“It never picks a side”: the wallet ended 92% of its markets betting on one, and that is where the profit came from
A post on X says a Bitcoin bot makes money by minting Up and Down pairs for $1 and selling both sides to the crowd for more. We read all 267,126 of the wallet's public records.
The claim“Every five minutes it pays Polymarket $1 per pair and gets one Up and one Down back… Then it sells the Up to people betting on a pump and the Down to people betting on a dump.” “192,111 trades since May, not a single buy.” “+$33K in October after four flat months.” “It never picks a side, it just prints both and lets the crowd fight over them.” — X (@rcom1337)
The short version
The post went up on X on October 23, 2026 and had about 4,650 views when we read it. It says a bot turns $1 into one Up and one Down token on every 5-minute Bitcoin market (Polymarket calls this a split), then sells the Up to people betting on a rise and the Down to people betting on a fall, keeping whatever is left on the winning side. It gives three figures: $1,500 to $2,000 minted on every market, 192,111 trades since May without a single buy, and +$33K in October after four flat months. It links the wallet's profile and ends: “It never picks a side, it just prints both and lets the crowd fight over them.”
Every trade on Polymarket is public, so we downloaded the wallet's full history from Polymarket's data API: 267,126 records from June 3 to October 28, 2026, every one in a 5-minute Bitcoin Up/Down market, 12,244 markets in all. Most of the post's figures hold. The wallet has made 232,089 trades and every one is a sale; it has never bought. By the moment of the post it had made 192,636 trades, close to the post's 192,111. In October up to the post it split pairs in 6,105 of the 6,534 markets that ran, $1,500 or $2,000 in 73% of them and $1,000 in the rest. Two details are off: its history starts on June 3, not in May, and the $1 is not paid to Polymarket; it is locked in the pair and comes back when the pair is redeemed.
The method is where the post goes wrong. Selling Up from a fresh pair is the same trade as buying Down at $1 minus the sale price, so every sale is a bet that the side sold will lose. If the wallet really sold both sides to the crowd for more than $1, it would profit whoever won. It did not: where it sold both sides of the same market, the two sides fetched 89.4¢ a pair on average, and that part of its trading lost about $175,000. It sold only about a fifth of the tokens it minted, and it ended 11,227 of its 12,244 markets (92%) holding more of one side than the other. Those one-sided bets made about $225,000. Whether the wallet chose those sides or was left with them by the crowd, it was picking a side in nearly every market, and that is where the money came from.
The bets were good ones, but only recently. Judged by the wallet's own sale prices, its one-sided bets should have won about 6,810 times; they won 7,385. That gap is far too large to be chance. But it is almost all October. From June to August the bets won about as often as their prices implied, and the wallet's trading lost $1,010; in September it made $3,314. In markets that started in October it made $47,420, as it raised its size. The post's +$33K matches Polymarket's own all-time figure for the wallet on the day of the post, $32,769, which includes $7,166 of fee rebates; October alone was about $30,500 by Polymarket's figures.
The wallet has kept making money since the post: +$24,538 in the 1,000 markets it has traded since, now minting $5,000 a market. In all, its trading made $49,724 after $12,634 of fees, and Polymarket paid it another $10,164 in fee rebates, $59,888 together. Polymarket's own figure is $59,941. Verdict: the trade counts, the no-buys claim and the October profit are real, but the bot does not make money by selling both sides for more than $1. It sold pairs for less than $1 and made its profit on the side it kept.
Technical appendix
Exhibit A What the wallet did
Wallet 0x56f609c609975454bf851f0d91079fbc4a5fe251 (arcsine-barrister), linked from the post. Every record from Polymarket's public /activity feed, June 3 to October 28, 2026 (UTC).
| Record type | Count | Amount |
|---|---|---|
| Split ($1 into one Up + one Down) | 12,261 | $14,528,155 |
| Sell | 232,089 | $2,343,847 |
| Buy | 0 | — |
| Merge (pair back into $1) | 15 | $6,030 |
| Redeem (winning tokens for $1 each) | 22,677 | |
| Fee rebates | 84 | $10,164 |
All 12,244 markets are 5-minute Bitcoin Up/Down markets, and all have resolved. Of the tokens it minted, it sold 19.6%; the rest were held to the close, where a held pair returns its $1. 80.5% of its sales by dollars were resting orders that paid no fee; the other 19.5% paid $12,634 in taker fees.
Exhibit B Where the profit came from
Selling a token from a fresh pair is the same as betting that side loses: a sale at price p makes p if that side loses and p − $1 if it wins. Per market, we split the wallet's sales into pairs (as many Up as Down sold) and the rest (the extra sales on one side, which leave the wallet holding the other). Pairs are valued at each side's average sale price in that market, so the split is approximate; the total is exact.
| Part of its trading | Amount | P/L |
|---|---|---|
| Sold both sides of the same market | 1,646,680 pairs at 89.4¢ average | −$174,879 |
| Sold more of one side, keeping a bet on the other | 11,227 markets | +$224,603 |
| Total, after fees | 12,244 markets | +$49,724 |
A pair sold for less than $1 loses money whichever side wins. The one-sided bets won 7,385 times in 11,227 markets; the wallet's own sale prices implied about 6,810. By chance alone, a gap that large would be about 14 standard deviations.
Exhibit C Month by month
Markets grouped by the month they started (UTC). P/L is trading after fees, before rebates. “Implied” is the number of wins the wallet's one-sided bets should have had at its own sale prices.
| Month | Markets | Typical split | One-sided bets won | Implied | P/L |
|---|---|---|---|---|---|
| June | 318 | $195 | 88 | 100.4 | −$396 |
| July | 257 | $55 | 62 | 67.8 | −$137 |
| August | 477 | $60 | 221 | 218.9 | −$477 |
| September | 4,087 | $325 | 2,184 | 2,082.1 | +$3,314 |
| October (to the 28th) | 7,105 | $1,500 | 4,830 | 4,341.1 | +$47,420 |
Of October's figure, $22,882 came from markets that started before the post and $24,538 from the 1,000 markets since. Since October 25 it has split $5,000 in most markets.
Exhibit D When and at what price it sold
Each sale's P/L, as in “Where the profit came from”, grouped by the net price received and by how much of the 5-minute window was left.
| Sale price | Sales | Received | P/L |
|---|---|---|---|
| under 10¢ | 39,506 | $61,762 | +$10,847 |
| 10–30¢ | 51,503 | $235,533 | +$1,909 |
| 30–50¢ | 43,279 | $479,636 | −$3,677 |
| 50–70¢ | 35,990 | $468,239 | +$16,342 |
| 70–90¢ | 28,662 | $378,955 | +$17,824 |
| 90¢ or more | 33,149 | $719,723 | +$6,479 |
| Time left in the window | Sales | Received | P/L |
|---|---|---|---|
| more than 4 minutes | 50,838 | $665,049 | +$30,503 |
| 2–4 minutes | 78,417 | $684,837 | +$4,151 |
| 1–2 minutes | 53,494 | $539,219 | −$1,792 |
| 30–60 seconds | 33,829 | $329,304 | +$8,222 |
| last 30 seconds | 15,253 | $123,961 | +$8,677 |
| after the close | 258 | $1,477 | −$37 |
Exhibit E The post's claims, one by one
| Claim | What the record shows |
|---|---|
| Splits $1 into an Up and a Down every five minutes | True. In October up to the post it split in 6,105 of 6,534 markets. The $1 is locked in the pair, not paid to Polymarket. |
| The “50¢” on every winning position | True. Polymarket records each split token at 50¢. |
| Sells the Up to people betting on a pump and the Down to people betting on a dump | Partly. It sold both sides in 7,573 markets, but for 89.4¢ a pair on average, and it sold only about a fifth of what it minted. |
| Whatever is left on the winning side pays out a dollar | Partly. In 92% of markets it was left with more of one side, and that side lost 3,842 times. |
| $1,500 to $2,000 minted on every market, all day | Mostly true for October before the post: $1,500 or $2,000 in 73% of markets, $1,000 in 27%. Since October 25, mostly $5,000. |
| 192,111 trades since May, not a single buy | True on the trades (192,636 by the time of the post, zero buys). Its history starts June 3. |
| +$33K in October after four flat months | Roughly. Polymarket's all-time figure on the day was $32,769, including $7,166 of rebates; October alone was about $30,500. June to September made $2,304 in trading. |
| It never picks a side | False. It ended 11,227 of 12,244 markets with a one-sided position, and those positions made all of its profit. |
Exhibit F The numbers we compared
| Figure | Value | Source |
|---|---|---|
| Trading P/L, all markets, after fees, before rebates | +$49,723.76 | our count |
| Fee rebates received | +$10,164.13 | activity feed |
| Total | +$59,887.89 | our count |
| Polymarket daily P/L history, latest | +$59,941 | data-api v2 user-pnl, Oct 28 |
| Polymarket leaderboard, all time | +$48,651 | data-api v2 leaderboard, Oct 28 |
| Polymarket daily P/L history, Oct 23 00:00 UTC | +$32,769 | data-api v2 user-pnl |
The leaderboard figure appears to leave rebates out, like our trading figure; the two are $1,073 apart.
Exhibit G Links in the post
The profile link in the post carries a via=rcom1337 tag, and the website on the author's X profile is a polymarket.com link with an r=rcom 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 these tags are referrals. The author's X profile describes them as a Polymarket poster and bot tester.