How Much Can You Earn From Polymarket Liquidity Rewards? 28 Days of Real Data

Whenever we explain how Polymarket's liquidity rewards work, the first question is always the same: fine, but how much? The honest answer is a dataset, not a number. Here is ours: every nightly payout since our account went live on July 21, 2026. The same feed updates on our landing page every day.

1The setup

• Capital: about $3,200 at stake, roughly $2,300 of free collateral plus $900 of open inventory at any given moment. The live split is on the front page.
• Coverage: around 70 distinct markets quoted in a typical 24 hours, selected and rotated automatically by reward pool vs. competition.
• Style: minimum-size, two-sided resting orders inside the qualifying band, repriced around the clock. No directional bets, ever.
• Income counted: everything Polymarket paid the account nightly, meaning liquidity rewards plus maker rebates.

2The raw nightly data

The first 12 days (Jul 21 to Aug 1) were the rough draft: $67.22 total, and only 6 of those 12 nights paid at all while the earliest strategy version found its footing. Since Aug 2 the account has paid 16 nights out of 16. One row per payout:

3What changed mid-August

The eye goes straight to the step-change. Aug 2 to 14 averaged about $15 a night. Aug 15 to 17 averaged about $61. Mid-August we shipped the current generation of the bot: denser coverage, faster repricing, and stricter selection of which markets are worth quoting at all. The last three nights are the new strategy; the flat stretch before it is the old one.

4The costs behind the payouts

Every payout above is gross reward income, and gross is not profit. The cost side is inventory: when resting orders fill, the bot holds real positions, and unwinding them isn't free. Over the 15 payout days ending August 17, rewards paid $349 while unwinds cost $479, so the account finished that window $130 down net. Rewards covered 0.73× the costs. Above 1.0 the strategy prints; below it, the rewards are subsidising the inventory.

That ratio is the entire engineering problem, and it is the number we watch rather than the payout headline. It is also live at our public stats API, so it can't quietly become a stale marketing figure: whatever it reads today, gross and net both come off the same two ledgers. Averages also hide the tails:

5What this scales to (and what it doesn't)

The recent pace, roughly $30 to $60 a night on $3,200, annualizes to a number we won't print, because it would be gross of the inventory costs in section 4 and because reward income does not scale like interest anyway. Your share of each market's fixed pool depends on competition; more capital mostly means more markets, not a bigger cut of the same ones. The proof that scale exists is on-chain: the top nine farming wallets collected $742,677 in 30 days They got there with breadth, uptime and engineering, not by parking a bigger balance.

Every number in this post is from our production ledger, and the losing day stays in. That's the standard you should hold anyone to when they show you yield.

Want the mechanics behind these numbers? Start with the full explainer. Want it running for you? Reward Farmer's waitlist is open.

Corrections

If we're going to claim our numbers are checkable, the fixes have to be public too. Everything below was caught by an outside reader going through the site against our own API.

• 2026-08-17: 15-day total was $30 too high. The front page showed $379.46 collected over the last 15 days. Adding up the same nightly figures on the same page gives $349.46. A typo in a hand-maintained constant. The number is now computed from the nightly series at build time and read live from the API in the browser, so it can't disagree with the chart above it again.
• 2026-08-17: the “2.2× rewards vs. costs” figure was wrong. The site claimed reward income ran about 2.2× inventory costs. Measured properly over the trailing 15 payout days it is 0.73×, which means the account is net negative over that window rather than comfortably ahead. The ratio is now computed in the database from the two ledgers and served by the API, gross and net side by side. Section 4 has the arithmetic.
• 2026-08-17: Aug 14 restated from −$17 to −$72. The losing day was worse than first published: $3.59 of rewards against $75.94 of unwind costs.
• 2026-08-17: capital restated. Earlier copy said “about $2,600 of working capital,” counting only free collateral and ignoring held inventory. Money at stake is closer to $3,200. The live split is on the front page.