🔍 Read the full analysis: Week Four’s AI Trading Experiment Finds A Viral Strategy Loses on ThorstenMeyerAI.com
Get everyday essentials delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
ThorstenMeyerAI.com reports that two simulated versions of a viral Polymarket trading strategy lost money across about 13,000 market windows on BTC, ETH, SOL and XRP. The strategy’s core double-fill happened three times in 9,486 paired attempts, while a post-close version’s four settled fills all lost.
Two simulated versions of a viral strategy for Polymarket’s 5-minute crypto markets lost money in a two-day test across about 13,000 market windows, according to a May 21 report from ThorstenMeyerAI.com. The test found that the strategy’s key condition—buying both sides at 2 cents—occurred just three times in 9,486 paired attempts; the report says no real funds were used.
The strategy followed a YouTube video showing a trader turning $1 into $50 on a Polymarket BTC Up/Down market. The report says the video included verifiable on-chain transactions, but the video’s creator had not established how often the trade could be repeated. The experiment described by ThorstenMeyerAI.com aimed to test that repeatability rather than dispute that the showcased trade occurred.
In the paired version, the bot placed 2-cent bids on both Up and Down. Because one side of the binary market pays $1 per share at settlement, buying 50 shares on each side for a total of $2 would return $50 if both sides filled, regardless of the outcome. The report describes the apparent opportunity as stale liquidity near market close: a trader might buy a resting order that another bot failed to cancel. But the required double-fill appeared in only 3 of 9,486 attempts, or 0.032%, while the bot filled only the losing side 1,297 times, or 13.7%. The simulated paired strategy recorded a net loss of $280.
The second version placed bids after the market closed, only on the side its price feed identified as the winner. It posted 3,482 times and filled eight times, a 0.23% fill rate. Four fills settled, and all four lost, the report says. The test ran both versions on BTC, ETH, SOL and XRP, using a paper-fill model in which a simulated bid filled only when a real taker sold into it.
AI Trading Experiment · Week Four
Week Four’s AI Trading Experiment Finds a Viral Strategy Loses
Two simulated versions of a viral Polymarket strategy lost money across roughly 13,000 five-minute market windows. The report tested whether a striking trade could be repeated, using paper fills and no real funds.
The core event behind the headline payoff appeared just 0.032% of the time.
Reported net loss in the simulated paired-bid experiment.
Published May 21, 2026
Five-minute market windows
BTC · ETH · SOL · XRP
Simulated experiments
A big payout depended on two rare fills
The viral example showed a trader turning $1 into $50 on a Polymarket BTC Up/Down market. The report says its on-chain transactions were verifiable; the experiment asked whether the underlying trade could happen often enough to repeat.
Buy both sides for 2¢ each
The bot posted bids on both Up and Down in a binary market. For 50 shares per side, the total cost would be $2 if both bids filled.
One side settles at $1
At settlement, one side pays $1 per share. If both 50-share bids filled, the combined payout would be $50 regardless of direction.
Catch stale close-time orders
The opportunity depended on buying a resting order near market close after another bot failed to cancel it in time.
One-sided losses were far more common
The paired strategy needed both bids to fill. A fill on only the losing side happened much more often, leaving the simulation with a net loss.
−$280 net
Across 9,486 paired attempts, the double-fill condition occurred three times. The bot filled only the losing side 1,297 times.
4 settled fills, 4 losses
The bot posted 3,482 times and filled eight times (0.23%). Four fills settled; all four were losses, according to the report.
The experiment tested repeatability
The video’s creator described rare high-multiple outcomes, including 50×, 48× and 100×. The report says the video did not establish how often the trade could be repeated.
Study the example
Review the video and reported transaction hashes.
Build two versions
Simulate paired 2¢ bids and post-close winner bids.
Run live-book tests
Use paper fills based on real taker activity across four coins.
Compare outcomes
Track fills and simulated settlements over two days.
“This is going to take time because this is a rare event…”
The YouTuber, as quoted in the report
A narrow test, with real limitations
These findings describe two simulated strategies under particular conditions. They do not establish how every market or future trading period would behave.
Paper fills, not live orders
The model filled a simulated bid only when a real taker sold into it. The report says no real funds were used.
Costs and execution are uncertain
The supplied account does not provide enough detail to independently reproduce every assumption or assess fees and execution delays.
Settlement can disagree
The post-close version relied on a price feed that could disagree with Polymarket’s official resolution. The report attributes its settled losses to this mismatch.
Short sample, four assets
Two days across BTC, ETH, SOL and XRP cannot forecast every market, period or order-book condition.
What the report does—and does not—say
What strategy was tested?
One version bid 2¢ on both Up and Down. Another bid after close on the side a price feed identified as the winner.
Why did the paired version lose?
Both sides filled in only 3 of 9,486 attempts, while the bot filled only the losing side 1,297 times. Its simulated net loss was $280.
Why did the post-close version lose?
The report says its price feed could disagree with Polymarket’s official resolution. All four settled fills were losses.
Does this prove the strategy can never work?
No. It reports losses for two versions in a two-day simulated sample, not outcomes for every market or condition.
Was real money involved?
No. The experiment used simulated money and paper fills; the report says no real funds were involved.
What comes next?
The article is Part 4 of an ongoing series. The supplied report names no date or details for a next installment.
Why the Double-Fill Rate Matters
The results address a basic question behind viral trading examples: can a striking return be repeated often enough to outweigh losses? In this test, the attractive payoff depended on both sides of a market filling at unusually low prices. That condition was rare, while fills on only the losing side occurred far more often. The reported figures show why a single large winning trade cannot establish that a strategy has positive expected returns.
The findings also point to a separate risk in trying to trade immediately after a market closes. The post-close version relied on identifying the eventual winner from a price feed, but the report says its feed could disagree with Polymarket’s official resolution. A strategy that appears risk-free under one timing assumption may not be risk-free when settlement is determined by a different price or oracle process. These are results from a particular simulated test, not a guarantee about every market or future trading conditions.
As an affiliate, we earn on qualifying purchases.
The YouTube video described in the report showed a 50-times return in a BTC five-minute market, alongside other reported outcomes of 48 times and 100 times. According to ThorstenMeyerAI.com, the creator used an AI coding agent to examine transaction hashes and propose how the mechanism might work. The video also acknowledged that the trade was rare and that it did not yet have results showing whether the approach could be repeated.
The May 21 article is identified as Part 4 of an ongoing series. Its authors say they rebuilt the strategy in two forms and ran 24 parallel experiments over two days. The test used simulated money and live order-book conditions, not actual trades with real capital. The report’s outcome counts therefore describe the experiment’s paper fills and settlements, rather than a record of a live account.
“This is going to take time because this is a rare event… probably I didn’t have anything for you in this video about the results.”
— The YouTuber, as quoted in the ThorstenMeyerAI.com report
As an affiliate, we earn on qualifying purchases.
Limits of the Simulated Results
The report does not establish how the strategies would perform with real orders, different execution delays, trading fees or changing order-book conditions. It says the test used simulated money and a fill model based on real taker activity, but the published material provided here does not give enough detail to independently reproduce every assumption or assess the impact of fees.
The post-close test also leaves a timing question: the authors attribute its settled losses to a mismatch between their price feed and Polymarket’s official resolution, but the available account does not give a full breakdown of the settlement discrepancies. The sample covers two days and four coins, so its findings should not be treated as a forecast for every market or period.
As an affiliate, we earn on qualifying purchases.
Further Tests and Series Results
ThorstenMeyerAI.com describes this as the fourth week of an ongoing series, but the supplied report does not name a publication date for the next installment or specify another test. Any follow-up would need to clarify whether the same pattern holds across longer periods and how execution costs, timing assumptions and settlement rules affect the results.
For now, the reported conclusion is limited to the tested versions and conditions: both lost money in the simulation. The article also stresses that its experiment involved no real funds and warns readers that attempting a similar strategy with real money could result in losses.
As an affiliate, we earn on qualifying purchases.
Key Questions
What strategy was tested?
The paired version placed 2-cent bids on both Up and Down in a 5-minute crypto market, aiming for both to fill and create a payout larger than the combined cost. A second version bid after close on the side a price feed indicated had won.
What did the paired-strategy test find?
ThorstenMeyerAI.com reports that both sides filled in 3 of 9,486 attempts. The bot filled only the losing side 1,297 times, and the paper strategy recorded a $280 net loss.
Why did the post-close version lose?
The report says the price feed used to choose the winner could disagree with Polymarket’s official resolution. Of the eight fills, four settled, and all four settled as losses.
Did the experiment use real money?
No. The report says the tests used simulated money and that no real funds were involved.
Does the test prove the strategy can never work?
No. It reports losses for two versions over a two-day sample across four coins. The results do not establish how the strategy would perform in every market or under different conditions.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
