Hyper — Hyperliquid Regime Book

• Size — $175k notional on $7.5k equity is 23× effective leverage. The exchange shows 25× on the position; the account has nothing else, so that is also the whole account.
• Distance — 2.2% to liquidation. ETH has moved 2.2% or more in a day on 37% of 2026 sessions, and UP 2.2% or more on 16% — roughly one day in six.
• Margin — $7,016 of $7,537 is committed; withdrawable is $0. There is no collateral for any other trade.
• Direction — The direction is fine — ETH is a war-sensitive asset (−0.24% per 1% oil, t −3.0) and has no rate beta. Only the size is wrong.

This is the single most important number on the page and the reason it leads. The regime book below is a set of moderately-sized positions that pay out over weeks; a 23× position pays out or dies in hours. If ETH squeezes 2.2% on a Tuesday morning — a Hormuz de-escalation headline, a Fed speaker, a short-squeeze in a market where the account is one of the larger shorts — the whole account is gone and every card below is a reading exercise.

The account’s own history says this is not its game. Its ETH record is 2,812 fills with 1,698 winners to 20 losers on closing fills — a scalper’s record, tiny profits taken hundreds of times, +$1,002 in total. It has paid $4,645 of funding on ETH all-time, more than on any other coin. A scalping edge and a 23× swing position are two different businesses, and the second one is being financed by the first.

The fix is mechanical: cut to 10 ETH. At 10 ETH ($24k, 3.2× equity) with ≈$2,000 of collateral left behind it in the main account, the liquidation price moves to ≈2,560 — a 6% move instead of 2.2% — and ≈$6,050 of margin is freed. Move ≈$5,500 of that to the HIP-3 (xyz) margin account, which is a separate collateral pool where nine of the ten tickets below trade. The residual 10 ETH stays as the book’s crypto-war leg (card #5).

• Rates — −2.82% per +10 bp on the 10-year, t −2.2 (135 trading days since listing). In the page’s scenario (+20 bp) that is −5.6% on its own.
• War — −0.44% per +1% in WTI, t −3.2. Korea imports essentially all of its crude; a Hormuz premium is a direct cost to the Korean trade balance and the won. Oil +10% → another −4.4%.
• Carry — Funding +36%/yr. On Hyperliquid positive funding is paid by longs to shorts, so a $3,000 short collects ≈$90 a month simply for being short.
• Book — $288M open interest, $224M daily volume — the deepest stock perp on Hyperliquid after SP500 and GOLD. The ticket is invisible to it.
• Trend — −16.1% over 90 days while up 4.7% over 30 — a bear-market bounce inside a downtrend, which is the entry the short wants.

Why a memory stock is a rates trade. SK Hynix is a memory maker whose value is almost entirely 2027–28 earnings from the AI-capex cycle — HBM for Nvidia, DRAM for the datacenter buildout. That makes it one of the longest-duration equities on the board: a small change in the discount rate moves a large fraction of its value, and a 10-year at 4.8% is precisely the discount-rate shock. The 2022 analogue is the cleanest: when the 10-year went from 1.5% to 4.2%, the KOSPI fell about a quarter and SK Hynix roughly 40% peak-to-trough. The data this year shows the same reflex at t −2.2.

Why the war doubles it. Korea is one of the five largest crude importers in the world and buys most of it from the Gulf. A Hormuz risk premium widens Korea’s trade deficit, weakens the won, and pulls foreign money out of the KOSPI — which is why the three liquid Korean instruments on the board (SKHX, Samsung, EWY) all carry significant negative oil betas — larger, per 1% of oil, than any US megacap’s. The combined scenario estimate is −10.0%, the third-largest on the board (after DRAM and RKLB) and the largest on a book over $100M.

Why this ticker and not Samsung or the ETF. Samsung (SMSN, −3.2%/10 bp) and the Korea ETF (EWY, −2.7) are the same trade. SKHX wins on two counts: its book is six times deeper than Samsung’s and fifteen times EWY’s, and its funding is +36%/yr against +5% on the other two — the market is paying a premium to be long SK Hynix specifically, which is what a crowded AI-memory long looks like. The book takes SKHX for depth and carry and EWY (#2) for the country expression, and does not add Samsung on top.

What would make it wrong. A memory-price upcycle strong enough to overpower the discount rate (the 2026 bounce was +4.7% in 30 days, so the bid exists), or the 10-year falling back through 4.55% — at which point the regime input is gone and the short is just a fade of a strong sector. The stop is set for the first; the invalidation rule handles the second.

• Rates — −2.72% per +10 bp, t −3.2, on 128 trading days since March. Scenario: −5.4% from rates alone.
• War — −0.32% per +1% oil, t −3.6. Scenario: −3.2% from oil. Combined estimate −8.6%.
• Fit — R² 25% — for a single country on daily data, unusually tight. This is not a story fitted to noise; it is a market that mechanically loses on both shocks.
• Book — $18.7M open interest, $32M daily volume. Funding +5%/yr, paid to the short.
• Trend — −11.3% over 90 days; +7.7% over 30 — the same bounce-inside-a-downtrend as SKHX.

The mechanism is the currency. A rising US 10-year pulls capital toward dollars and widens the rate gap against the Bank of Korea; the won weakens; foreign holders of Korean equities lose on the currency and sell the equities. At the same time, oil at $95 is a tax on an economy that imports its energy and exports manufactured goods to customers who are also paying more for energy. EWY sits at the intersection of both flows, which is why its two betas are the most consistent pair on the board outside the S&P itself.

Why EWY rather than KR200. KR200 (the KOSPI 200 perp) has a rate beta of −3.4 and looks like the stronger expression. It has $0.6M of open interest. EWY has $18.7M. A $3,000 order into a $600k book is a market-moving event; into an $18.7M book it is a rounding error. The page prices the trade it can execute, and lists KR200 as “right idea, wrong book” in the screen.

How it fits with #1. SKHX and EWY are one position wearing two tickers — SK Hynix is one of the two largest weights in the KOSPI, and their daily returns are highly correlated. The book holds both because they diversify the idiosyncratic risk (a memory-price headline hits SKHX, a Korean political headline hits EWY) while keeping the same macro exposure; it counts them as one $6,000 Korea position for risk purposes and sizes them accordingly.

• Rates — −4.33% per +10 bp, t −2.1, on 83 trading days since May. Scenario −8.7% from rates alone; −10.4% combined, the worst on the board.
• War — Oil beta −0.17, not significant — this is a pure rates/duration short, not a war short.
• Book — $82M open interest, $82M daily volume. Funding +5%/yr, paid to the short.
• Trend — −13.2% over 90 days, +5.7% over 30. 81% annualised vol.
• The account — SNDK −$4,201, MU, SKHX — 1,602 + 1,951 + 355 fills, $0.81M bought long against $0.01M short. Same sector, wrong side, all summer.

Why memory has the longest duration on the board. DRAM and NAND pricing is a two-to-three-year cycle, and the stocks are priced off the peak of the next cycle, not this quarter’s earnings. The market’s 2026 story is that AI datacenter demand makes the next peak higher and later; every dollar of that value is discounted from 2027–28 back to today, which is why a 20 bp move in the 10-year is worth almost 9% of the index. Rate-sensitive growth stocks usually have this property; memory has it and cyclicality on top.

Why #3 rather than #1. The history is short (83 days) and the coefficient overlaps heavily with #1 — the DRAM index and SK Hynix trade the same memory-price cycle (the index’s exact composition is not published in the API; the page treats it as memory-sector beta). The page does not want to triple-count the same exposure, so DRAM is sized at $2,000 rather than $3,000 and the three Korea/memory legs are risk-managed as one position.

Why it is on the page at all. The Trade Review shows the account bought this complex all summer and lost on it — SNDK is the single worst instrument in the ledger, and the open-long notional across SNDK, MU and SKHX was 80 times the open-short notional. The regime book says the account was on the wrong side of its favourite sector, and this ticket is the explicit reversal: do not hold this short and a long in SNDK, MU or SKHX at the same time.

• Rates — −0.55% per +10 bp, t −3.4, 116 trading days. Scenario −1.1% from rates.
• War — −0.10% per +1% oil, t −5.2 — the highest oil t-stat on the board outside energy and Japan. Scenario −1.0% from oil.
• Fit — R² 38% — the index is the best-explained instrument on the board after energy.
• Book — $382M open interest, $229M daily volume, 50× max leverage. Funding +5%/yr, paid to the short.
• Vol — 11% annualised — a −2.1% scenario move is a real move at this vol, and the low vol is what lets the leg run 10×.

Why a small beta is still the biggest ticket. The single-name shorts above have betas of −3 to −4 per 10 bp but 80–100% vol and books of $20–300M. The index has a beta of −0.55 with 11% vol and a book that dwarfs every other rate loser (1.3× SKHX, 20× EWY). Per unit of risk the index is the most efficient way to own “rates up, oil up, equities down” — which is why the book sizes it at $6,000 and levers it 10×, so that its scenario P&L is comparable to the single names’ while its path risk is a fraction.

The transmission argument. The Treasuries tab on this site prices a 6% long bond against the S&P at 25× forward earnings: a 4.0% earnings yield versus a 5.4–5.6% ten-year in that world (further than this page’s +20 bp scenario) is an equity risk premium of roughly −1.4%, not seen since 2000, and partial normalisation alone maps to SPX 5,300–6,100. This ticket does not need that world — it needs the next 20 bp — but it is the instrument that pays without single-name risk if that world arrives.

Why it is called a hedge. The book’s other shorts are Korean and memory names that can rally on their own news while the index does nothing. The index short is what makes the book pay if the regime plays out as “everything with duration drifts lower” rather than “Korea specifically breaks”. It has done nothing for 90 days (+0.9%) while the 10-year rose 30 bp — earnings have been fighting the beta — so it is sized as insurance on the book, not as a view.

• Rates — BTC +0.45% per 10 bp (t 0.8, not significant); ETH 0.00; SOL +0.21. The “long-duration asset that dies when rates rise” story is not in the 2026 data.
• War — BTC −0.21% per 1% oil (t −3.6), ETH −0.24 (t −3.0), SOL −0.25 (t −3.2), LINK −0.24 (t −3.1), ZEC −0.48 (t −3.1). Every major sells off on the days crude spikes.
• Carry — BTC/ETH funding +11%/yr paid to the short; SOL 0.
• Position — After card #0: 10 ETH short, ≈$24k — already 4× what the book would choose on its own.

What the data says and does not say. The page expected crypto to be the obvious rates short. It is not: the rate betas are small, mixed in sign and statistically zero. What crypto is, in 2026, is a risk asset that trades off on war headlines — the oil beta is negative and significant for every major, which fits a market that de-risks when Hormuz is in the news and re-risks when it is not. So the honest call is a war short, and a war short is what the account already has.

Why no new crypto tickets. The residual 10 ETH after card #0 is ≈$24k of notional, four times the $6,000 the book would allocate to a single leg. Adding BTC or SOL shorts on top would push crypto to half the book’s gross on one factor. ZEC (the most oil-sensitive, −0.48, after +82% in 90 days) and SOL are the watch-list shorts for the next oil spike — a Brent close through $100 is the trigger — not tickets tonight.

The account’s own evidence. ETH is the coin the account has paid the most funding on (−$4,645 all-time) — it has historically been long ETH and paid for it. A short that collects +11%/yr is, for once, on the right side of the funding.

• War — Brent to WTI beta 0.83 at t 42 — it is the same barrel. Scenario (oil +10%) → +8.3%. Brent is up 4.5% on the day of the walk on the Iran tape.
• Rates — +0.67% per +10 bp, t 3.5. In 2026 the yield back-ups have been energy-led inflation scares, so oil and yields rise together. Scenario +1.3%.
• Carry — Funding −36%/yr: the perp trades under its oracle and shorts pay longs. A $3,000 long collects ≈$90/month. WTI (xyz:CL) prints +3% — same barrel, 39 points of annual carry difference.
• Book — $169M open interest, $199M daily volume — DEEP.
• Role — The hedge against the Korea/memory shorts being right for the wrong reason: a risk-off that takes oil down with equities.

Why Brent and not WTI. The Iran war is a Hormuz story, and Hormuz is a Brent story: Brent is the seaborne, globally traded benchmark that a Gulf supply shock reprices first, while WTI is landlocked US crude with its own inventory dynamics. On the day of the walk Brent’s premium over WTI was $4.66. More importantly for a perp, Brent’s funding is −36%/yr and WTI’s is +3%: the Brent perp is trading below its oracle price (a war-day pattern — the market is net short the perp against physical or futures), so longs are paid to provide the other side. That is 39 points of annual carry between two instruments that move together at β 0.83.

Why the rate beta matters. Most assets that hedge the war (gold, the yen) lose when rates rise. Brent gains — +0.67% per 10 bp, t 3.5 — because in 2026 the rate scares have been inflation scares with an energy component. That makes Brent the only instrument on the board with a positive coefficient on both regime inputs, which is why it is the book’s primary long rather than a diversifier.

What it protects. The book’s shorts assume rates and oil rise together. If the war escalates into a growth scare instead — oil spikes, then equities and oil both fall as demand fears take over — the Korea shorts still pay and Brent gives some back; if the war de-escalates, Brent falls and the shorts are left holding a rates-only thesis. The long is sized at $3,000 to hedge the first path without dominating the second.

What would make it wrong. A ceasefire print (Polymarket ceasefire-continues-Sep above 90¢), or the funding flipping positive above +10% — which would mean the war premium has unwound in the perp market before the spot market, and the reason to prefer Brent over WTI is gone.

• Rates — +0.21% per +10 bp, t 2.5, 165 trading days, R² 11%. Scenario +0.4%.
• War — +0.02% per +1% oil, t 2.5. Japan imports its energy; a dearer barrel is a weaker yen. Scenario +0.2%.
• Vol — 5% annualised — the calmest instrument on the board. 10× leverage gives it single-stock volatility with a documented, two-factor driver.
• Carry — Funding +3%/yr paid by the long — cheap.
• Book — $15.2M open interest, 50× max.

The macro case. A 4.8% US ten-year against a Bank of Japan that cannot hike into $95 oil is the widest rate differential of this cycle, and USDJPY is the direct expression of a rate differential. The dollar side is the page’s first input; the yen side is the second input working through Japan’s energy bill (Japan imports nearly 90% of its energy, and a Gulf war is a direct hit to its trade balance). Both point the same way, and the 2026 data confirms both at t 2.5.

Why leverage is appropriate here and nowhere else on the page. The single-name shorts have 80–100% vol; a 10× position in them is a coin flip on tomorrow’s open. USDJPY has 5% vol: a 10× position has roughly half the daily risk of an unlevered memory stock, with a much better-behaved return distribution. The book puts $6,000 of notional here — the second-largest leg — precisely because the volatility is low enough to size.

The risk that sets the stop. 160 is where Japan’s Ministry of Finance has intervened before, selling dollars in size on a Tokyo morning and moving the pair 3–4% in minutes. That is the one scenario in which a 10× yen position behaves like a 10× stock position. The stop is placed under 156.5 — the intervention zone — as a daily close, and the invalidation is the Fed: if the September meeting passes without a hike (Polymarket Sep-hike under 30¢), the rate-differential leg of the thesis is gone.

• NATGAS · war — +0.24% per +1% oil, t 4.0, 155 trading days. The LNG-arbitrage leg of a Gulf war: Qatar’s LNG transits Hormuz, and a disruption reprices gas globally.
• NATGAS · rates — −0.32 per 10 bp, not significant — rates-neutral.
• XLE · both — +1.01% per +10 bp (t 3.2) and +0.23% per +1% oil (t 4.7), R² 49% — the best two-factor long on the board (scenario +4.3%).
• XLE · book — $1.5M open interest, $0.4M daily volume — THIN. The page will not put a real ticket into a $400k-a-day book.
• Carry — Both pay: NATGAS +5%/yr, XLE +11%/yr — the wrong direction, another reason Brent (#6) leads.

Why they are on the page. They add breadth to the energy long, not conviction: NATGAS is the war trade through a different molecule (gas, not oil — the Qatar LNG route), and XLE is the war trade through equities that also happen to like higher rates (energy producers carry little duration and benefit from the inflation that drives yields). If Brent (#6) is the primary hedge, these are the diversifiers that make the energy long less dependent on one contract’s funding print.

Why they are small. Books. XLE’s coefficients are the best on the board and its open interest is $1.5M — a $1,000 order is small, but the whole book trades $400k a day and a stop would be filled against almost nothing. NATGAS has a $10M book and takes $1,000 without noticing. The rule on this page is that size follows the book, not the t-statistic, and that rule is what keeps these two at $1,000 each.

• EUR · war — −0.03% per +1% oil, t −4.4 — one of the most significant oil betas on the board outside energy. Europe imports its energy and its currency shows it.
• EUR · rates — −0.12% per +10 bp, t −1.8. Scenario −0.5%; ×10 leverage → −5%.
• EUR · carry — Funding +4%/yr, paid to the short. 6% vol, 50× max, $15.3M book.
• COPPER · rates — −0.85% per +10 bp, t −2.5, 159 trading days. The growth-scare metal: higher rates → slower construction and manufacturing → less copper. Scenario −1.7% from rates (−2.4% combined).
• COPPER · carry — Funding +5%/yr, paid to the short. 21% vol, $20.2M book. Flat over 90 days while the 10-year rose 30 bp.

What these two add. The single-name shorts pay if the regime arrives as a shock; these two pay if it arrives as a grind. A dollar that gets bid on rate differentials and an energy-importing euro that softens are slow-moving, low-vol trades — which is exactly why they can be run at 10× and 5× without the account’s survival depending on tomorrow’s open. Copper is the industrial confirmation: if rates are biting, demand for the metal that goes into every building and grid shows it before the equity index does.

Why EUR at 10×. Same logic as USDJPY (#7): a 6%-vol instrument at 10× has less daily risk than an unlevered memory stock, and the two currency legs (long USDJPY, short EURUSD) are both long the dollar — a deliberate concentration in the one asset the first input (US rates) most directly supports. The stop on EUR is 1.4% away as a daily close, which at 10× is a 14% loss on the margin posted, not on the account.

Why copper is the smaller of the two. Its rate beta is real (t −2.5) but its oil beta is not, and its 90-day return (+0.1%) says the beta has been fighting Chinese demand all summer. The ticket is $2,000 with a 7.4% stop — a position that confirms the thesis if it works and costs little if China outbids the discount rate.

• para:10Y — A perp on the 10-year yield itself, 4.814 at the walk. $0.3M open interest, $0.2M daily volume, 14 days of candles. Funding −287%/yr: the perp trades far under the oracle and shorts pay longs enormously — what a one-sided, illiquid market looks like.
• USBOND — The bond-future perp: the direct rates short. $0.5M open interest, four days of candle history, funding +4%.
• SMALL2000 — Small caps — the classic rate loser. $1.5M open interest, four days of history, funding +47%/yr paid to the short.

Why they are listed and not sized. A reader who agrees with the thesis will ask why the page does not simply buy the 10-year-yield perp. The answer is that a $2,000 order is 0.1–0.7% of the open interest in each of these markets and a large share of the day’s volume; the page would be the market, and the funding rates are telling you what happens to whoever is stuck on the wrong side of a market with no other participants.

What to do with them. A $500 long in para:10Y as a funding experiment — the carry is roughly 24% a month if it persists, and it will not, but $500 is the cost of finding out how the market behaves. Nothing in USBOND. SMALL2000 on the watch list until its open interest clears $5M; it is the one of the three that could become a real ticket, because small-cap equity is exactly the rate loser this page is looking for.

• Gold (xyz:GOLD, $4,301) — −0.53% per +10 bp (t −1.4); oil beta −0.07 (t −1.7) — the wrong sign for a war hedge. In 2026 gold has traded as a real-rates asset, not a fear asset, and 4.8% nominal / ~2.5% real caps it. NEUTRAL.
• Circle (xyz:CRCL, $89) — Screens SHORT at −5.9 (oil t −2.6, +47% in 30 days). But Circle earns Treasury yield on its stablecoin reserves — a September hike is revenue. The beta is crypto-flow; the fundamentals are the opposite. The one place the page overrides its own screen. NEUTRAL.
• BTC ($77,540) — Rate beta +0.45 per 10 bp (t 0.8) — positive and insignificant. Oil beta −0.21 (t −3.6). A war short, not a rates short; card #5 covers it.
• Microsoft (xyz:MSFT, $499) — +0.39 per 10 bp (t 1.0), oil t −1.8, +16.7% over 90 days. The megacap the regime does not hurt. Not a long (no edge); just not a short.

Why gold is not the war trade this year. The intuition is that a Gulf war bids gold. The 2026 data says the bid has been small and, if anything, negative on oil-spike days (t −1.7) — because the same days push yields up, and gold’s dominant driver this year has been the real rate. The Polymarket Treasuries tab finds the gold-$5,000 rung cheap on its own ladder; that is a cheaper way to own gold upside than paying 5%/yr of funding on a perp to hold a position the data does not support.

Why Circle gets an override. The screen is a co-movement measurement and Circle co-moves with crypto flows, so it screens short. But Circle’s business is holding reserves in Treasury bills and earning the yield; a hiking Fed raises that yield directly. When a screen and a business model point in opposite directions on a regime input, the page trusts the business model and says so.