Skip to content

RohOnChain ranking book

Paper bot · Retired

Forward test of RohOnChain's ranking book on his "Coin 30": each day, long the 5 strongest of the 30 most-traded crypto perps and short the 5 weakest.

Historical record unavailable

Live record lost in the 24 Oct 2026 server move.

Simulated paper trading only — orders are not sent to Hyperliquid.

Ranking book

Last rebalance none yet

Next rebalance Tue 17 Nov, 00:15 UTC (in 15h)

Universe Coin 30: the 30 Hyperliquid perps with the highest 24h volume (at least $2M, 40+ daily candles), picked on the first rebalance of each UTC month. The first list is picked at the next rebalance.

Today's legs

No open legs yet. The first book opens at the next rebalance.

Equal weight, 80% of equity in total, long the top 5 and short the bottom 5.

Ranking

No ranking yet; the first one is made at the next rebalance.

Score = momentum z (average of the 7- and 30-day returns) + MACD z (histogram ÷ price), each ranked across the coins on yesterday's close.

Drawdown vs the halt

Drawdown —

At 20% below its peak the bot closes everything and stays flat until reviewed by hand.

Backtest (2020-2026, Binance daily closes, 41 coins): Sharpe 1.01 since Aug 2022, 1.06 before; positive every calendar year (1.8, 0.6, 0.5, 0.6, 2.1, 0.8, 1.2); beta to BTC about 0.0; max drawdown 37% at 1× gross. The 24 most liquid coins (2-12 bps slippage): 0.89 before, 1.17 since. The coins are today's survivors and these are backtests: this is a forward test. Checked point-in-time on 2026-10-09 (each month's 24 most-traded Binance perps, incl. coins later delisted): the same daily book scored Sharpe 0.2-0.3 with 50-70% drawdowns, so the survivor backtest above flatters it.

In plain words

TL;DR

Forward test of RohOnChain's ranking book on his "Coin 30": each day, long the 5 strongest of the 30 most-traded crypto perps and short the 5 weakest. Solid backtest, but on today's surviving coins.

How it bets

RohOnChain's article ends with an example strategy: "A ranking idea, cross-sectional, on Coin 30. Combine short-term momentum and price-normalized MACD. Long the top 5, short the bottom 5, rebalance daily, equal weight." We coded it exactly that way, with no AI model. Coin 30: on the first rebalance of each month the bot picks the 30 Hyperliquid perps with the highest 24-hour trading volume (at least $2M, with 40+ days of price history) and keeps that list for the month. A coin that is delisted or trades under $2M on a day sits that day out. Every day at 00:15 UTC the bot scores those coins on yesterday's close: momentum (average of the 7-day and 30-day returns) plus the MACD histogram divided by the price, each ranked against the other coins. It buys the top 5 and sells short the bottom 5, equal weight, 80% of its equity in total, at the live order book with fees and slippage (wider for less-traded coins). Hourly perp funding is booked: longs usually pay it, shorts receive it. Safety: a 20% fall from the equity peak closes everything until we review it by hand; a short 40% against us is closed at once. Our backtest on 2020–2026 daily prices of 41 coins (fees, slippage, real funding): Sharpe 1.06 to mid-2022 and 1.01 since, positive every calendar year, almost no link to Bitcoin's moves, worst drawdown 37% at full size. On the 24 most liquid coins: 0.89 and 1.17. Those coins are the ones still trading today (dead coins are missing), so the live record is the real test. Credit: RohOnChain's article (the strategy, Coin 30 and the drawdown halt).

Example

Say SOL, HYPE, SUI, ENA and TAO rank top and DOGE, PEPE, WLD, ARB and UNI rank bottom with $1,000 equity: the bot holds about $80 long in each of the top 5 and $80 short in each of the bottom 5. Next day, if UNI moves up the ranking and NEAR drops to the bottom 5, it closes the UNI short and opens a NEAR short; the other legs stay unless they drift more than 25% from $80.

Why it might work

Crypto coins that have been rising relative to the others tend to keep doing so for a while, and holding both sides cancels most of the market's moves.

Why it might fail

The backtest only includes coins that survived to today, momentum can reverse sharply (the worst backtest drawdown was 37%), and small coins cost more to trade than the model assumes on a bad day.

Exact rules

No new bets since 9 Oct 2026 (retired); open legs were closed at market. The rules it traded by: Paper bot on Hyperliquid perpetuals, coding the example strategy from RohOnChain's article (https://x.com/RohOnChain/status/2107469298382090680): "A ranking idea, cross-sectional, on Coin 30. Combine short-term momentum and price-normalized MACD. Long the top 5, short the bottom 5, rebalance daily, equal weight." No AI model is involved. Universe (Coin 30): the 30 Hyperliquid perps with the highest 24-hour trading volume, at least $2M and at least 40 closed daily candles, picked on the first rebalance of each UTC month and fixed until the next month. On a day a listed coin is delisted, trades under $2M in 24 hours or lacks 40 closed daily candles up to yesterday, it sits out (its leg is closed); with fewer than 8 usable coins the bot holds nothing. Every day at 00:15 UTC it reads Hyperliquid's daily candles, closed days only. Momentum = the average of the 7-day and 30-day returns; MACD = (12-day EMA − 26-day EMA) ÷ price, minus its 9-day EMA (the histogram). Each is turned into a z-score across the coins and the two are added. It goes long the top 5 and short the bottom 5 (with fewer than 30 usable coins, n ÷ 6 rounded per side), equal weight, 80% of equity gross (40% long, 40% short). Rebalance: legs that dropped out are closed, new ones opened, and a kept leg is resized only when it is more than 25% off its target; at most 60 fills per rebalance. Fills use the live Hyperliquid order book, never better than mid ± 2 bps (BTC, ETH), 4 bps (SOL, XRP, DOGE) or, for other coins, by 24-hour volume: 4 bps (≥ $100M), 6 bps (≥ $20M), 8 bps (≥ $5M), 12 bps (less), plus the 4.5 bps taker fee. Hyperliquid funding is booked every hour at the current rate (longs pay a positive rate, shorts receive it). Risk: if equity falls 20% from its peak, everything is closed and the bot stays flat until we review it by hand (Roh's drawdown halt); a short more than 40% against us on the hourly mark is closed. Our backtest (Binance daily closes 2020 to Oct 2026, the same score, 4.5 bps taker fee plus slippage, real Binance funding): on 41 coins, Sharpe 1.06 from Jul 2020 to Jul 2022 and 1.01 from Aug 2022 to Oct 2026, positive in every calendar year (1.8, 0.6, 0.5, 0.6, 2.1, 0.8, 1.2 for 2020–2026), beta to BTC about 0, maximum drawdown 37% at 1× gross (hence 0.8× here); on the 24 most liquid coins with 2–12 bps slippage, 0.89 and 1.17. The coins are today's survivors (coins that died are missing) and these are backtests, so this is a forward test, not proof.

Source: x.com ↗

Bets

No published bets yet: this record has no settled, closed or open bets.