited with 3R take-profit on May 10 at 16:00, net gain +7.42%. A short on May 15 exited when it hit 3R take-profit on May 17, net gain +8.03%. SOL had the strongest trend elasticity, but it also had the highest trading noise.

5.3 Trade distribution: low win-rate structure; a few large trades determine returns

Among the 58 trades in the dual-direction strategy, there were not many profitable trades. BTC’s win rate was 22.2%, ETH’s was 38.9%, and SOL’s was 22.7%. Strategy gains came from a small number of large trend trades, while losses were controlled by EMA12 exits and fixed stop-losses.

Looking at cumulative profit per trade, the strategy’s net value in early May oscillated upward. In mid-May, short trades in ETH and SOL boosted performance. In late May, short trades in BTC and SOL continued to contribute gains. Losses were concentrated in the phase where longs and shorts alternated repeatedly. The system characteristics were low win rate and high payoff-to-risk, suitable for markets where trend legs are clear, but not suitable for dense oscillations.

Splitting contributions between long and short better explains the month’s return sources. BTC’s long contributions were negative, while its short contributions were positive. ETH’s long contributions were negative, while its short contributions were significantly positive. SOL had positive contributions from both longs and shorts, with shorts being more stable. The main line in May was a downtrend after failed attempts to push higher.

Win rate, expected return, and maximum drawdown show that SOL had the highest per-trade expected value, ETH was second, and BTC was the weakest. BTC had the highest density of false breakouts; ETH’s direction signals were cleaner; and SOL had stronger elasticity.

5.4 Exit mechanism: EMA12 controls noise; 3R preserves trend gains

Exit-reason analysis shows that EMA12 exits accounted for the highest proportion. Many trades did not end by stop-loss; instead, after a breakout failure, they exited when price returned near EMA12. The EMA12 rules reduced the time held on erroneous signals and limited loss spread.

The number of stop-loss trades was limited, and losses were concentrated. There were very few 3R take-profit trades, but their contribution to returns was significant. This structure follows the logic of trend-following strategies: most trades produce small losses or small gains, and a small number of trend trades contribute the majority of returns. In May, if the 3R take-profit was removed, the large-trade gains in SOL and ETH would have been cut off earlier. If EMA12 exits were removed, losses during the choppy phase would have expanded.

The signal timeline shows that long signals were dense in early May, short signals began to increase in mid-May, and long/short signals interleaved in late May. Dense signals do not necessarily mean dense opportunities; truly effective signals were concentrated in short windows with clear direction.

5.5 Enhanced filtering of failures: volume breakouts in May are traps

The report also tested an enhanced filtering version. The conditions included: 7-day volatility not higher than 1.15 times the 30-day volatility; volume not less than 0.9 times the moving average of 20 candles (1 candle = 4 hours); long entries near 20 swing highs; and short entries near 20 swing lows. This version performed worse. BTC’s enhanced dual-direction strategy return was -3.40%, ETH -5.03%, SOL -2.58%, and the equal-weighted three-asset portfolio was -3.63%.

The failures happened because in May, volume breakout signals often occurred near local tops. For example, BTC’s enhanced long signal entered on May 4 at 80,322.9 USD was stopped out within 4 hours, for a net loss of -2.58%. ETH’s enhanced long signal entered on May 6 at 2,410.39 USD was also stopped out when the stop-loss condition was triggered, for a net loss of -2.58%. SOL’s enhanced long signals on May 4 likewise triggered stop-losses.

Volume expansion indicates participation, not necessarily the quality of capital. The volume spikes in May were mostly driven by topping, leverage liquidations, and short-term chasing. Effective filtering should also incorporate ETF flows, spot volume share, active buy-sell ratio, perpetual volume share, and US stock risk appetite. Price and volume can only identify volatility; they cannot identify trend absorption.

5.6 Conclusions by asset

BTC is a regime anchor. BTC’s monthly decline was smaller than ETH’s, and its drawdowns were more controllable. BTC’s dual-direction strategy return was -2.83%, indicating BTC itself was not the best returning asset in May. It is more suitable to use as a tool for judging the market’s risk budget. If BTC cannot regain the EMA12 and 30-moving-average area, the long weight for ETH and SOL should be reduced.

ETH is a weak mainline. Its monthly decline was -12.09%, and the success rate of long breakouts was extremely low; the dual-direction strategy relied on short-side profits. After ETH failed to push higher near 2,400 USD, it kept breaking below 2,300, 2,200, and 2,100 USD. The subsequent reassessment of long weights requires first repairing the 2,100 to 2,200 USD area.

SOL is a trading-oriented asset. Its monthly closing decline was only -1.74%, but its intramonth path was extremely volatile. SOL’s dual-direction strategy returned +6.05%, clearly higher than BTC and ETH. SOL is suitable for trend following and not for passive holding. Low win rate, high elasticity, and trend-trade concentration are SOL’s core characteristics in May.

5.7 June strategy framework

In June, we will continue using the dual-direction 4H moving-average cluster breakout system. We should reduce the weight given to one-way chasing longs. BTC serves as a state filter; ETH and SOL serve as profit assets after confirming relative strength. If BTC regains the EMA12 and 30-moving-average area, ETF outflows slow down, and the active buy-sell ratio returns to above 1, then the weight of long signals can be increased. If BTC remains below the 74,000 to 76,000 USD area, the market will stay in a weak repair state.

The US stock filter layer needs to be maintained. Nasdaq and AI leaders remain strong, and if BTC ETF outflows slow down, the cross-asset risk budget is improving again. If Nasdaq is strong while BTC ETFs continue to see outflows, it suggests capital is favoring US tech leaders with higher earnings certainty, and crypto long breakouts still need caution. When US stocks and crypto weaken in sync, short-signal priority for ETH and SOL increases.

Position rules should remain mechanical. Risk per trade is 2.5%, take-profit is 3R, and the EMA12 exit remains unchanged. Breakout signals alone cannot trigger heavy position sizing. When spot absorption is weak, ETF outflows are ongoing, the share of perpetual trading volume is high, and active buy orders are insufficient, then reduce the weight of upside breakout signals and increase the weight of downside breakdown signals.

6. Conclusion

In May, the crypto market completed a state transition from recovery to failure. BTC, ETH, and SOL surged at the beginning of the month; the quality of trends declined after mid-month; and low-volatility compression took over at month-end. Stablecoins and DeFi underlying liquidity remained, mainstream ETF absorption weakened, derivative trading weight increased, and price discovery leaned toward leverage markets.

The strategy results provide a clear answer. Buy-and-hold performed the worst. Long-only breakouts could not adapt to the trend failures after mid-May. The dual moving-average cluster breakout achieved the best results. Equal-weight buy-and-hold across the three assets returned about -6.09%; long-only about -3.65%; and the dual-direction strategy +2.11%. Gains came from the short trend legs of ETH and SOL, as well as the early-month long trend leg of SOL.

US stock factors offer a fuller explanation. AI tech leaders still have earnings-driven support, and stocks like Nvidia back the US stock risk appetite. Meanwhile, BTC and ETH ETFs saw continuous outflows, weakening mainstream crypto asset absorption. The high correlation between BTC and the S&P 500 indicates that crypto is still influenced by US risk budgets and macro interest-rate expectations.

The trading focus in June is not on forecasting direction in advance. A better path is to identify the regime, execute dual-direction signals, control risk per trade, and preserve trend gains. After moving-average cluster compression, both upside and downside breakouts can form valid trades. The EMA12 exit mechanism protects the strategy from being dragged down by false breakouts, while 3R take-profit allows a small number of large trades to cover most small losses. Under current conditions, a disciplined dual-direction system outperforms subjective chasing.

References

• Gate,

• Investor,

• DeFiLlama,

• CMC,

• BlackRock,

• Coinglass,

• K33,

Gate Research Institute is a comprehensive blockchain and cryptocurrency research platform that provides readers with in-depth content, including technical analysis, hotspot insights, market reviews, industry research, trend forecasting, and macroeconomic policy analysis.

Disclaimer

Investing in the cryptoasset market involves high risk. Users are advised to conduct independent research and fully understand the nature of the assets and products they purchase before making any investment decisions. Gate is not responsible for any losses or damages resulting from such investment decisions.