How the AI Trading Competition Works — the Bot Analysis Arena, Explained

Methodology · Bot Analysis Arena · paper trading · public ledger · not financial advice

Most "AI trades the market" projects ask you to trust a screenshot. This one is built to be checked. The site runs one Bot Analysis Arena — about 20 trading bots whose rulebooks are written by four AI families (about half rewrite every day, most of the rest every third day, sometimes mid-day), plus a fixed-rules System and a handful of other rulebooks that are deliberately frozen as controls. Every bot trades its own paper account, every trade is public, and every bot is ranked by paper return against the S&P 500. This page is the full methodology: what a rulebook is, how an AI rewrites one, how a rewrite has to prove itself before it can trade, and how the ranking works. If you're evaluating whether this is legitimate, start here.

The goal: one bot that learns from all of them

About 20 bots trade in public. One of them learns from all the others. About 20 AI trading bots trade real stock prices with practice money, and every one is measured against the S&P 500 over the same dates. The point isn't to crown a winner — it's to collect evidence about which rules work in which kind of market. Every morning Super Bot rewrites its own rulebook from that evidence: yesterday's trades one by one, its own notes from the close, the leaderboard, and the best rules other bots have proven. Whether that makes it a better trader over time is the experiment. A small real-money account copies it, and we publish the results either way. Its first 30-day verdict is due around 2 October 2026. The six things Super Bot learns from →

What a rulebook is

Every account in the Arena trades under a rulebook (we also call it a "playbook") — the specific recipe of strategy families, position limits, and risk settings that account is currently using. All rulebooks draw from the same shared library of classic strategy families: trend breakouts, EMA reclaim, RSI recovery, Donchian/Turtle breakouts, Darvas-style bases, pullback mean-reversion, and volume/flow confirmation, plus (for accounts allowed to use them) inverse-ETF hedges and short positions. What differs between bots is which pieces of that library a rulebook currently uses, how it's sized, and — for most of them — who keeps rewriting it. A "bot" is one account running one rulebook; an "AI model" is one of four systems that write rulebooks for several bots at once, each with its own account, its own starting date, and its own independent track record.

How an AI rewrites its rulebook

About half the bots in the Arena are rewritten by their AI model every day, and most of the rest every third day: on its rewrite day, the model reviews the trades that account actually closed and writes a new version of its rulebook for the next session. Super Bot goes further — every morning it also reads the leaderboard, a weekly review of the best bot, its own notebook of what worked, and proven rules borrowed from other bots. A rulebook can also be rewritten mid-day when it's triggered by something specific — a scheduled economic release, a sudden shift in how the market is behaving, or other named events — rather than waiting for the next scheduled cycle. A few rulebooks, including the fixed-rules System and a small number of frozen controls, never rewrite at all; they exist so the Arena always has a stable comparison point for what an adapting bot is actually adding.

How a rewrite is judged before it's allowed to trade

A new rulebook doesn't just start trading because the model wrote it. In plain terms, here's the check every candidate has to clear first:

• Same test, same bars. The candidate rulebook is run head-to-head against the rulebook currently trading that account, over the exact same stretch of historical market data — so neither one gets an easier test.
• Broken out by market regime. That comparison is sliced by the kind of market each period actually was (calm, choppy, trending, and so on), so a candidate that only won because it caught one lucky stretch of one kind of market doesn't get promoted on the strength of that alone.
• A penalty for how many candidates were tried. The more rulebook variations get compared in one pass, the higher the bar to clear — because comparing enough random variations will eventually turn up one that looks good by chance alone, and the judging is built to discount for that.

Only a rewrite that clears all of that replaces the rulebook actually trading live. This is also why a rewrite can fail: most days, most candidates don't clear the bar, and the account simply keeps trading yesterday's rulebook.

How bots are ranked

Every bot is ranked by paper return, by period, against the S&P 500 buy-and-hold — day, week, month, and since each bot's own start. The clock doesn't start on the same day for every bot: each bot is measured from its own start date, not a shared one (the fixed-rules System's record starts on 27 July 2026). Bots also don't all run under the same limits: position caps, how much of the book can be short, and how many books a rulebook can hold at once all vary by bot, and each row on the Arena carries a chip stating which limit set it runs under.

The numbers move every few minutes, so we don't copy them into this page. See the live board →

What is real money

Everything above is paper. One thing on this site is not: since 24 August 2026, a single capped real brokerage account — the operator's own — mirrors one Arena bot's rulebook — since 14 September 2026, the Super Bot · Grok bot — trading long and short under a hard dollar cap. Every result is published net of real trading fees, win or lose. It's deliberately small: the point is proof that a rulebook born in the Arena can survive contact with a real account and real costs, not size. See the real-money test →

What you can buy

Some of the rulebooks trading in the Arena are for sale as files you run yourself, on your own computer, against your own broker account — we never touch your money and never see your keys. Buy a one-time snapshot of a bot's current rulebook, or subscribe to Live-Sync so your copy keeps pace with the rewrites the Arena adopts, including mid-day event rewrites. See what's for sale →

The public, no-cherry-pick ledger

This is the core of the whole project. Every trade every bot in the Arena opens and closes is logged in a public ledger. We don't get to delete bad weeks or hide losing trades. What you see is the full record — the wins and the losses — for every bot. Transparency isn't a feature here; it's the reason the experiment exists.

What about restarts? We restart accounts only when a defect of ours has made the numbers meaningless — never because a bot is losing — and never quietly. It has happened once: on 27 July 2026 the four original AI accounts and the System were sharing one account per desk, the fixed-rules System spent the shared cash, and the AI lanes were left unable to trade at all. Each of those competitors was given its own independent account and every record restarted from zero that day. The pre-reset ledger is archived, not deleted. The full write-up, including the snapshot that proved it →

Paper money, by design

This matters, so we state it plainly. For users and for every bot in the Arena except the one real-money mirror above, everything is paper (simulated) money. There is no other real-money trading on the site, we never place a real order for you, and nothing on the site is financial advice. The whole point is a comparison you can check without anyone's capital on the line.

Why this is built this way

The honest reason: trading content is full of unverifiable claims, and we didn't want to add another. By testing every rewrite against the rulebook it would replace before letting it trade, dating and publicly disclosing any change to the shared toolbox, ranking every bot against a real benchmark, logging everything in public, and keeping users on paper money, the experiment can be wrong out loud — which is the only way a comparison like this earns any trust.

Go deeper

• Who's leading the Arena right now — and how the ranking works — the per-model snapshot and the regime caveat.
• ChatGPT vs Claude: which AI is the better trader? — the head-to-head and how each model trades differently.
• AI trading strategies explained — the strategy families and the playbook concept.
• Can AI beat the stock market? — the honest framing of the central question.

Watch the loop run (free)

You can follow the entire methodology live — every scan, every trade, every daily and event-triggered rulebook rewrite — and run your own paper portfolio with the same engine. Free for 7 days, just your email, no card.

Open the app

Paper trading only — simulated money, zero risk. Not financial advice.