Plug Claude, GPT, Gemini or a model running on your PC into the Lab. It searches on your machine and your graphics card. Then an independent judge replays the strategy on data it has never seen and says whether it exists or is a mirage.
Same tool, same data, a sealed period no AI had seen. Top averages on October 7, 2026:
7 of 23 scored runs beat buy-and-hold once risk is counted.
Round 1 was played on a past period (Oct 2023 → Sep 2026) the models may know; round 2 will be played on future data. Open the leaderboard →6 of the 6 leaderboard strategies, 234 of 234 trades, give the same trades one by one in the real TradingView, quantities included (Lab 236 export, test of 8 October 2026).
TradingView's Strategy Tester is the author's tool. The judge is the second opinion: same trades first, then data nobody has seen. How the judge works → Open TradingView (partner link) Affiliate link. Partner link: Strategy Arena may earn a commission on an eligible purchase. It never changes a verdict.1,000,000 strategy variants in about 20 seconds on an RTX 4080 Super (Lab 235 measurement, 7 October 2026, legacy execution semantics). Every row that is kept is replayed on the CPU reference engine.
Small models running on your own PC complete the whole path in 6 calls: start_here, then datasets, validation, backtest, sweep and Pine export.
“Winner or mirage?”: the sealed proof keeps a test period unseen during the search, compares with buy-and-hold on the same windows and counts every variant tried. Plug in my AI →Launch video · 73 s
Silent video with on-screen captions. An illustrated opening (Grok Imagine), then real captures of the Lab window: in the Colosseum, Claude, Cursor and DeepSeek take one challenge while a period stays sealed; the GPU sweeps 200,000 variants in 2.8 s and the CPU replays the ranking with 0 mismatches; Night mode delivers its morning report (verdict: Mirage); the TradingView judge replays 234 / 234 identical trades; the reveal shows a training winner that is a mirage on the sealed period; a published badge is recomputed by the server on strategyarena.io/verified. Past results, not a forecast.
One month of BTC. The future is hidden. Choose your move before the chart reveals it.
MARKET CHALLENGE · BTC/USDT · 30 DAYS
One historical market, three choices, the same starting capital. Decide first; see the dates after the replay.
Loading historical prices…
The future is still hidden.
Change one choice above to compare the same period.
Strategy Arena may earn a commission if you later purchase an eligible TradingView subscription. No purchase is needed for the chart exercise.
BTC/USDT hourly prices from a local Binance-loader archive, reduced to the last close of each complete UTC day. These two 30-day windows were chosen after seeing the outcomes; replaying them is educational, not an unseen validation or a forecast. Starting capital: 10,000 USDT. Hypothetical commission plus slippage: 15 bps per fill. The trend rule uses the previous day’s close against its 20-day average and fills at the next daily close. Displayed portfolio values include a hypothetical final exit cost. Cash is held at a constant 10,000 USDT; stablecoin, tax and custody risks are outside this exercise.
First chart and price alert: Open TradingView — first account. No real money required.
Android prototypes: Pocket Lab & Time Machine · Experimental, offline APKs.
Two desires. Two different products.
Quick Check and the Lab stay free. Pay 9.90 € when you want a sealed server report you can reopen. Open TradingView when you want to see the Pine on a live chart — first web account only; existing accounts stay neutral.
The Lab lives on your PC. 9.90 € buys an independent sealed holdout and a private URL you can reopen. Payment never turns REJECT into KEEP.
First web account here: live chart, price alert, paper trade. Start with the free account for this exercise. Already have an account? Use the neutral path — no partner cookie.
Affiliate disclosure: Strategy Arena may receive a commission if a first-time user later buys an eligible TradingView web subscription. Partner attribution is for first-time users and web purchases only. No paid subscription is required for the first exercise. Builder is a separate Lab subscription; no audit purchase is required. TradingView account choice · Terms
Free check. Paid report. Optional later.
Quick Check prevalidates Pine compatibility for free; it does not run stress tests or issue a verdict. The Windows Lab remains free for owner-operated local proof. An independent private Survival Audit costs 9.90 € once. Builder and Operator are separate subscriptions, with no audit prerequisite; they never improve its verdict or include an independent Audit.
Paste Pine for a compatibility-only Quick Check, or install the Windows Lab for an owner-operated local proof. No stress test, score or verdict is produced by the Quick Check. No card and no broker connection.
Strategy Arena runs one independent server-bound execution against the exact locked Pine, dataset and protocol, then delivers a durable private evidence report. An unfavorable result is still delivered; only a technical compute failure is refunded.
Choose a subscription for your research workflow. No audit purchase is required. Builder preserves private work. Operator adds bounded supervised research under your control. Neither subscription includes or replaces an independent Survival Audit.
“Try to kill my strategy” is an entry to the Audit, not a separate product. TradingView remains a separate, disclosed and human-controlled first-account path for new users only.
Loading the latest published edition…
The Lab guides one complete workflow on your machine: select, search, verify on unseen data, then inspect the report before TradingView.
Proof, not spectacle: demo REJECT · +172% that lost 37% · World Arena facts · Pine converter
Current Arena facts come from fingerprinted receipts. Missing evidence stays visibly missing instead of being replaced by marketing claims.
Watch live simulations if you want. They do not replace a local KEEP / REJECT on your idea.
Watch AIs, gurus and algorithms fight live market data. Every win, loss, mutation and corpse is public. Only survivors move on.
Query the first official MCP server with Claude, ChatGPT or Cursor. Read the leaderboards, stress your algos, export the survivors to Pine Script v5. We do the dirty work (Monte Carlo, real fees), you get clean code.
The Pine export carries the arena's fees and slippage settings. Verified: the converted script compiles without errors in the TradingView editor. Not measured yet: trade-by-trade parity with the arena.
Streamable HTTP · POST /mcp · start with get_product_contract
Each one hunts the market its own way. Some survive. Others mutate. Living vitals — not profit promises.
A deep-sea fund organism. Distributes capital across the surviving strategies and tightens its grip when the regime turns.
Keeps the 50 patterns that survived statistical filtering (5% FDR, 5 walk-forward folds) on a year of BTC, ETH, SOL and BNB. Out-of-sample Brier 0.2512: close to a coin flip, so it gives context, not a forecast.
Cut one head, two grow back. A recurrent network still in training — volatile, hungry, not yet ready to be released.
Hundreds of decision trees in armored formation. Slow to excite, almost impossible to fool — it votes, then it strikes.
Nine minds vote on what comes next. It sleeps until summoned — then opens its eye and weighs the odds.
The audit and subscriptions are separate purchases. No audit is required to subscribe.
9.90 € buys a sealed server report you can reopen — not a better verdict. TradingView is for seeing the Pine on a live chart if you still need a first web account. The Lab stays free on your machine.
Assemble atoms in Lab (Lab 0) free. Keep reports — Builder keeps private report history.
The Lab runs free on your PC. Paid options only keep or expand the work — they never change the verdict.