RN1 Doesn't Need to Know Who Wins
Polymarket · Polygon · 10 Aug – 10 Nov 2026 · 92 days
(This describes what one wallet did on chain. It is not trading advice.)
Picture a guy in a bar on a Sunday afternoon in late September, three beers in, watching the Ravens play the Cowboys on a screen that is slightly too far away. Baltimore does something good, the room makes a noise, and he is already thumbing the buy button on Polymarket at whatever price the app shows him, because waiting thirty seconds for a better one feels like watching the bus pull away. He does not know who took the other side of that trade and he would not care if you told him. I went looking anyway, and in that game the order waiting for him very often belonged to a wallet called RN1, which made 452 fills there and let 419 of them come to it.
Ravens vs. Cowboys
Each dot is one of RN1's fills: when it happened, the price it paid and, by the dot's size, how many shares. Filled dots are maker fills and rings are taker fills. The lower chart tracks the shares RN1 held of each side.
I have no idea who our man in the bar is, and he is imagined, but I know his money well, because I pulled every one of RN1's 911,860 fills from 10 August to 10 November into a warehouse and spent a long time staring at what came out. Most of what gets written about RN1 on X is screenshots of it winning, usually next to a tool that promises to let you ride along, and the assumption underneath all of it is that the wallet knows who is going to win. My reading of the fills is that it mostly does not know and has arranged its whole business so that it rarely needs to. RN1 collects a small toll from people like our man in the bar, about a cent and a half per pair, and the rest of this piece is about how that toll is collected, who tops it up, who manages to dodge it, and a night in Barranquilla I still cannot explain.
Two Halves Make a Whole
The plumbing first, because everything else depends on it. On a Polymarket sports market, one share of each outcome held together always redeems for exactly 1 pUSD1, since one side pays 1 and the other pays 0, so anyone who can buy both sides for less than 1 combined has locked in the difference before kickoff and can stop caring about the score. The way to buy both sides cheaply is to rest bids on each outcome and wait for impatient people to fill them, and live sport produces impatient people on an industrial scale, which is why 89.75% of RN1's fills were as the maker2, the side whose order was already waiting.
The impatient person is rarely a seller. On one sampled evening, 90.5% of RN1's resting fills came from someone buying the other outcome, which in the bar goes like this: our man buys the Ravens, RN1's bid on the Cowboys fills, and the exchange mints a new pair, one share each.
Over the 92 days RN1 paid 0.985 pUSD per pair on average, fees included, and the pairs locked in +2.1M pUSD, while the shares that never found a partner gave back −486K of it.
Profit split into pairs and leftovers
Put together, RN1 spent 222.5M pUSD across 87,642 markets, every one of them sports, and its fills kept +1.6M, a return of 0.71% on what it spent that only makes sense when the same capital turns over every few hours.
The part people miss is that RN1 does not really set its own toll. Its bids ask for one cent, the least a one-cent grid allows: on the two outcomes they sum to 0.99 in 72% to 79% of the cases where I can see both and ask for more in 17% to 23%, though only bids that filled can be seen. The game decides what it keeps, since a pair comes out cheap when the side it bought first rallied before the second side filled and dear when that first side sank, and when pairs cost 1.05 or more, both halves lost.
Fills are not all the money. On each of the 92 days pUSD also reached the wallet outside any trade, 935K in all from four addresses, which lifts the total to 2.53M.
These daily payments are 37% of that total and lift the return to 1.13%. The charts and the figures below leave them out unless I say otherwise.
Most of it is not a reward for resting orders. One address sends, each day, 15% of the fee takers paid on RN1's maker fills the day before, and a second sends back half of RN1's own taker fees; together they are 85% of the money. So in most markets part of the fee the exchange charges people like our man is RN1's by the next day, about a quarter of a cent per pair.
A third address pays mostly after days of US football, at no fixed rate, and may be the reward for resting orders, which is a guess; a fourth paid four times and matches nothing in the fills. The chain shows who paid and how much, not what Polymarket calls each payment or why it makes it, and the two rules above come from a first read I have yet to rerun.
One Wallet, Four Jobs
When I split the fills by whether RN1 was resting an order or hitting someone else's, and by whether the game had started, four separate strategies fell out, and they are not equally good at making money. The chart counts fills alone; below I add the 15% share to the quoting rows and the refund to the taking rows, which leaves 152K that fits no row.
Profit to payout by strategy
Quoting during games carries the whole operation, with 795,715 fills earning a thin +0.66% that adds up to +1.13M pUSD, or +1.56M with its payments, and it is so dominant that 97.1% of all RN1's fills happen after kickoff.
Quoting before games loses 283K pUSD, or 244K after its payments, and stays a loss even if every pUSD that fits no row goes to it. The data only shows that it loses, and my guess, still a guess, is that a bid left on the book before kickoff is mostly a gift to whoever reads the team sheet first. Two things lean that way: 227K of the loss comes from fills an hour or more before the start, and RN1's own pre-game taking earns all its profit inside the last hour.
You could reasonably object that the real skill must sit somewhere cleverer than resting bids, and per pUSD you would be right, because when RN1 crosses the spread itself and pays the fee, it earns +1.36% during games and +7.36% before them. In pUSD that is +497K and +243K, or +779K and +277K once half its fees come back.
Its resting quotes, meanwhile, are close to perfectly fair, with profit per share within about a cent of zero in every 10-cent price band, which suggests the bids are not trying to be smart and that RN1 saves its judgment for the rarer moments when it takes.
Profit per share bought, by price band
The Fifty Who Don't Pay
Most drivers pay the toll, and our man in the bar is one of them. 42,885 different wallets filled RN1's resting orders, and against everyone outside its 50 busiest takers RN1 made +1.4M pUSD.
Those fifty took 504K pUSD back, and the busiest of them filled 37,654 of RN1's orders and walked off with 169K. I suspect they have faster prices or better information than RN1's quotes reflect, but fills only show that they win, not why.
The busiest 50 against everyone else
InferredThe busy takers trade with better information or faster prices than RN1's quotes. A copy of this strategy has to spot them and widen or pull its quotes when they arrive.
The second leak is RN1's own doing. A tidy market maker keeps its book flat by buying whichever side it holds less of, and RN1 does that on only 42.87% of the buys that change its balance, so one side tends to pile up. Its quotes do not lean against the pile: they are as tight when its holdings are lopsided as when they are level, and on one sampled evening its usual order size was the same on both outcomes in 95 of 99 markets.
Extra shares on the side it bought more cheaply lost 10.2M pUSD and extra shares on the dearer side won +9.6M, two enormous numbers that nearly cancel into the −486K leak, and the cheap side is usually cheap because its price fell while RN1's bids sat there, which in sport tends to mean it belongs to the team that is losing.
Profit on the unpaired shares
InferredThe outcome RN1 ends up holding cheaply is often the one whose price fell while its bids waited, and a falling price usually means a losing side.
Where it loses is telling too. Tennis is RN1's biggest market and, taken as a whole, its weakest, with ITF tennis, the lower tour, losing 331K pUSD, worse than any other series, while college football quietly brings in +242K.
Profit to payout by series
The ride is not smooth either, since on fills alone RN1 lost money on 39 of 92 days and its worst drop from a running high was 323K pUSD. The daily payments cut those to 34 days and 246K.
Profit to payout, running total
Barranquilla, 01:20 UTC
Which brings me to Barranquilla. At 01:20 UTC on 8 September, a tennis match between Julieta Pareja and Lucrezia Stefanini started there, and RN1 put 1.0M pUSD into it, its largest market by spend, across 67 fills, all but one of them taken at other people's prices.
It looks like the wallet's biggest bet, yet it does not read as a bet on the result, because for two hours RN1 paid 0.50 to 0.51 for both players alike. The match then resolved 50/50, every share paid 0.50, and RN1 walked away 36K poorer. About 24.8K of that loss is fees, and half of those should come back under the refund rule.
Barranquilla: Julieta Pareja vs Lucrezia Stefanini
Each dot is one of RN1's fills: when it happened, the price it paid and, by the dot's size, how many shares. Filled dots are maker fills and rings are taker fills. The lower chart tracks the shares RN1 held of each side.
The wallet's cash is the stranger part. It fell from 1.18M pUSD at midnight to 168K, the low of the 92 days, and the last buy came the second cash first crossed under 200K. For the next 41 minutes RN1 held 886K pUSD of pairs it could have merged into cash and merged none, while its bids elsewhere kept filling.
Four days later it did the same in Guadalajara, in another match that resolved 50/50, and a merge on 13 September freed 269K pUSD from that market. Why RN1 paid over par for both sides I do not know: a view on the match does not fit, a reason tied to volume or a fault would, and both are guesses. On the night it looked most like its screenshots, RN1 was backing nobody.
Somebody Is Holding the Mouse
For a machine running hundreds of thousands of fills, RN1 is strangely relaxed about collecting its money, since 98.65% of what came back to it was paid by Polymarket's auto-redeemer3 a median 12 seconds after each market resolved. It merged pairs early on only 3 days, and its two big sweeps came in the two hours when it had the most cash tied up of the whole period, 1.88M and 1.87M pUSD.
Cash plus open positions ran from 1.19M to 2.66M, with a median of 1.92M, and stood at about 2.07M at each sweep.
pUSD tied up in markets not yet resolved
Both sweeps began soon after cash crossed under 200K pUSD, 102 and 78 seconds after, which reads like a floor built into the bot. Yet in Barranquilla cash sat under that line for 41 minutes and nothing merged.
The first big sweep opened with a single small transaction of 3 merges worth 109 pUSD, then RN1 redeemed shares itself, the only time it did, and the other 20 transactions followed 213 seconds after the test.
Its three withdrawal days open the same way, with exactly 1 pUSD and then round sums of 250K or 500K, between 10:40 and 12:45 UTC, to a bridge. The account is a standard Polymarket proxy wallet, and its orders, in the samples I read, fit only Polymarket's own Python client or a copy of it. To me all of this looks less like a bot crossing a threshold and more like a person testing a script on pocket change before committing the rest, though I cannot prove it and would not stake much on it.
Two fair questions remain. The first is whether this is skill or a lucky summer, and testing the daily results against zero gives a 99% chance the combined gain from fills is more than luck, and about 99.98% with the daily payments, a rough test over 92 days, though I find it more persuasive that the 2.1M pUSD from pairs does not depend on any result at all.
The second is whether you can copy it, and I doubt the screenshots help much: the edge on fills is about 1.5 cents per pair, 37% of the total arrives as daily payments, the pre-game quotes lost money, and fifty takers cut the maker profit from about 1.35M to 0.85M, none of which shows up in any single trade.
Is the profit more than luck?
What I Still Can't See
There is still no order book on chain. A signed order shows up only in the transaction that fills it, so I have posted size and signing time for orders that filled, in samples so far, and nothing on the orders nobody hit, the ones RN1 cancelled, or whether it sets the best price or joins others at it. I do not know what two of the four payers pay for, or why RN1 bought both sides above par in two tennis matches. Beyond the payments, no transfer or fill links RN1 to its first funder, its signing key or the payers, which says nothing about sister wallets, and these are 92 days of an account that has traded since July 2025. The much larger all-time RN1 numbers circulating on X come from somewhere else.
The toll, at least, is now a counted figure and not a floor: 2.53M pUSD, 1.6M of it from fills. What I would want next is the order book itself, because every line here about where RN1 quotes is drawn from bids that filled, and those are the only ones our man in the bar ever meets.
How I counted: every OrderFilled log on Polymarket's exchange where RN1 is maker or taker, 10 Aug to 10 Nov 2026 UTC, Polygon blocks 91,743,315 to 97,042,228. Each share is valued at what it paid out, less what it cost with fees; 206 unresolved markets are left out. The warehouse's own ledger gives +1.7M pUSD over the same markets, against +1.6M from fills alone. The daily payments and cash figures come from the wallet's pUSD transfers; signed orders were read from the calldata of fill transactions, in samples. The payers' rules, every figure that counts the payments, the cash and merge timings, the split by time before the start and the detail of the two tennis matches are first reads, not yet rerun. Figures are rounded. The four strategies are rebuilt as pseudo code in the follow-up.