Trade Note: Long ETH/BTC Relative Value

1. The Trade

Long ETH/BTC at ~0.0296, expressed as long spot ETH / short BTCUSDT perp, equal USD notional. Entry in two tranches: half at market (~0.0296), half resting at 0.0290–0.0294 to average in on a pullback. Risk budget: ~50bp of NAV to the stop, sized small on purpose. This is a real, verified signal, not a high-conviction one.

2. Thesis

Two things are true and independently verified, not narrative:

The ratio has drifted from ~0.0289 to ~0.0296 (+2.4%) over 30 days while BTC went nowhere in dollar terms (~$64,670 → ~~$64,642) and ETH quietly gained 2.3% (~~$1,871 → ~$1,914).
Binance perpetual funding shows a mild, not stretched, ETH premium: BTCUSDT funding annualizes to roughly 1%, ETHUSDT to roughly 5%. For reference, a genuinely crowded long print looks like 20-50%+ annualized. There's a real bid for ETH exposure and essentially no leverage froth behind it.

What makes this a trade rather than a trivia fact: ETH is outperforming inside a fearful tape. Fear & Greed sits around 35 ("Fear"), BTC is pinned in a $3.7k range, and reported ETF flows are net negative: a risk-off backdrop. ETH is the higher-beta asset in this market; higher-beta assets aren't supposed to quietly out-earn BTC when sentiment is scared. It's doing it anyway. That could mean the marginal ETH seller is exhausted relative to BTC's, or that there's a persistent, low-key bid nobody's talking about yet. The trade doesn't require knowing which. The relative price action and the funding differential agreeing with each other is the actual signal.

3. What the Market Is Currently Pricing (and Not Pricing)

Positioning right now is unusually clean, which is itself informative. BTC funding near flat with ~~107,048 BTC (~~$6.9B) of open interest on Binance alone reads as two-sided flow, not a leveraged-long pile waiting to unwind. Nobody is paying a real premium for upside on either asset. That's a defensive tape, not a euphoric or a capitulating one.

That backdrop cuts two ways for this trade. It means a naked directional long (BTC or ETH) here is really a bet that fearful, low-flow positioning resolves upward, with the Fed sitting at 3.50%-3.75% "higher for longer" as the macro judge of that bet. That's a beta call this note isn't making. A relative-value pair strips most of that out: broad macro moves wash through both legs roughly symmetrically, except through the differential beta between ETH and BTC, which is precisely the trade's real risk (Section 7), not something it avoids entirely.

What the market does not appear to be pricing at all is continued, quiet ETH outperformance specifically. The tape's attention is on whether BTC's range holds or breaks, not on the cross.

4. Entry and Structure

Long spot ETH / short BTCUSDT perp: not perp-perp, and not spot-spot.

The reasoning is carry, not preference. A long ETH perp pays roughly 5% annualized funding right now; a short BTC perp receives roughly 1%. Run this as perp-perp and funding alone costs about 4%/year of carry, against a trade whose entire observed 30-day edge so far is 2.4%. The carry would eat the thesis on anything but a fast move. Structuring as spot ETH against a short BTC perp flips that to roughly +1%/year of positive carry, and removes the position's exposure to ETH funding ratcheting higher if the long side gets crowded later. (Spot-spot is cleaner in principle but adds BTC borrow friction most retail/prosumer setups don't have cheap access to. Perp-spot is the practical version of the same idea.)

Entry: ~50% at market near 0.0296; ~50% resting at 0.0290–0.0294.
Stop: daily close below 0.0286 (see Section 5 for why this level specifically).
Target 1: 0.0305 (partial trim). Target 2 / full exit: 0.0310–0.0311 (the 30-day ratio high).
Sizing: risk to the stop from a ~0.0293 blended entry is roughly 2.5-4%. A 50bp NAV risk budget implies roughly 12-15% of NAV per leg (~25-30% gross). Start with half of that and add only on confirmation (Section 6).

5. Invalidation

Hard stop: a daily close below 0.0286. That sits below the 30-day drift's origin point (0.0289) with a small buffer. That level isn't just "down a lot," it specifically means the entire month's ETH/BTC divergence has round-tripped and the rotation was noise, not signal. That makes this stop a statement about the thesis being wrong, not merely about the trade being unlucky.

Two conditions to exit before the hard stop is even touched:

A funding-based thesis stop, independent of price: if ETH funding ratchets above roughly 15-20% annualized while the ratio stalls rather than extending, the "uncrowded" premise that makes this trade attractive is dead. Exit regardless of where P&L sits at that moment.
A funding-reversal early warning: if ETH funding drops below BTC funding while the ratio is trading at or below ~0.0290, the marginal bid for ETH that the whole thesis rests on has disappeared. That's a cue to cut aggressively rather than wait for the hard stop to be hit.
A time stop: no new 30-day ratio high within four weeks means the drift isn't persisting. Flatten and re-underwrite rather than pay opportunity cost holding a pair that's gone quiet.

6. What Would Make This Much Higher Conviction

A daily close above 0.0305-0.0311 on spot-led ETH volume, with funding still holding under ~10% annualized: a breakout without leverage froth behind it. That would justify sizing up to roughly 2x, and extending the target toward 0.0320-0.0325.
Confirmed ETF flow composition showing ETH products taking inflows while BTC products are flat or bleeding: actual rotation evidence at the issuer level, rather than an inference drawn from price and funding alone.
Primary-source open interest, split by asset, across venues beyond Binance (OKX, Bybit, dYdX): rising ETH-side OI with funding still mild would mean spot-led accumulation, not leveraged chasing.

Any one of these moves the trade from a 2/5 to roughly a 3/5 and justifies the larger size band; two of them together would justify going further.

7. What Actually Kills This Trade (Including the Data Behind It)

In order of what genuinely worries me about this position:

The signal itself is modest. A 2.4% 30-day ratio drift and a ~4-point funding spread is real, but it is not a screaming edge. It can simply be noise that reverts. That's the reason the size is small; no amount of clever structuring changes that fact, and it shouldn't be dressed up as more than it is.
Beta-gap risk in the pair itself. A pairs trade "feels" hedged, but the ETH/BTC ratio has a fat tail on real risk-off shocks: in a genuine macro scare (a hawkish surprise from a Fed already at 3.50%-3.75% counts), ETH historically underperforms BTC fast, and the ratio can gap 5-8% in a matter of days. The 30-day ratio low (~0.0277) sits uncomfortably close to the stop level; a gap straight through the stop is a real possibility, not a tail-risk footnote. This is the main reason size stays capped at 50bp of NAV rather than something larger.
Two of the inputs behind this note are secondhand and unverified. The ~$250M Aug 12-14 ETF outflow figure and the "market-wide futures OI near a two-month high" claim both came from aggregated news search, not from issuer daily reports or exchange APIs directly. The BTC/ETH prices, 30-day ranges, funding prints, and Binance BTC open interest figure were pulled directly from primary sources and should be trusted more. Before adding size to this position, those two secondhand figures need confirming against issuer-level ETF flow data and cross-venue OI/funding (not just Binance). Sizing a real position off aggregated-search data points is a good way to become someone else's exit liquidity. If the ETF outflow turns out to be ETH-led rather than BTC-led, or if OI is actually building aggressively in ETH longs on venues this note didn't check, the "positioning isn't crowded" premise weakens or breaks.
Carry flip. If BTC funding goes negative, the short-BTC leg's +1%/year carry becomes a small cost instead. Minor on its own, but worth monitoring alongside the funding-reversal warning in Section 5.

Bottom line: a small, defined-risk drift trade on a divergence that was verified directly rather than assumed, in a tape with unusually clean (non-crowded) positioning on both sides. It targets roughly 1.2:1 to 1.9:1 reward-to-risk depending on fill, and costs about 50bp of NAV if wrong. Nobody gets rich on this note. The point of including it in a portfolio isn't the return, it's demonstrating the process: verified data over secondhand claims, honest sizing of a modest signal, carry-aware structuring, and a hard invalidation defined before the trade rather than argued after the fact.

Appendix: Where the Panel Agreed and Where It Added Distinct Value

All three models converged, independently, on the same trade: long ETH/BTC via spot ETH against a short BTC perp, in the same 0.0296 entry zone, with invalidation clustered tightly at 0.0284-0.0288 and a first target around 0.0305-0.0311. That's a meaningfully tighter convergence than the EigenLayer piece's directional agreement. Here even the specific price levels lined up within a few basis points of each other, which is reassuring (the underlying data supports one clear reading). As with the first piece, this convergence is treated as supporting evidence rather than proof. Three models reasoning from the same numbers with similar trading-desk conventions will tend to converge on the textbook answer.

Distinct contributions folded into the synthesis above:

DeepSeek was tightest and most execution-focused: the cleanest statement of the tranched entry/add logic and the clearest one-line articulation of why BTC's range plus ETH's position near its own range high lets the ratio work without needing broad crypto beta to cooperate.
Kimi built the most rigorous risk framework: an explicit conviction score (2/5) tied to a specific risk budget, a funding-based "thesis stop" that's independent of price entirely, a time stop, and the sharpest articulation of gap risk on the ratio's fat tail during beta shocks. The reasoning behind capping size at 50bp rather than something larger is Kimi's framing, adopted directly.
Qwen contributed the funding-reversal early-warning condition (exit if ETH funding drops below BTC funding even before the hard stop is touched) and was the most thorough on listing every confirmation that would justify sizing up, plus the clearest statement of the base-rate risk that BTC tends to be "the cleanest shirt" in a liquidity-driven crypto selloff, capping how much protection the short-BTC leg actually provides against an ETH-specific de-rating.

No disagreement surfaced on the trade direction, structure, or core risk. The only real variance was precision: stop levels ranged from 0.0284 to 0.0288 across the three, and this note uses 0.0286 as the middle of that range with a stated rationale (just under the 30-day drift's origin) rather than picking whichever model's number happened to be most convenient.