2026-09-20PressBULLISH
8.0
“higher gas prices "a very inexpensive price to pay for what we've done”
Geopolitics escalationUSOIL ▲ 4/5
Build a rule from thisGOLD4,074.20-0.96%
WTI90.76+0.27%
BTC75,926.20-0.14%
NASDAQ26,629.81-6.36%
SPX7,353.73-4.36%
AlphaTrade · event-driven paper trading
Demo stageSIMULATED
A plain-English idea, read by AI into a strategy blueprint, risk-checked and backtested on a verbatim, sourced archive — then paper-traded in the open. Never live money here.
The gap
0
statements read and scored by the AI.
Each one is stored with its time and source, a relevance score and a read on gold and oil.
010.0 s
median time for the AI to read one statement.
Measured on statements it has actually scored. Reading time only: not a trade, and not a promise about prices.
020
strategies tested on the archive.
Each one is replayed on stored statements and prices, then stress-tested. The verdict can be “sample too small” or “fragile”, and we say so.
03Counted from our own archive. Updated 2026-11-06.
AlphaTrade closes that gap: the AI reads every statement, Studio tests your idea on the archive, and the Desk keeps risk limits on. All of it on paper, and we sell no signals.
WATCH THE LOOP
The whole loop, in two minutes.
One statement, read by AI, turned into rules, risk-checked, backtested on the archive and paper-traded — with nothing hidden in between.
Event AI
Strategies that read what the world says.
The AI reads each public statement in our archive and scores it for gold and oil. Your rules can use that score: archive first, paper only.
From a statement to a score: 0.0 s.
Median time for the AI to read one statement · not an end-to-end trade time
EVENT AIStatement scoringAI
Each archived statement gets a strength, relevance and direction for gold and oil.
Open Radar to see every statement the AI has read and scored.
EVENT AIArchive backtestAI
Your rule is replayed on stored statements and prices, then stress-tested (Monte Carlo).
Open Studio: say the idea in plain words, get a backtest and a verdict.
EVENT AIRisk checkAI
Per-trade and daily limits and a kill switch are checked before any paper position.
Paper only. No broker is connected and no real money moves.
Studio
Say the idea in plain words. Studio asks about what is unclear instead of guessing, reads your rules back to you, and tells you the truth about the test — including a No, with the reason.
Asks firstOne clear question when something is ambiguous — never a silent guess.
Rules in plain wordsYour rules come back as a short plain list, assumptions flagged, before anything runs.
An honest NoA too-small sample or a fragile result is called exactly that, with the reason — never rounded up.
Prefer to talk? Studio takes dictation, where your browser supports it.
StudioEXAMPLE conversation- Buy gold after a strong statement about interest rates. Hold a few days.
- One thing I won’t guess: every statement about rates, or only the ones the AI scores as high-impact?
- Only high-impact — 4 or 5 out of 5.
- Clear. Here are your rules, in plain words.
- WHEN: A statement about interest rates, AI strength 4 or 5
- DO: Go long Gold at the next daily open
- EXIT · ASSUMED: Stop 1%, target 2.5%, or after 3 days — change any of it
Written for this page. Not a real user, not a real result.
Want to see the AI read a real sentence first? Try the desk ↓
TRY THE DESK
Hand it a sentence. Watch the desk read it.
Paste any public statement, or pick an example. The same engine that reads the archive scores it against Gold and Oil in a few seconds, then lines it up with what the archive actually recorded afterwards.
0 / 600⌘/Ctrl + Enter
Sandbox: nothing is stored or traded.
The desk is waiting for a statement. It reads the words, classifies them, scores Gold and Oil, then checks the archive.
AI read — not a trade signal. Sandbox statements are not stored, not published and never reach the paper book.
Your route
01
Idea
Say it in plain words, typed or dictated. One sentence is enough to start.
TYPE DICTATE
02
Rules
Studio asks what is unclear and writes the rules back. You edit, then confirm.
TRIGGER — statement
FILTER — AI strength
ACTION — enter trade
03
Archive backtest
Your rules walk the statement archive and its price bars. A Monte Carlo reshuffle gives one verdict.
Sample too smallFragileRobust
Backtested so far: 0
04
Risk check
Loss and drawdown rules are checked first. On paper, size is throttled at 70% of the daily limit and halted at 85%.
DAILY LOSS LIMIT USED
- THROTTLE 70%
- HALT 85%
05
Paper
New matching statements open paper orders. Every fill is logged next to the backtest.
PAPER
ORDER
FILL
CLOSE
On paper now: 0
06
Arena
Ranked by AlphaScore once a strategy has 20 verified paper trades. The best can climb the virtual Capital Lane.
ALPHASCORE — after 20 paper trades
Beta · paper only. Every step runs on the archive — no real money, no broker.
The Desk
A risk desk for paper books. Every strategy trades inside limits set before its first paper trade, and the desk steps in as it nears them. Paper only: no funded accounts, no payouts.
01
Set the limits
A daily loss limit, caps on open positions and size, and a news blackout around a statement. All written before the first paper trade.
02
Throttle, then halt
At 70% of the daily loss limit the desk cuts size. At 85% it stops the book and closes its positions.
03
Kill switch and audit log
One switch pauses a strategy, a trader or the whole firm. Every limit change and every stop is logged: who, what, when.
The Arena
Strategies are ranked by AlphaScore, a risk-adjusted score, and only after 20 verified paper trades. Paper only: no prizes, no screenshots, no stories.
Strategy rankingEXAMPLE ROWS
1.Strategy AEXAMPLEAlphaScore0.00
AlphaScore · last 8 weeks
EXAMPLE
2.Strategy BEXAMPLEAlphaScore0.00
AlphaScore · last 8 weeks
EXAMPLE
3.Strategy CEXAMPLEAlphaScore0.00
AlphaScore · last 8 weeks
EXAMPLE
- Strategy DEXAMPLEpending — 0 of 20
- Strategy EEXAMPLEpending — 0 of 20
0 strategies live on paper0 backtested on the archive20 paper trades to be ranked No prizes, no money
Example rows are illustrative. Real strategies appear here once they pass 20 verified paper trades.
FROM THE ARCHIVE
The latest statements the AI has read.
Verbatim, dated, sourced. Click one to build a rule from it.
2026-09-17PressBULLISH
10.0
“big decision" coming — whether to annihilate the Iranian regime”
Geopolitics escalationUSOIL ▲ 5/5
Build a rule from this2026-09-16PressBULLISH
6.0
“I'm relying on Kevin, but he's got, you know, a very tough board, he's got a board that was put there by other people. And I told, I talked to Kevin, and I said, 'You might as well…”Build a rule from this
PLANS
Two plans. The same three caps.
Compare plansSimple is $100 a month. The Full package is $1,000 a year. Both are enforced by the engine — the cap is the product, not a promise.
Simple
billed monthlyFor one trader proving a system.
$100/ mo
- live strategies
- 5
- backtests / month
- 100
- open paper positions
- 3
Full package
billed yearlyFor a desk running several systems at once.
$1000/ yr
- live strategies
- 25
- backtests / month
- 500
- open paper positions
- 10
CUSTOM LIMITS · TRIAL · INCLUDED
Replay a real episode
Watch an archive statement, the AI read stored on it and the market reaction the desk recorded.
Open the replaySee the live numbers
Counts, latency and the share of verdicts the engine refused to endorse — computed from the database when you open the page.
Open the proofWhy now
Language models can now read a statement and say what it means. Statements and prices can be kept in an archive and replayed. Retail trading has plenty of ideas and few fair tests — AlphaTrade is built around that gap, on paper.
Running today
0
statements read and scored by the AI — each kept with its time and source
0ms
median time for the AI to read one statement
0
strategies backtested on that archive, each ending in a verdict
Live counters from the app itself — none of them typed by hand.
Models that can readREAD
A language model can read a statement and say what it means: which asset, which direction, how strongly — with its reasoning on show.
An archive to test onREPLAY
Statements and prices can be kept side by side with their timestamps, so an idea is tested against what actually happened.
Tests that can say noPROVE
Costs and a Monte Carlo check go into the backtest, and the verdict is allowed to read “Sample too small”. Arena ranks a strategy only after 20 verified paper trades.
THE RULES WE KEEP
What you will never see here.
- 01
No live money. Paper books only, labelled PAPER and SIMULATED wherever a number appears.
- 02
No invented numbers. Every backtest figure comes from the engine; every statement has its source link.
- 03
No signals for sale. The product is the loop — read, build, test, paper-trade, get ranked.
- 04
Small samples say so. Three trades is "sample too small", not a track record.
Type yours the way you would say it. Studio asks what it needs to know, tests the idea on the archive and gives you a verdict. No account needed to try.
or replay a real episodeBETA · PAPERWe sell no signals