House funding carry book
Paper bot
A house book on 24 established coins: long the ones whose futures show the fewest crowded buyers, short the most crowded, rebalanced every two weeks.
24 trades recorded · 10 settled
Simulated paper trading only — orders are not sent to Hyperliquid.
House funding carry book returned +0.35% from its first trade on 25 Oct. Buying the S&P 500 returned −1.1%, measured from its 23 Oct base close. +1.4 pts vs the S&P 500.
All recorded dates · 25 Oct – 16 Nov
1M and 3M unavailable: only 22 days recorded so far.
- +0.3%Lifetime · from $1,000.00
- +1.4 ptsLifetime · S&P base close 23 OctS&P −1.1% from that close
- $1,003.48Paper account value
- 2410 settled
- 1.8%Lifetime · worst fall from a peak
- 23Since the paper record began
House funding carry book
Rule Each of the lab's 24 established coins is ranked by its Binance funding summed over the last 7 closed days. Long the 6 least crowded (lowest funding), short the 6 most crowded (highest), equal weight, 80% of equity in total; rebalanced every 14 days, held in between. A coin under $5M of 24h Hyperliquid volume sits out that rebalance (k = a quarter of the coins left, at least 8 coins); no leg above 6.7% of equity.
Schedule Every 14 days at 00:15 UTC, held in between.
Last rebalance Sun 8 Nov, 00:15 UTC (8d ago)
Next rebalance Sun 22 Nov, 00:15 UTC (in 6d)
Open legs
| Coin | Side | Weight | Entry | Now | P&L |
|---|---|---|---|---|---|
| ETH | long | 6.7% | $2,237.90 | $1,793.45 | −$13.36 |
| WLD | long | 6.7% | $0.3336 | $0.2628 | −$14.29 |
| NEAR | short | 6.7% | $3.71 | $3.16 | +$9.93 |
| XRP | short | 6.7% | $1.16 | $0.9445 | +$12.59 |
Funding ranking for 2026-11-07
| # | Coin | Funding 7d | Book |
|---|---|---|---|
| 1 | WLD | −0.168% | long |
| 2 | ETH | 0.113% | long |
| 3 | ZEC | 0.167% | — |
| 4 | SOL | 0.176% | — |
| 5 | BTC | 0.177% | — |
| 6 | PEPE | 0.187% | — |
| 7 | UNI | 0.187% | — |
| 8 | ENA | 0.195% | — |
| 9 | XRP | 0.198% | short |
| 10 | NEAR | 0.199% | short |
Binance funding summed over the last 7 closed days; least crowded (lowest) first. Left out: AAVE (low volume), ADA (low volume), ARB (low volume), AVAX (low volume), BNB (low volume), CRV (low volume), DOGE (low volume), INJ (low volume), JUP (low volume), LINK (low volume), LTC (low volume), ONDO (low volume), SUI (low volume), TAO (low volume).
Drawdown vs the halt
Drawdown 0.6% of 35% halt
At 35% below its peak the bot closes everything and stays flat until reviewed by hand. A leg 40% against its entry is closed on the hourly mark.
Equity $1,003.40 · peak $1,009.50
Closed legs 10 · realized +$7.88 · funding +$0.77 · fees $0.70
Backtest (2020-06..2026-10, Binance USDT-M daily closes and funding, point-in-time universe (each month's 24 most-traded perps listed >= 1 year, incl. later-delisted coins)): Sharpe 0.96 since Sep 2022, 1.55 before; 22% a year; max drawdown 26%; 32% of months losing; worst month −12.9%; correlation with the BTC trend 0.03.
Risk review: nearby settings median Sharpe 0.40; bootstrap 5th percentile 0.37; top 5 months 64% of the profit. We budget for a Sharpe of about 0.4. Stop if: Sharpe below 0 after 6 months, or the profit concentrated in one or two coins.
Picked as one of 3 survivors among 18 candidates tested; the live signal is Binance funding, positions pay Hyperliquid funding; a forward test, not a proven edge.
In plain words
TL;DR
A house book on 24 established coins: long the ones whose futures show the fewest crowded buyers, short the most crowded, rebalanced every two weeks. A forward test; we budget for a modest result.
How it bets
On perpetual futures, buyers pay sellers a small fee (funding) when too many traders are long. Researchers at the Bank for International Settlements ("Crypto Carry", 2023) found that heavy funding marks crowded coins that tend to lag afterwards. This bot is our own version, built by us with no AI model. Every two weeks at 00:15 UTC it ranks the research lab's 24 established coins by the funding paid on Binance over the last 7 days. Coins with under $5M of daily trading on Hyperliquid sit out. It buys the quarter with the least crowding and sells short the quarter with the most (6 and 6 on 24 coins), equal weight, 80% of its equity in total, no leg above 6.7% of equity, at the live order book with fees and slippage. Between rebalances it holds; hourly funding is booked. Safety: a leg 40% against us is closed at once, and a 25% fall from the equity peak closes everything until we review it by hand. Our point-in-time backtest (each month's most-traded coins, including ones later delisted, real costs): Sharpe 1.55 to Sep 2022 and 0.96 since, worst drawdown 26%, and almost no link to the house trend follower (0.03). A risk review found nearby settings did much worse (median about 0.4) and five months made most of the profit, so we budget for a Sharpe of about 0.4 and will stop it if it is below 0 after 6 months or its profit comes from one or two coins. Credit: Schmeling, Schrimpf & Todorov, "Crypto Carry" (BIS working paper, 2023).
Example
With $1,000 equity, if ZEC, PEPE, TAO, ENA, WLD and SUI paid the most funding last week and BTC, ETH, LTC, BNB, ADA and XRP the least, the bot holds about $67 short in each of the first six and $67 long in each of the second six for two weeks.
Why it might work
Crowded longs pay to stay in and tend to unwind; the book also collects funding on its shorts, and holding both sides cancels most of the market's moves.
Why it might fail
The backtest edge was fragile to small changes in settings, most of the profit came in a few months, crowded coins can keep squeezing higher, and the signal reads Binance funding while the positions pay Hyperliquid's.
Exact rules
Paper bot on Hyperliquid perpetuals, built by us (no AI model) as one of the RohOnChain desk's house strategies, beside the house trend follower. Idea: Schmeling, Schrimpf & Todorov, "Crypto Carry" (BIS working paper, 2023): perp funding shows how crowded a coin's longs are, and crowded coins tend to do worse afterwards. Universe: the research lab's 24 established coins (BTC, ETH, SOL, XRP, DOGE, AVAX, NEAR, UNI, LINK, LTC, ADA, AAVE, BNB, ARB, SUI, INJ, ENA, ZEC, CRV, PEPE, WLD, JUP, ONDO, TAO), not the most-traded list, because shorting freshly listed meme coins can lose more than 100% in a squeeze. At a rebalance a coin with less than $5M of 24-hour Hyperliquid volume sits out. Each coin is ranked by its Binance funding summed over the last 7 closed days (the lab's own daily data, closed days only). It goes long the quarter with the lowest funding and short the quarter with the highest (6 and 6 on 24 coins; with fewer than 8 usable coins it holds nothing), equal weight, 80% of equity in total, and no leg is ever above 6.7% of equity. The first decision is made at 00:15 UTC after the bot starts; the book is rebalanced every 14 days from then and held in between (a rebalance missed for lack of data is made up on the next day, then the 14-day schedule continues unchanged). A kept leg is resized only when it is more than 25% off target. Fills use the live Hyperliquid order book, never better than mid ± the coin's slippage (2–12 bps), plus the 4.5 bps taker fee; a coin whose live price is more than 25% from the last daily close is not traded that day. Hyperliquid funding is booked every hour (longs pay a positive rate, shorts receive it); the signal reads Binance funding, the positions pay Hyperliquid's. Risk: a short more than 40% above its entry or a long more than 40% below its entry is closed on the hourly mark; if equity falls 35% from its peak (above the backtest's own 26% maximum drawdown, below a 2x crisis), everything is closed and the bot stays flat until we review it by hand. Backtest (Binance USDT-M daily closes and funding, point-in-time universe: each month's 24 most-traded perps listed at least a year, including coins later delisted; real costs): Sharpe 1.55 from Jun 2020 to Sep 2022 and 0.96 from Sep 2022 to Oct 2026, 22% a year, maximum drawdown 26%, 32% of months losing, worst month −12.9%; 14-day funding with weekly rebalancing 0.99; on the lab's fixed 24 coins about 1.1; daily correlation with the BTC trend follower 0.03. An independent risk review found that this exact setting flatters it: across nearby settings the median Sharpe was about 0.40, a block-bootstrap 5th percentile 0.37, and the 5 best months made about 64% of the profit. So we budget for a Sharpe of about 0.4. Kill criteria: Sharpe below 0 after 6 months, or the profit concentrated in one or two coins. It was picked as one of 3 survivors among 18 strategies tested, so this is a forward test, not a proven edge.
Source: bis.org ↗
Bets
Signed amounts show gains and losses. Open amounts are unrealized; settled results retain their recorded status.
14 settled and open bets on this record.
−$14.11
Unrealized
Open · Now worth $53.20
@ $0.3336 · now $0.2638 · −20.9% · $67.28 + $0.03 fee
−$13.34
Unrealized
Open · Now worth $53.97
@ $2237.90 · now $1795.05 · −19.8% · $67.28 + $0.03 fee
+$12.57
Unrealized
Open · Now worth $79.88
@ $1.16 · now $0.9443 · +18.7% · $67.28 + $0.03 fee
+$9.44
Unrealized
Open · Now worth $76.75
@ $3.71 · now $3.18 · +14.1% · $67.28 + $0.03 fee
Closed −$19.37
@ $0.4690 · $66.67 + $0.05 fee · exit @ $0.3331
Closed +$2.94
@ $6.87 · $66.67 + $0.06 fee · exit @ $7.18
Closed −$4.11
@ $740.09 · $66.67 + $0.06 fee · exit @ $695.12
Closed −$1.10
@ $0.1702 · $66.67 + $0.06 fee · exit @ $0.1676
Closed −$3.36
@ $137.98 · $66.67 + $0.06 fee · exit @ $131.15
Closed −$3.01
@ $1203.93 · $66.67 + $0.06 fee · exit @ $1257.10
Closed +$3.29
@ $2356.18 · $66.67 + $0.06 fee · exit @ $2237.90
Closed +$12.64
@ $0.1950 · $66.67 + $0.05 fee · exit @ $0.1579
Closed +$15.95
@ $0.3521 · $66.67 + $0.05 fee · exit @ $0.2676
Closed +$4.01
@ $80745.35 · $66.67 + $0.06 fee · exit @ $75814.66
Show raw fills · 24 recordedHide raw fills · 24 recorded
All 24 recorded fills available on this page are shown (up to 100).
HL WLD long
$67.28
notional
long at 0.334 · 201.65237996858204 shares · $0.03 fee
HL ETH long
$67.28
notional
long at 2237.897 · 0.03006332242378729 shares · $0.03 fee
HL XRP short
$67.28
notional
short at 1.162 · 57.90472406388828 shares · $0.03 fee
HL NEAR short
$67.28
notional
short at 3.705 · 18.158504898006452 shares · $0.03 fee
HL ZEC Close short (rebalance)
$69.61
notional
short at 1257.104 · 0.05537433335912947 shares · $0.03 fee
HL WLD Close long (resize)
$47.34
notional
long at 0.333 · 142.13127639813084 shares · $0.02 fee
HL UNI Close long (rebalance)
$69.66
notional
long at 7.179 · 9.70399957067177 shares · $0.03 fee
HL ETH Close short (rebalance)
$63.32
notional
short at 2237.897 · 0.028294402082277858 shares · $0.03 fee
HL ENA Close short (rebalance)
$53.97
notional
short at 0.158 · 341.81389501204336 shares · $0.02 fee
HL CRV Close short (rebalance)
$50.66
notional
short at 0.268 · 189.3390902262395 shares · $0.02 fee
HL BTC Close short (rebalance)
$62.60
notional
short at 75814.660 · 0.0008256409634194525 shares · $0.03 fee
HL BNB Close long (rebalance)
$62.62
notional
long at 695.120 · 0.09007888549401687 shares · $0.03 fee
HL ARB Close long (rebalance)
$65.62
notional
long at 0.168 · 391.58318416657204 shares · $0.03 fee
HL AAVE Close long (rebalance)
$63.37
notional
long at 131.147 · 0.4831604872590411 shares · $0.03 fee
HL WLD long
$66.67
notional
long at 0.469 · 142.13127639813084 shares · $0.03 fee
HL UNI long
$66.67
notional
long at 6.870 · 9.70399957067177 shares · $0.03 fee
HL BNB long
$66.67
notional
long at 740.092 · 0.09007888549401687 shares · $0.03 fee
HL ARB long
$66.67
notional
long at 0.170 · 391.58318416657204 shares · $0.03 fee
HL AAVE long
$66.67
notional
long at 137.980 · 0.4831604872590411 shares · $0.03 fee
HL ZEC short
$66.67
notional
short at 1203.927 · 0.05537433335912947 shares · $0.03 fee
HL ETH short
$66.67
notional
short at 2356.179 · 0.028294402082277858 shares · $0.03 fee
HL ENA short
$66.67
notional
short at 0.195 · 341.81389501204336 shares · $0.03 fee
HL CRV short
$66.67
notional
short at 0.352 · 189.3390902262395 shares · $0.03 fee
HL BTC short
$66.67
notional
short at 80745.348 · 0.0008256409634194525 shares · $0.03 fee
Last hourly run
Considered — · Trades 0 · Skipped —
Skips are normal when the market fails the bot's published filter, the price is outside its band, the book is empty, or the full stake cannot fill.