AI vs 1,076 Human Traders

Jay Sharma · Founder, Botsfolio

Published October 16, 2026

When an AI lab releases a new model, it publishes benchmark scores. Math tests, coding tests, reasoning tests, each one a number you can compare with other models and with people.

Trading AI almost never gets a test like that. You get a screenshot of a good month, a backtest with no drops in it, or a promise of 2% a day.

So we built a benchmark for trading. The opponent is real people trading real money, and every one of them gets matched with Botsfolio AI on the same money and the same dates. Here are the scores.

The test

A comparison is only fair if both sides play the same game, so every matchup follows the same rules.

• The human traders. We picked 2,500 accounts at random from the 32,839 that have traded at least $50,000 on Hyperliquid, a crypto exchange where every trade is public on the blockchain. The data comes from its public leaderboard and public API. We removed market makers and bots, meaning accounts that trade 150 or more times a day or around the clock, and accounts where money moved in and out too often to measure. That left 1,076 human traders.
• The window. Each trader is measured over their own active period, from their first trade to their last, between May 2024 and October 2026. The typical window was about a year.
• Botsfolio AI. For every trader, Botsfolio AI, running as the BTC & ETH Cycle Trader, starts with the same money on the trader's first day and runs to their last.
• Fees. The traders' results are after their own Hyperliquid fees and funding. They paid a median of 0.034% per trade, measured from their actual trades (Hyperliquid's fee schedule). Botsfolio AI was charged 0.10% on every trade, about three times as much. At Hyperliquid's own spot fees, its results improve by less than a point.
• The scores. Return is the gain or loss as a share of the most money the trader ever had in the account. Worst drop is the biggest fall from a high. Made money means the window ended above the starting money.

One question comes up first, so here's the answer up front: these are trading gains and losses, not deposits. The profit figure is the exchange's own trading profit and loss, from closed trades, fees and funding. Money added to or taken out of the account doesn't count. We checked it by rebuilding the profit from the actual trades for 59 accounts, and the answer to "did this trader make money" matched for 58 of them.

Botsfolio AI's benchmark scores

The biggest gaps are in losing. The typical trader's worst drop was 89.8%, against Botsfolio AI's 18.3%. Half of the traders lost 90% or more of their account at some point. Botsfolio AI never did.

Losing less is what Botsfolio AI does best, and it's also where human traders struggle most.

The head-to-head results

The typical trader in this group ended about $12,300 down. Half of them, 537 traders, lost 90% or more of their account at some point. That's what leverage does when a trade goes the wrong way: the account doesn't just dip, it gets wiped.

What if they had let Botsfolio AI trade?

Watch five real traders from the study against Botsfolio AI, with the same $10,000 over their exact dates: a typical one, a wiped-out one, a winner, a careful one, and one of the few who beat Botsfolio AI on both counts.

Across the whole test, Botsfolio AI made more money than 844 of the 1,076 traders. Here is what those 844 accounts looked like, next to the same money with Botsfolio AI over the same dates:

Botsfolio AI didn't win every window. 82 of these 844 still ended in a loss, and its worst window lost 17.9%. The difference is the size of the losses. Its worst window cost less than a fifth of the money, while most of these traders lost most of their account at some point.

The 2.4% who beat it, and the 19% who made more

Some traders are genuinely good. 26 of the 1,076 made more than Botsfolio AI and fell less while doing it. If you're one of them, you don't need us.

Another 206 traders, 19.1%, made more money than it, by a median of 28 percentage points. But to get there, their typical worst drop was 49%, against Botsfolio AI's 10% over the same dates. With leverage, a 49% fall is often a few bad days away from liquidation. Some of those traders will keep winning. Others are one bad week from joining the half who lost 90%.

That's why we score return per 1% of drop, not just return. A trader who makes 30% with a 50% drop and an AI that makes 20% with a 10% drop didn't take the same risk to get there. Botsfolio AI had the better return per 1% of drop against 85.4% of the traders.

Careful traders vs everyone else

We also split the 1,076 traders by their own worst drop and checked Botsfolio AI against each group:

Careful traders hold their own: 71% of them made money, and some beat Botsfolio AI on both counts. Once a trader let a drop pass 25%, not one of them beat it on both. That's 85% of all the traders in the test.

This period was kind to Botsfolio AI, though. Tested from every monthly start since 2018, its typical biggest drop was about 37%, so expect deeper falls than this test shows.

Why most human traders lose

The results above aren't unusual. They match what other data shows about how people trade crypto:

• Leverage wipes people out. On October 10, 2025, 1.6 million traders were liquidated in 24 hours, $19.1 billion in positions, about 90% of them bets that prices would rise, according to CoinGlass.
• The crowd stays long, even when prices fall. In Bybit's public data, most BTC futures accounts were betting on a rise on 89% of days since January 2022. Over the last 12 months, while Bitcoin and Ethereum together fell 29%, it was 94% of days.
• Emotion drives the trades. In a Kraken survey of 1,248 US crypto holders, 84% said they had acted on fear of missing out, and 63% said emotional decisions had hurt their portfolio.
• It's not just traders. The Bank for International Settlements found that most crypto app users lost money on bitcoin between 2015 and 2022, because they bought after prices had already risen.

Most of it comes down to time, emotion and leverage. An AI doesn't get tired, scared or greedy, and Botsfolio AI doesn't use leverage at all. We wrote more about the habits behind it in why crypto traders lose money.

Who Botsfolio AI is for

The 2.4% who beat it don't need it. Botsfolio AI is built for the rest:

• People who trade on the side. 656 of the traders in this test had between $1,000 and $100,000 in their account. 26.4% of them made money. Botsfolio AI made money in 92.1% of the same windows, and $10,000 in this group became $2,455 for the typical trader against $12,472 with Botsfolio AI.
• People who have been liquidated, or are tired of watching leverage. Botsfolio AI trades spot coins with no leverage, so a bad day can't wipe the account.
• Holders who believe in Bitcoin and Ethereum but can't sit through 60% drops. Holding both fell 53% at its worst in the last year. You keep the coins you believe in, with a plan for the falls.
• People who want rules instead of hunches. Every trade follows the same rules, every time, at 3am and on the worst day of the month.
• People who mainly want to stop losing. Botsfolio AI can be told to keep falls to a minimum. More on that below.
• Beginners. You can watch it trade for free before putting in any money.

It's not for everyone. If you want 10x in a month, want leverage, or are already in the 2.4%, this isn't your product.

When we told Botsfolio AI to keep falls to a minimum

Botsfolio AI can also be told to put small falls first. That version runs as the BTC & ETH Steady Trader. We ran it against the same 1,076 traders, with the same money over the same dates:

It almost never loses big. Its typical worst drop was 9%, a typical window had no month that lost 5% or more, and only 12 traders, 1.1%, beat it on both return and worst drop.

The price is growth. Its median gain was 3.3% over about a year, and about a quarter of its windows ended slightly down, the worst by 9.9%. Tested from every monthly start since 2018, its typical biggest drop was about 11%, against about 37% for the main Botsfolio AI. It suits people who mainly want to stop losing, more than people who want their money to grow.

How to try it

• Watch it first. Watch Botsfolio AI trade live: every buy, sell and stop it makes, in public, next to holding the same coins.
• Look at the AI from this test. The BTC & ETH Cycle Trader shows its full record and every trade. The minimum-fall version runs as the BTC & ETH Steady Trader.
• Start free with paper money. Build your own AI portfolio on the coins you pick and watch it trade on live prices. No card needed.
• Use real money when you're ready. Run the same AI on your own Coinbase, Kraken, Binance or Binance.US account. Your money stays on your exchange, the key can't withdraw, and you pay only when a trade fills. See pricing and how connecting works.

Method and limits

How we picked the AI, how its portfolios are tested, and the tests they didn't pass are on the portfolio methodology page. To understand the drop numbers, read our guide to drawdown.

FAQ

Educational analysis, not financial advice. Trader results come from Hyperliquid's public data. Botsfolio AI's results are tested on past prices. Trading crypto can lose money. Past performance does not predict future results.