Gate Research Institute: ETF Outflows Suppress Risk Appetite, Two-Way System Navigates Weak Market

marsbitPublished on 2026-06-18Last updated on 2026-06-18

Abstract

Gate Institute Research Report: May 2026 Crypto Market Review & Strategy Analysis In May 2026, the crypto market shifted from an early-month rally to a mid-month correction, concluding with low-volatility consolidation. BTC, ETH, and SOL peaked in early May before declining. The primary market dynamic was a divergence between weakening spot ETF inflows and persistently high leverage-driven perpetual trading volume. A dual-direction moving average cluster breakout strategy outperformed, returning +2.11% for an equally-weighted BTC/ETH/SOL portfolio. This contrasted with a -6.09% return for buy-and-hold and -3.65% for a long-only version of the strategy. Profits were primarily generated from short positions on ETH and SOL during the mid-to-late May downtrend, demonstrating the month's suitability for two-way trend trading. Market structure evolved in three phases: an initial surge (May 1-6), a failure and reversal (starting May 7), and low-volatility compression (May 22 onward). While stablecoin supply remained stable, significant outflows from mainstream BTC and ETH ETFs created selling pressure. Concurrently, high correlation with the S&P 500 (~0.6) and stronger performance from AI equities like Nvidia highlighted crypto's position as a high-beta risk asset within a broader risk-budget framework, lacking independent momentum. The successful strategy employed a 4-hour chart system using a cluster of six moving averages (EMA6,12,24 & SMA6,12,24). A breakout signal was trigg...

• The cryptocurrency market shifted from a rally at the beginning of May to a mid-month pullback and low-volatility consolidation at the month's end. BTC, ETH, and SOL all formed interim highs in early May before entering corrections. Mainstream ETF demand weakened, while perpetual futures trading volume remained high, showing a market structure characterized by weak spot performance and leverage dominance.

• The two-way moving average cluster breakout strategy performed best. The equal-weighted buy-and-hold return for the three assets was approximately -6.09%, the long-only strategy return was about -3.65%, and the two-way strategy return was about +2.11%. Profits mainly came from short trend phases in ETH and SOL, verifying that the May market was more suitable for two-way trend trading.

• Under low-volatility compression, disciplined trading outperforms subjective judgment. The EMA12 exit mechanism effectively controlled false breakout losses, and the 3R take-profit preserved trend gains. The current market remains in a phase of directional selection; identifying the state, controlling risk, and executing two-way signals is a superior trading framework to subjectively chasing rallies.

The primary contradiction in the May crypto market was the divergence between price action and market structure after the initial rally, with weak spot buying and active leverage trading diverging. BTC, ETH, and SOL all formed interim highs in early May before entering pullbacks and low-volatility consolidation. BTC retreated from its 4H closing price of $77,117.4 at the start of the month to $73,684.0 at the month's end, a monthly return of -4.45%. ETH fell from $2,283.02 to $2,007.0, a monthly return of -12.09%. SOL retreated from $83.90 to $82.44, a monthly return of -1.74%. SOL's closing decline was smaller, but it reached a high of $98.40 during the month before retreating to around $80.00, indicating significantly higher intra-month trading volatility than suggested by its monthly return.

The strategy backtest results for the month are clear. The equal-weighted buy-and-hold return for the three assets was approximately -6.09%; the long-only moving average cluster breakout strategy return was about -3.65%; the two-way moving average cluster breakout strategy return was +2.11%. The two-way strategy achieved an excess return of about +8.2% compared to buy-and-hold. Profits mainly came from short trend legs after mid-May, with ETH and SOL being the most significant contributors.

The effective trading framework for May was: first, identify moving average cluster compression; then, allow price to complete its directional choice, either up or down; use EMA12 to manage failed signals, a fixed 2.5% stop-loss to limit single-trade losses, and a 3R take-profit (3 * 2.5%) to preserve trend gains. This framework suited May's market structure of low win rate, high reward-to-risk ratio, and concentrated trend legs.

U.S. stock market factors reinforced this view. In May, AI-heavyweight stocks had earnings and industry momentum support. Nvidia reported strong quarterly earnings, with Q1 FY2027 revenue around $81.6 billion, and its market cap briefly touched the $5 trillion milestone again. The correlation between BTC and the S&P 500 remains high, with public samples showing the 30-day correlation reached about 0.74 within 2026 and was still around 0.6 near the end of May. Cryptocurrency assets did not detach from the U.S. stock market risk budget framework in May. Continuous BTC ETF outflows, weakening ETH ETF demand, and a high share of perpetual trading volume collectively led to crypto underperformance compared to U.S. tech giants.

I. Market Structure: Early-Month Rally, Mid-Month Weakness, End-Month Low Volatility

The first phase of May occurred from May 1st to May 6th. BTC rose from $77,117.4 to $82,828.2, ETH rose from $2,283.02 to $2,423.99, and SOL extended its rally to touch $98.40 on May 11th. During this phase, short-term moving average clusters shifted upward, volatility was within a controllable range, and the market exhibited characteristics of a recovery rally. SOL showed leading elasticity, indicating a willingness among capital to take higher risk exposure at the start of the month.

The second phase began on May 7th. BTC failed to stabilize above $82k, ETH could not sustain above $2,400, and SOL formed its monthly high near $98. Upside breakout signals began to fail frequently, with prices falling back to the EMA12 and repeatedly triggering exits. Long trades in BTC triggered stop-losses after May 14th, long trades in ETH consistently failed after May 6th, and SOL entered a clear downtrend leg after May 15th.

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

Intra-month drawdowns verified asset roles. BTC's maximum intra-month drawdown from high to low was about 12.5%, ETH's was about 18.8%, and SOL's was about 18.7%. BTC serves as the risk anchor, while ETH and SOL act as amplifiers of risk appetite. After BTC weakened, ETH and SOL's drawdowns were faster; at a strategic level, this necessitates reducing the long-side weighting of high-Beta assets.

II. Capital Structure: Stablecoins Remain, Mainstream ETF Demand Weakens

As of May 31st, the total market cap of stablecoins was approximately $3.2 trillion, and DeFi TVL was about $2.51 trillion. The underlying US dollar liquidity did not experience a systemic retreat. CEX 24-hour spot trading volume was approximately $1.242 trillion, while CEX perpetual trading volume was about $8.944 trillion, with perpetual volume being roughly 7.2 times that of spot volume. Price discovery is increasingly borne by the derivatives market.

ETF flows became a significant pressure point in the latter half of May. Public news indicates that BTC spot ETFs experienced net outflows for nine consecutive trading days, totaling about $2.8 billion, including a single-day net outflow of approximately $649 million at one point, with BlackRock's IBIT seeing a single-day outflow of about $448 million. ETH ETFs also faced pressure, with a net outflow of about $241 million in the last week of May.

However, capital did not completely leave the crypto asset space. Altcoin ETFs like SOL and XRP saw small net inflows, and new ETF narratives like HYPE also gained attention. Capital migrated from mainstream BTC and ETH ETFs towards thematic ETFs and higher-beta assets. This structure indicates the core issue was cooling demand for mainstream spot assets, with capital participating in local rotations and short-term trades.

Derivatives data is consistent with this. Active buy/sell ratios for BTC, ETH, and SOL were all below 1, indicating slightly stronger active selling pressure. Funding rates were around 0.01%, not reaching extreme crowding levels. The May market was in a typical state: active leverage trading, insufficient spot follow-through, weak active buying pressure, and price breakouts prone to turning into false moves.

III. U.S. Stock Linkage: AI Giants Support Nasdaq, Crypto ETF Outflows Create Pressure

May's crypto performance must also be viewed within the U.S. stock market risk appetite framework. The correlation between BTC and the S&P 500 remained high, with the 30-day correlation reaching around 0.74 within 2026 and still around 0.6 near the end of May. On a broader scale, BTC behaved as a high-Beta risk asset this month, lacking stable, independent safe-haven characteristics.

The core support on the U.S. stock side came from AI and large-cap tech stocks. Nvidia reported strong quarterly earnings in May, with Q1 FY2027 revenue around $81.6 billion; its stock price hit new all-time highs and briefly retouched the ~$5 trillion market cap milestone. AI-heavyweight stocks supported Nasdaq risk appetite through earnings confirmation. Cryptocurrency assets lack a comparable earnings anchor, with prices influenced by ETF flows, derivatives leverage, and liquidity expectations.

This cross-asset divergence impacted May strategy results. U.S. tech giants have earnings drivers, while BTC and ETH mainstream ETFs saw outflows, leading to a reallocation of capital within large risk assets. Traditional capital continued buying AI giants with higher earnings certainty, reducing willingness to allocate to BTC ETFs. Crypto long breakouts lacked spot demand, causing high-volume breakouts to easily turn into sell-the-news events.

May's macro data calendar also compressed the risk budget. Key monthly data included NFP, CPI, PPI, the second GDP estimate, and PCE. Employment, inflation, and growth data directly impacted U.S. Treasury yields, the dollar, and Nasdaq valuations, which then transmitted to BTC ETF flows and perpetual funding rates. Towards month-end, the market began focusing on early June data releases like employment, ISM, JOLTS, ADP, as well as the FOMC meeting and options expiry. Crypto's low-volatility compression occurred around these events; position unwinding and reduced risk budgets are plausible explanations.

IV. Volatility: Short-Term Compression Forms, Price Strength Lacking

By the end of May, BTC's 7-day 4H realized volatility was approximately 0.46%, and its 30-day 4H realized volatility was about 0.64%; ETH's were around 0.7% and 0.81% respectively; SOL's were around 0.76% and 1% respectively. Short-term volatility for all three major assets was lower than medium-term volatility, indicating the market had entered low-volatility compression.

Low-volatility compression suggests the market is approaching a directional choice, but does not imply an upward breakout. At May's end, BTC closed at $73,684.0, with the EMA12 around $73,776.35; ETH closed at $2,007.0, with the EMA12 around $2,016.34; SOL closed at $82.44, with the EMA12 around $82.39. BTC and ETH remained in weak territory, while SOL had just returned near the EMA12. Price strength is insufficient; low volatility appears more like consolidation after a decline.

Moving average cluster width shows a similar state. At month-end, BTC's cluster width was about 0.57%, ETH's about 0.63%, and SOL's about 0.58%, all below the 2.2% strategy threshold. This environment frequently triggers breakout signals. May's data shows that after moving average cluster compression, two-way trading must be permitted. Trading only long breakouts systematically misses downtrends.

V. Strategy Backtest: 4H Moving Average Cluster Compression Breakout System

The strategy uses six moving averages to form a cluster: EMA6, EMA12, EMA24, SMA6, SMA12, SMA24. The cluster width equals the maximum value minus the minimum value of the six moving averages, divided by the current closing price. If the previous candle's cluster width is below 2.2% and the current candle closes above the cluster's upper bound, go long at the next 4H open. If the previous candle's cluster width is below 2.2% and the current candle closes below the cluster's lower bound, go short at the next 4H open.

Exit rules are fixed. Exit long positions if price closes below the EMA12. Exit short positions if price closes above the EMA12. Single-trade stop-loss is set at 2.5%, and take-profit is set at 3R, i.e., 7.5%. If both take-profit and stop-loss are triggered on the same candle, priority is given to stop-loss. A round-trip transaction cost of 8 basis points is deducted. Any open position at month-end is closed at the last 4H closing price.

This report tests two strategy variations. The long-only version only trades upside breakout signals. The two-way version trades both upside and downside breakout signals. May's results show the two-way version better fits the market state.

5.1 Long-Only Strategy: Quality of Upside Breakout Signals Declines

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

BTC's long-only profits were concentrated in the first two trades. Entry on May 1st, exit on May 4th, net profit +2.09%. Entry on May 4th, exit on May 7th, net profit +0.92%. Signal quality declined afterward; the long trade entered on May 14th triggered a stop-loss, net loss -2.58%.

ETH's long-only performance was weakest. Entry on May 1st to exit on May 5th, net profit +3.17%. All subsequent 9 long trades resulted in losses. ETH's upside breakouts were mostly weak rebounds, not trend expansions.

SOL's long-only strategy generated a slight profit, coming from two trades. Entry on May 5th, exit on May 8th, net profit +3.95%. Entry on May 8th, triggered 3R take-profit on May 10th, net profit +7.42%. Most other signals were losing trades. SOL was the only asset with a positive return via the long-only strategy this month, with profits highly concentrated.

5.2 Two-Way Strategy: Short Trend Legs Contribute Most Profits

The two-way strategy significantly improved results. BTC's two-way strategy return was -2.83%, ETH's was +3.14%, and SOL's was +6.05%. The equal-weighted two-way strategy return for the three assets was +2.11%, compared to an equal-weighted buy-and-hold return of about -6.09% over the same period.

BTC's two-way strategy still resulted in a loss, but the loss magnitude was smaller than the long-only strategy. BTC had 18 trades total, a win rate of 22.2%, and a maximum drawdown of -10.74%. Two short trades contributed most: short entry on May 15th, exit on May 20th, net profit +2.35%; short entry on May 26th, exit on May 30th, net profit +3.42%. BTC had numerous false signals in mid-May, causing losses from frequent long/short switches.

ETH's two-way strategy returned +3.14%, with 18 trades, a win rate of 38.9%, and a maximum drawdown of -8.26%. Key trades were: short on May 15th, triggered 3R take-profit on May 17th, net profit +8.03%; short on May 26th, exit on May 29th, net profit +2.68%. ETH's long signals failed, with short legs constituting the main profits.

SOL's two-way 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. Long on May 8th, triggered 3R take-profit at 16:00 on May 10th, net profit +7.42%. Short on May 15th, triggered 3R take-profit on May 17th, net profit +8.03%. SOL's trend elasticity was strongest, but trading noise was also highest.

5.3 Trade Distribution: Low Win Rate Structure, Few Large Trades Determine Profits

Among all 58 trades in the two-way strategy, the number of profitable trades was limited. BTC's win rate was 22.2%, ETH's was 38.9%, and SOL's was 22.7%. Strategy profits came from a few large trend trades, while losses were controlled by the EMA12 exit and fixed stop-loss.

Cumulative profit per trade shows the strategy's net value oscillated upward in early May, was boosted in mid-May by ETH and SOL short trades, and continued to benefit from BTC and SOL short trades in late May. Losses concentrated during phases of frequent long/short switching. The system is characterized by low win rate and high reward-to-risk ratio, suitable for markets with clear trend legs, not dense consolidation.

A breakdown of long/short contributions further clarifies the profit source for the month. BTC's long-side contribution was negative, short-side positive. ETH's long-side contribution was negative, short-side significantly positive. SOL had positive contributions from both long and short sides, with the short side more stable. May's main theme was a downtrend following failed rallies.

Win rate, expected return per trade, and maximum drawdown show SOL had the highest single-trade expectation, followed by ETH, with BTC weakest. BTC had the highest density of false breakouts, ETH's direction was cleaner, and SOL's elasticity was strongest.

5.4 Exit Mechanism: EMA12 Controls Noise, 3R Preserves Trend Profits

Analysis of exit reasons shows the EMA12 exit accounted for the highest proportion. Many trades did not end via stop-loss but exited when the failed breakout returned near the EMA12. The EMA12 rule reduced holding time and loss expansion from erroneous signals.

The number of stop-loss trades was limited, but losses were concentrated. The number of 3R take-profit trades was very small, but profit contribution was significant. This structure aligns with trend-following strategy patterns: most trades are small losses or small gains, with a few trend trades contributing the majority of profits. In May, removing the 3R take-profit would have prematurely cut SOL and ETH's large profitable trades; removing the EMA12 exit would have expanded losses during consolidation phases.

The signal timeline shows dense long signals in early May, increasing short signals in mid-May, and alternating long/short signals in late May. Dense signals do not equate to dense opportunities; truly effective signals concentrated in short windows with clear direction.

5.5 Enhanced Filter Failure: Volume Breakouts Were Traps in May

The report also tested an enhanced filter version. Conditions included: 7-day volatility not exceeding 1.15 times the 30-day volatility, volume not below 0.9 times the 20-period (1 period = 4 hours) average volume, long signals near the 20-period high, short signals near the 20-period low. This version performed worse. BTC's enhanced two-way strategy return was -3.40%, ETH's was -5.03%, SOL's was -2.58%, and the equal-weighted three-asset return was -3.63%.

The failure reason was that May's volume breakouts often occurred near interim tops. BTC's enhanced long signal on May 4th entered at $80,322.9, triggered stop-loss within 4 hours, net loss -2.58%. ETH's enhanced long signal on May 6th entered at $2,410.39, triggered stop-loss on the same candle, net loss -2.58%. SOL's enhanced long signal on May 4th also triggered stop-loss.

High volume indicates participation, not quality of capital. In May, high volume more often resulted from topping churn, leverage liquidations, and short-term chasing. Effective filters should incorporate ETF flows, spot trading volume share, active buy/sell ratios, perpetual trading volume share, and U.S. stock market risk appetite. Price and volume alone can identify volatility, not trend sustainability.

5.6 Per-Asset Conclusions

BTC is the State Anchor. BTC's monthly decline was smaller than ETH's, and its drawdown was more controlled. BTC's two-way strategy return of -2.83% indicates BTC itself was not the best asset for returns in May. It is better suited for gauging overall market risk budgets. If BTC cannot reclaim the EMA12 and 30-period moving average area, long-side weighting for ETH and SOL should be reduced.

ETH is the Weakness Mainstream. Monthly decline of -12.09%, extremely low win rate for long breakouts, two-way strategy reliant on short-side profits. ETH failed near $2,400 and broke down through $2,300, $2,200, $2,100 successively. It first needs to repair the $2,100 to $2,200 zone before long-side weighting can be reassessed.

SOL is the Trading Asset. Monthly closing decline was only -1.74%, but the intra-month path was very volatile. SOL's two-way strategy return of +6.05% was significantly higher than BTC and ETH. SOL is suited for trend following, not passive holding. Low win rate, high elasticity, and concentrated trend trades were SOL's core characteristics in May.

5.7 June Strategy Framework

Continue using the two-way 4H moving average cluster breakout system in June. Reduce weighting for one-sided long chasing. Use BTC as a state filter, with ETH and SOL as profit assets confirmed by relative strength. If BTC reclaims the EMA12 and 30-period moving average area, while ETF outflows slow and active buy/sell ratios return above 1, long signal weighting can be increased. If BTC remains below the $74k-$76k zone, the market is still in a weak recovery state.

Keep the U.S. stock filter layer. If Nasdaq and AI leaders remain strong while BTC ETF outflows slow, it indicates improving cross-asset risk budgets. If Nasdaq is strong while BTC ETF outflows persist, it suggests capital favors U.S. tech giants with higher earnings certainty; caution is still needed on crypto long breakouts. If both U.S. stocks and crypto weaken, priority should increase for short signals on ETH and SOL.

Maintain mechanical position sizing rules. Single-trade risk 2.5%, take-profit 3R, EMA12 exit unchanged. Breakout signals alone should not trigger heavy positions. When spot demand is weak, ETF outflows persist, perpetual trading share is high, and active buying is insufficient, reduce weighting for long breakout signals and increase weighting for short breakout signals.

VI. Conclusion

The May cryptocurrency market completed a state transition from recovery to failure. BTC, ETH, and SOL all rallied early in the month; trend quality deteriorated after mid-month, entering low-volatility compression by month-end. Stablecoin and DeFi underlying liquidity remained, but mainstream ETF demand weakened, derivatives trading gained weight, and price discovery leaned towards leveraged markets.

Strategy results provide a clear answer. Buy-and-hold performed worst; long-only breakouts could not adapt to the trend failure after mid-May; the two-way moving average cluster breakout achieved the best results. Equal-weighted buy-and-hold for the three assets returned approximately -6.09%, long-only about -3.65%, and the two-way strategy +2.11%. Profits came from short trend legs in ETH and SOL, as well as SOL's long trend leg early in the month.

U.S. stock market factors provide a more complete explanation. AI tech leaders still had earnings drivers, with heavyweights like Nvidia supporting U.S. stock risk appetite; BTC and ETH ETFs experienced continuous outflows, weakening demand for mainstream crypto assets. The correlation between BTC and the S&P 500 remained high, indicating the crypto market is still influenced by U.S. stock risk budgets and macro interest rate expectations.

The focus for June trading is not about predicting direction in advance. A superior path is to identify the state, execute two-way signals, control single-trade risk, and preserve trend profits. 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 the 3R take-profit allows a few large winners to cover many small losses. In the current environment, a disciplined two-way system is superior to subjective rally-chasing.

References

• Gate, https://www.gate.com/trade/BTC_USDT

• Investor, https://investor.nvidia.com/news/press-release-details/2026/NVIDIA-Announces-Financial-Results-for-First-Quarter-Fiscal-2027/default.aspx

• DeFiLlama, https://defillama.com/stablecoins

• CMC, https://coinmarketcap.com/charts/

• BlackRock, https://www.blackrock.com/us/individual/products/333011/ishares-bitcoin-trust

• Coinglass, https://www.coinglass.com/etf/bitcoin

• K33, https://k33.com/research/articles/waiting-and-seeing

Gate Research Institute is a comprehensive blockchain and cryptocurrency research platform providing readers with in-depth content, including technical analysis, hotspot insights, market reviews, industry research, trend forecasts, and macro-economic policy analysis.

Disclaimer

Investing in the cryptocurrency market involves high risks. Users are advised to conduct independent research and fully understand the nature of purchased assets and products before making any investment decisions. Gate bears no responsibility for any loss or damage caused by such investment decisions.

Related Questions

QAccording to the article, what were the key factors that led to the crypto market's weakness in May compared to US tech stocks?

AThe article identifies two main factors: 1) Crypto-specific: BTC and ETH ETFs experienced consecutive outflows (e.g., BTC ETFs had net outflows for 9 days, totaling approx. $2.8B), weakening spot market support. Simultaneously, perpetual futures trading volume remained high (approx. 7.2x spot volume), shifting price discovery to the leverage-dominated market. 2) Broader risk-on framework: US tech/AI stocks like Nvidia had strong earnings support, attracting traditional capital and sustaining Nasdaq risk appetite. This led to a capital rotation *within* risk assets away from crypto ETFs and towards high-conviction AI stocks. Consequently, crypto lacked similar fundamental anchors and underperformed despite high correlation with the S&P 500.

QDescribe the performance difference between the 'long-only' and 'bidirectional' versions of the 4H moving average cluster breakout strategy in May's market.

AThe performance difference was significant and demonstrated the bidirectional strategy's superiority in the May market structure. For an equal-weighted portfolio of BTC, ETH, and SOL: the Buy & Hold return was approximately -6.09%. The Long-Only strategy return was -3.65%, still negative but with less drawdown. The Bidirectional strategy returned +2.11%, generating a +8.2% alpha over Buy & Hold. The profit for the bidirectional strategy primarily came from short positions taken on ETH and SOL during the downtrend leg after mid-May. This result validated that the market structure (low win rate, high payoff ratio, concentrated trend legs) was more suitable for two-way trend trading.

QWhat were the core rules for entry, exit, and risk management in the bidirectional 4H moving average cluster breakout system?

AThe core rules were: 1) Entry: Trigger when the previous 4H candle's moving average cluster 'width' (range of 6 MAs: EMA6,12,24, SMA6,12,24) is below 2.2%. A long signal triggers if the current candle closes above the cluster's upper bound; a short signal triggers if it closes below the lower bound. Entry is at the next 4H open. 2) Exit: Long positions exit if price falls below the EMA12; short positions exit if price rises above the EMA12. 3) Risk Management: A fixed 2.5% stop-loss per trade. A 3R (3 * 2.5% = 7.5%) take-profit level. If a candle triggers both stop-loss and take-profit, stop-loss takes priority. An 8bp cost is deducted per round-trip trade.

QHow did the article characterize the roles of BTC, ETH, and SOL within the May market structure and strategy results?

AThe article characterized them as follows: 1) BTC: The 'state anchor' or risk benchmark. It had the smallest monthly decline (-4.45%) and most controlled drawdown (~12.5%). Its bidirectional strategy return was still negative (-2.83%), indicating it wasn't the best profit asset but crucial for gauging overall market risk appetite. 2) ETH: The 'weakness leader'. It had the largest monthly decline (-12.09%). Its long-only breakout signals had very low win rates, and its bidirectional strategy profit (+3.14%) depended entirely on short positions. 3) SOL: The 'trading asset'. It showed high elasticity (volatility), with a modest monthly close decline (-1.74%) but significant intra-month swings. Its bidirectional strategy delivered the highest return (+6.05%) from both long (early May) and short (mid-May) trend legs, fitting a high-payoff, low-win-rate trend-following model.

QWhy did the 'enhanced filter' version of the strategy (with volume and volatility conditions) perform worse in May, and what does this imply about market behavior?

AThe enhanced filter version performed worse because its conditions (e.g., volume above 90% of the 20-period average, price near 20-period high/low) often identified breakouts that occurred near local tops, which then quickly reversed. For example, enhanced long signals on May 4th (BTC) and May 6th (ETH) triggered stop-losses within the same candle. This failure implies that in May, high-volume breakouts did not signify strong, sustained buying pressure (high-quality capital) but rather represented activities like profit-taking, leverage unwinding, and short-term chasing. The market structure was such that price and volume signals alone were insufficient to identify genuine trend continuation; additional filters like ETF flows, spot vs. perpetual volume ratios, and active buy/sell ratios were needed to gauge underlying conviction.

Related Reads

CICC: How to view the trends of US stocks and the AI hardware industry chain?

U.S. stocks suffered their largest single-day drop since the April 2025 tariff shock on November 5, with the S&P 500, Nasdaq, and Dow falling 4.4%, 5.5%, and 3.4%, respectively. AI-related names led the decline: the Nasdaq 100 dropped over 6% and the Philadelphia Semiconductor Index plunged about 12%, both marking their worst single-day performances since March 2020. The selloff was broad-based—all 11 S&P sectors fell, including defensive staples and utilities—signaling possible panic and stampede conditions. The VIX rose to 28, a dangerous level last seen during the Iran situation earlier this year, and the gap between spot VIX and three-month futures widened to its largest since April 2025, a sign of market fear and potential further pressure. The direct trigger was heightened fear that Democrats, after midterm election wins, might increase AI regulation and halt data center construction. Comments from newly elected Texas Governor Gina Hinojosa and federal-level calls by Sanders to pause national data center construction amplified concerns about the pace and sustainability of AI infrastructure investment. Markets have whipsawed as they "choose" a direction to trade amid uncertainty over which Democratic agenda items will advance. At this point, short-term moves are no longer fundamentally driven. With U.S. leverage at record highs—FINRA margin debt implying leverage of 3.4 and CBOE put/call open interest ratio at 1.33, the highest since end-2018—markets are highly sensitive to price and volatility swings, raising the risk of margin calls, forced deleveraging, and contagion from AI names to other holdings. So far the stress is confined to equities; rising Treasury yields and a stronger dollar would signal more acute liquidity risk. AI is in a transitional phase lacking new demand or model narratives and catalysts, magnifying credit and regulatory risks. High capex is squeezing cloud providers' free cash flow, fueling worries about AI revenue realization and ROI. Stabilization requires natural clearing or a strong catalyst to reverse pessimistic expectations. The first half of next year is a key verification window—markets need to see breakthroughs akin to Anthropic's outperformance in Q1 this year. Otherwise, patience may run out and trading could shift toward recessionary positioning. Structurally, AI remains the main theme, but investors should favor segments with more sustainable and certain earnings. Valuations are not at extreme highs overall, but AI hardware names such as optical modules, semiconductor manufacturing, and liquid cooling have priced in higher growth expectations, while cloud providers look relatively reasonably valued. Cloud providers have room for further gains, but optical module names risk further valuation cuts if high growth cannot persist. The path of least resistance in trading has converged on Hang Seng Tech and gold: Hang Seng Tech has low AI exposure and benefits from eased geopolitical risk, lower oil prices, and reduced rate-hike pressure, though its problems are internal and southbound selling plus rising short interest signal growing divergence. Gold faces fewer obstacles and remains supported by political risk and expectations of weaker dollar and Treasury yields. On tech hardware: near-term headwinds for overseas AI infrastructure, but continued optimism on domestic computing power during the 15th Five-Year Plan. Most Chinese AI hardware names have corrected over 50% since July. Near-term North American data center changes may alter the slope of hardware demand but not the long-term uptrend; domestic computing demand may be even stronger. Overseas-chain valuations have digested premiums—the SW Communications Equipment index trades at 28.92x PE (TTM), below the 10th percentile of the past three years—and the sharp selloff has squeezed out bubble valuation. Q3 2026 capex at the top four U.S. cloud providers grew 84.2% year over year, with Azure and AWS revenue growing strongly, showing demand converting to revenue; 2027 should still be a year of active cloud investment. For the domestic chain, demand for compute chips and servers remains robust, with constraints on the supply side; the overseas pause provides a catch-up window for domestic hardware, and core localization logic for foundries, semiconductor equipment, and components is unchanged. After the correction, large manufacturing valuations are attractive, and the domestic compute chain is favored during the 15th Five-Year Plan. Risks: AI progress below expectations, slower domestic hardware penetration, and policy changes.

marsbit4m ago

CICC: How to view the trends of US stocks and the AI hardware industry chain?

marsbit4m ago

Wall Street Collectively Turned Bullish on US Stocks After the Midterms, Only to Be Slapped in the Face One Day Later

After the US midterm elections, major Wall Street trading desks—including JPMorgan, Citadel Securities, and UBS—collectively turned bullish on US equities, arguing that compressed pre-election positioning and fading uncertainty would drive capital back into the market. The S&P 500 rose nearly 2% and the Dow gained over 3% on Wednesday, briefly validating that view. But the rally reversed within a single day: anxiety over AI investment returns and computing infrastructure costs erupted, wiping $2.3 trillion from Nasdaq 100 market value and sending the S&P 500 down 4.4%, with every sector falling. Goldman Sachs' John Flood said Thursday was the mirror opposite of Wednesday's busiest trading day, with long-only investors frozen again. Barclays' Alex Altmann, one of the few skeptics, had warned that the election would not resolve the core issues压制ing stocks—AI return prospects and financing costs for compute buildout. The selloff was amplified by systematic strategies: volatility-control funds may have cut roughly 30 percentage points of equity exposure in a single day, implying about $135 billion in futures selling over 24-48 hours, while Thursday was likely the largest leveraged ETF rebalancing sell day on record. Retail investors, previously a dip-buying force, also retreated, with Vanda Research data showing net sales of $338 million by 3pm Thursday, potentially the largest single-day net selling since the pandemic. Despite the blowup, bullish views were not fully abandoned—S&P 500 futures rose 0.3% and Nasdaq 100 futures jumped 1.5% Friday morning—but the episode underscored how even top institutions struggle to predict market direction amid structural AI valuation repricing.

marsbit5m ago

Wall Street Collectively Turned Bullish on US Stocks After the Midterms, Only to Be Slapped in the Face One Day Later

marsbit5m ago

This Round of Sharp Decline: Technology Giants Clearly Oversold

The article argues that the recent market selloff—triggered by Democratic wins in the midterms and new state governors in data-center hubs like Texas and Ohio pledging to pause approvals—has conflated two very different things: companies whose revenue is tightly bound to new data-center construction, and companies holding scarce assets or whose demand doesn't depend on that path. The former deserve repricing; the latter are being wrongly sold off. The author applies three criteria—scarcity, order visibility, and reasonable or undervalued pre-crash valuation—to identify the truly mispriced names. TSMC is the clearest case, meeting all three: it is the only scalable 3nm/2nm pure-play foundry with ~90% AI chip foundry share, its advanced-node capacity is booked past 2028, and only about 30% of revenue ties to AI data centers, with HPC at 68% but smartphones, auto, and IoT providing a buffer. Its ADR trades near 24x PE after the drop, low for semiconductors. Nvidia fell ~10% overnight and 25% in a month; its 17x forward PE is a three-year low and order visibility is strong (management cites chip supply, not demand, as the constraint), but with data-center revenue over 90% of the total, it is a high-beta bet rather than a defensive mispricing. Apple is defensive—capex is only ~2.5% of sales and on-device AI runs on already-sold chips—but its 36x forward PE sits at the 99th percentile of its own history, so it fails the valuation test. Google and Amazon have huge backlogs ($514B and $496B respectively) but their conversion depends on compute supply, so they aren't fully decoupled. Meta, at 18–19x forward PE near a three-year low, is partly mispriced: its three flagship US projects (Ohio, Louisiana, Texas) are already past permitting, though new pipelines will slow. Broadcom, AMD, and Intel are falling for real exposure reasons, not mispricing. In China, SMIC best fits the criteria, with 89.6% China revenue tied to domestic fabless demand rather than US cloud orders; North China Innovation has strong visibility but a fuller valuation; ChangXin Memory shows an emotional discount versus fundamentals. Among model makers, US pure-plays are hit hardest, while Chinese firms' structural cost advantages are amplified—Zhipu, running GLM-5.3 on AWS Bedrock with revenue sharing, benefits from existing cloud capacity and may be somewhat mispriced. The conclusion: only TSMC truly satisfies all three criteria; the broad "anything AI rises" phase is over, and differentiation will favor companies with real orders, revenue, and compute reserves.

marsbit34m ago

This Round of Sharp Decline: Technology Giants Clearly Oversold

marsbit34m ago

Breaking News: Claude Cracks the First Prokaryotic Giant Protein, a 40-Year Mystery Structure Revealed

In a breakthrough spanning less than four months, Anthropic's Claude helped solve a 40-year-old biological mystery: the first three-dimensional structure of a prokaryotic "Vault" particle. Discovered in 1986 by Leonard Rome's UCLA team, vaults are among the largest protein complexes in cells—three times the mass of a ribosome—but their function remained unknown, since common model organisms like yeast, fruit flies, and nematodes lack them, and mice lacking vault genes develop normally. The team used Claude to search roughly 15 billion protein sequences, expanding the known vault protein family from 41 to 1,507 members—a 36-fold increase—across 29 bacterial phyla and 4 archaeal phyla, often near antiviral defense genes. Of 44 candidate proteins, 15 produced soluble protein in E. coli. Claude initially opposed using size-exclusion chromatography, correctly noting that misfolded protein aggregates would co-elute with vaults, but after being overruled it helped optimize the approach—including the counterintuitive suggestion to load crude lysate directly onto the column, cutting purification from three days to just 90 minutes. Eight candidates yielded clear vault particles, the first ever seen outside eukaryotes. Using a multi-agent framework, Claude ran RELION across hundreds of GPUs to reconstruct the phage L05 vault in just 34 hours, reaching 3.2 Å resolution within a week. The structure revealed 96 identical chains forming a shell 25% larger inside than human vaults, plus a hidden coenzyme A molecule tucked into each chain's inner pocket—hinting the vault serves as a molecular container during viral attack or metabolic stress. Engineered single-amino-acid changes opened the shell into flat "protein lotus" arrays, while inserting repeat modules expanded interior volume by 50% over human vaults. All code and computational analysis were performed by Claude, which the paper formally credits. Leonard Rome himself praised the work as opening new paths to understanding vault function, signaling a "Moore's Law" moment for basic science discovery.

marsbit36m ago

Breaking News: Claude Cracks the First Prokaryotic Giant Protein, a 40-Year Mystery Structure Revealed

marsbit36m ago

Tibo Reassigned: The GPT Reset Button Has a New Owner

Tibo, OpenAI's "reset god" known for frequently resetting usage limits, has been reassigned to lead OpenAI's internal systems security project, according to Reuters. His replacement appears to be Andrew Ambrosino, a founding member of the Codex App team, who has already begun handling user complaints and product feedback publicly. Tibo's move comes amid longstanding employee concerns about operational security at OpenAI, including overly broad access to sensitive code, insufficient security reviews of internal tools, and insider threats. In October, an engineer was prosecuted after planting listening devices in an executive meeting room and hiding a recorder in toilet paper. Since August, Tibo had already reassigned roughly a quarter of OpenAI's production engineers to security infrastructure work, and he has publicly announced forming a world-class cybersecurity team. Andrew Ambrosino studied computer science at UMass Lowell and previously worked at MITRE and Microsoft, founded fintech startup Catch where he served as CTO/CPO for seven years, and later worked at Noyo before joining OpenAI in 2025 as a core member of the Codex App team. He has since led that team and expanded into ChatGPT product work. Recently he announced restored double usage limits for Pro and Business Premium tiers and pledged to improve ChatGPT's simplicity and performance. Users have flocked to Andrew's comments with requests, including better treatment for Plus users, keeping Chat and Work separate, removing the five-hour usage limit, opening Dots to Plus users, and—most notably—begging him to hit the reset button on day one. OpenAI has not officially confirmed Andrew's full takeover of Tibo's former responsibilities, but his public actions suggest he is already stepping into the role.

marsbit38m ago

Tibo Reassigned: The GPT Reset Button Has a New Owner

marsbit38m ago

Trading

Spot
活动图片