The "$4,000 in 10 days" Grok arbitrage bot: the wallet never bought both sides once
A beginner's guide on X says a simple bot that buys Yes and No for under $1 made its author $4,000+ in ten days, and links the wallet as proof. We read all 14,168 of that wallet's public records. It never made the trade the guide teaches, and its trading lost money once the guide was out.
The claim"I run this exact strategy on BTC markets only, nothing fancy, and it's made $4,000+ in 10 days": buy Yes and No together for less than $1, merge the pair back to $1, repeat. — X (@bored2boar)
The short version
The article, "From $0 to your first arbitrage bot with Grok Build", was posted on X on September 3, 2026 and has been viewed about 226,000 times. Its idea is the classic one: on a two-outcome market, Yes and No together always settle for $1, so if you can buy both for 97¢ you lock in 3¢ whatever happens. It tells beginners to watch Bitcoin "Up or Down" markets, buy both sides when they add up to less than $1, merge the pair straight back into $1, rest orders instead of paying taker fees, and never hold one side on its own. Its proof: "I run this exact strategy on BTC markets only… and it's made $4,000+ in 10 days", with a link to a public wallet and a screenshot reading "$4,028.09, past month". It ends with an invitation to the author's Telegram channel.
Every trade that wallet made is public, so we downloaded its full history from Polymarket's data API: 14,168 records from August 22 to September 29, 2026, all in 15-minute Bitcoin Up/Down markets. Across 1,455 markets, it never once bought both Up and Down. It didn't buy pairs below $1 at all. Instead, 1,350 times it turned cash into Up+Down pairs at exactly $1 each, $249,803 in all (Polymarket calls this a split), then bought and sold one side while keeping the other. It merged a pair back into $1 twice, for $3 in total. It paid taker fees on 94.5% of the money it spent buying, about $2,446, against the guide's own advice to rest orders.
The result is a run of one-sided bets, not arbitrage. 1,016 of its 1,455 markets ended with the wallet holding more of one side than the other, and that heavier side won 48.1% of the time, about as often as a coin flip. The screenshot in the article bears this out: every win it shows is a single "Up" position, such as 300 shares bought at 33.6¢ and returning 185%, a payout no locked-in 3¢ pair could produce.
The money tells the same story. Counting every dollar in and out, the wallet's trading made $2,261 in the markets it opened before the article went up, then lost $2,681 in the 472 markets it traded afterwards, so its trading is about $419 down overall. That matches Polymarket's own public leaderboard, which shows the wallet at −$441 all time. The account is only ahead, by $1,738, because Polymarket paid it $2,158 in fee rebates. The best ten days we can find in the record, rebates included, came to $3,621. The wallet last traded on September 28 and now holds nothing.
Would the guide's strategy work for you? Its own description is accurate: when Yes and No really cost less than $1 together, buying both locks in a profit. The problem is how rarely that happens at prices you can actually trade. We watched these same Bitcoin markets for exactly this gap: buying both sides below $1 at the same moment was available in none of 4,195 scans, and even buying each side at its cheapest price of the whole market life cleared fees in 0.2% of 21,512 markets. Verdict: the wallet offered as proof never ran the strategy, and once the guide was published its trading lost money.
Technical appendix
Exhibit A What the wallet actually did, market by market
Wallet 0x4cea338e37835dfe624ee77d1031c73cf4f71286, linked from the article. Each of its 1,455 markets falls into one of six patterns. P/L is cash in minus cash out for that market, after fees, before rebates (rebates are paid to the wallet as a whole).
| What it did in the market | Markets | P/L |
|---|---|---|
| Split cash into an Up+Down pair, then bought and sold one side while holding the other | 647 | −$862.13 |
| Split into a pair and did nothing else (redeemed for what it cost) | 399 | $0.00 |
| Split into a pair, sold one side, kept the other | 181 | −$34.70 |
| Split into a pair, then bought more of one side | 113 | +$306.66 |
| No split; bought and sold one side | 64 | +$251.61 |
| No split; bought one side and held it | 51 | −$80.92 |
| Bought both Up and Down (the guide's trade) | 0 | — |
| Total | 1,455 | −$419.48 |
A split creates a pair at exactly $1, the opposite end of the guide's trade, which only pays when a pair costs less than $1. Once one side of a split pair is sold, what is left is a bet on the other side.
Exhibit B Before and after the article
Split at the moment the article was posted (September 3, 2026, 04:54 UTC), by when each market's trading began.
| Markets | Trading P/L (after fees) | Taker fees paid | Rebates received | Net | |
|---|---|---|---|---|---|
| Before | 983 | +$2,261.45 | $1,281.81 | $1,062.79 | +$3,324.24 |
| After | 472 | −$2,680.92 | $1,164.02 | $1,094.93 | −$1,585.99 |
| All | 1,455 | −$419.47 | $2,445.83 | $2,157.72 | +$1,738.25 |
Taker fees are measured on each buy as the amount paid above price × shares. The wallet's buying was 94.5% taker by dollars. After the article it traded on 11 more days, the last on September 28, and its balance is now $0.
Exhibit C The numbers in the article
| Figure | Value | Source |
|---|---|---|
| "Past month" profit in the article's screenshot | $4,028.09 | Polymarket profile, date of capture not shown |
| Best 10 calendar days, all cash in and out, rebates included | $3,621 | our count, Aug 25 – Sep 3 |
| Cash gained up to publication, rebates included | $3,324 | our count |
| Polymarket leaderboard, all time | −$441.11 | data-api v2, Oct 27 |
| Trading result, all time, before rebates | −$419.47 | our count |
We could not reproduce the $4,028.09 exactly. In our earlier checks the profile figure tracked profit before fees, and this wallet's trading before fees, up to publication, was about $3,543. The leaderboard's −$441.11 is within $22 of our own trading-only total; it appears to leave rebates out.
Exhibit D Our own tests of the same trade
We have tested buying both sides of these exact Bitcoin Up/Down markets below $1, at prices that could actually be traded:
- At the same moment: 0 hits in 4,195 scans.
- At any time in the market's life: even buying each side at its cheapest price of the whole market cleared fees in 46 of 21,512 markets (0.2%).
- Resting bids on both sides (the maker version the guide recommends): −31.6% over 15,366 settled markets. The orders that fill are the ones where the price is running against you.
See the Lab entries on crypto-candle arbitrage and the passive two-sided bid grid.
Exhibit E Links in the article
The article closes with an invitation to the author's Telegram channel, "Bored Alpha", described as free and shown with 281 subscribers. The wallet link in the article carries a via=bored2boar tag. Polymarket says any profile page can be shared as a referral link, and that referrers earn 10% of the trading fees their sign-ups pay for 30 days (Polymarket's referral program); its documentation does not say how those links are formed, so we cannot confirm whether this tag is one.
The guide's mechanics are described accurately, including its warnings about fees and about holding one side alone. The wallet offered as proof did both of the things it warns against.