Support for season 1.5 of @the_nof1 AI trading competition is now live 🤖
These models are exclusively trading US equities which you can copy trade or track with SuperX
The first SuperX Trading competition is now over! 🏆
Stats include:
• 165 traders participated
• $30M in total trading volume
Congratulations to all 20 winners - rewards will be distributed over the next 72 hours
$15M trading volume & 120 entries one week into the SuperX Trading Competition
With 6 days left, there's still plenty of prizes and rewards left up for grabs
1/🧵
$200M in total trading volume on SuperX 🏆
As trading increasingly moves on-chain, we’re proud to be one of the few teams building the tools that onboards new users.
Thank you to all our users for your support so far.
Jobs not finished.
The competition will run between the 3rd November to 17th November
Players will have the opportunity to compete in two tracks, based on the best ROE, and the highest trading volume
Rewards and prizes will be given to the 10 top traders in either track, creating a total of 20 guaranteed winners
Bonus rewards in the form of USDC & Hyperliquid NFTs will be raffled during the competition window and airdropped to winners
You won't want to miss this.
Sign up here --> https://t.co/O45A2qjQTS
$30K SuperX Trading Competition is back 🏆
20 guaranteed winners, exclusive bonus prizes, NFT drops, USDC raffles, and plenty more rewards up for grabs
Registration details below
We analysed how the AI models in the Nof1 Alpha Arena (@the_nof1) competition traded on Hyperliquid using data from SuperX
Here are some of the interesting insights we pulled:
1. Specialisation. The top models focused on a clear directional bias. DeepSeek was 96% long, Qwen 73% long and rarely flipped back & forth. The worst performers had the greatest bias towards going short.
2. Selectivity. The best models traded less and sized properly. DeepSeek traded only around 1.5 times per day compared to Gemini’s 18 times and still achieved the highest RoE. Lower frequency, higher quality trades won.
3. Holding time. The strongest returns came from trades held 1–2 days. DeepSeek’s average hold was around 48 hours, long enough to capture trend, short enough to avoid chop. Gemini’s was the lowest at just 7 hours.
4. Trend alignment. Every profitable model was net long while BTC and ETH trended up. The worst performers flipped short mid trend.
5. Consistency. DeepSeek’s equity curve shows steady compounding without big spikes. Grok and Claude started strong but lost momentum, suggesting their strategies couldn't adapt to changing market conditions.
Overall, the best performing models didn’t overly trade, they picked a direction (long) and held on to capture the returns
We just added support for @the_nof1 AI trading competition on SuperX 🤖
Track or copy trade the best performing AI models trading autonomously on Hyperliquid
Available on Web & Telegram
HIP-3 Markets are now live on SuperX
Trade XYZ100, an index of the top 100 U.S. non-financial companies directly from the terminal
24/7 markets on-chain, powered by Hyperliquid