LZLZL/Prediction markets/Rebates
FREEJUDGE IT B · RebatesMechanism

Four different things
all called "rebate"

2026-08-21 · rules per official docs; checking date noted

Polymarket runs four entirely separate return mechanisms. English calls them all rebates; Chinese calls them all 返利. They are four rulebooks, four pools, four payout times, with unrelated eligibility. A guide that does not name which one is essentially saying nothing.

1Four faces

#NameWho gets itEarned byPaid
Maker rebatethe posteryour quote being fillednext day ~00:45Z
Taker ladderthe taker30-day rolling weighted volumedaily 00:00Z
LP rewardsthe posterquote quality — no fill neededdaily ~00:45Z
Referralthe referreryour recruit paying taker feesdaily 00:00Z

① and ③ both pay posters, and their triggers are exact opposites: ① needs a fill, ③ does not. That alone sends two strategies in opposite directions — §5.

2Prerequisite: the taker fee

Three of the four are funded directly or indirectly by the taker fee, so price it first. Posters pay nothing; the fee sits entirely on the taker:

fee = 0.07 × p × (1 − p) × shares p = your fill price (0–1)

Counter-intuitively the fee is heaviest at p = 0.50 and cheaper toward both tails. 50¢ costs $1.75 per 100 shares; 95¢ costs $0.33.

⚠ The 0.07 is not in the documentation

I recovered it by reconciliation: predict a market-making wallet's full-day rebate from the formula, compare against what landed. Predicted $45.23, received $46.5897% (the gap is 43 unrecorded minutes at the start).

Also: the CLOB config's maker_base_fee/taker_base_fee read 1000. That is a raw config field, not a rate. Do not compute with it.

3① Maker rebate: 20% of the fee flows back to the filled poster

When your quote is hit, 20% of that taker's fee returns to you — about +0.25¢/share, settled daily around 00:45 UTC.

The load-bearing sentence: it exists only on fills, and is unrelated to how long you quoted. So wanting this money means wanting to be filled — and wanting to be filled means accepting adverse selection, where the fills you get are the ones you did not want. This subsidy is a fixed 0.25¢/share racing a variable cost.

4② Taker ladder: tiered on 30-day volume, automatic

No registration and no code — hit the volume and you tier up. The unit is not notional but a weighted volume:

wV = stake × (1 − entry price) × category weight taker fills only · maker fills do not count at all
CategoryWeight
Crypto2.3
Econ · Culture · Weather1.7
Politics · Finance · Tech1.3
Sports1.0
Geopolitics0 (excluded)
Tier30-day wVRebatePromotion bonusEffective uplift
Bronze$2k3%+$100.05pp
Silver$20k8%+$500.14pp
Gold$200k18%+$2500.32pp
Platinum$1M32%+$1.5k0.56pp
Diamond$4M44%+$7.5k
Obsidian$10M50%+$25k

It returns a percentage of the fees you paid, not of notional — the column people misread. Paid daily at 00:00Z straight into balance, $1 minimum.

5③ LP rewards: paid without a fill — but not on every board

Scored on quote quality — how close to mid, how much, how long — filled or not. Settled daily around 00:45Z.

clobRewards:[] does not mean the board has no rewards

An empty array from the events endpoint does not imply "no pool". The authoritative source is the CLOB's sampling-markets endpoint. Likewise rewardsMinSize/maxSpread are metadata — their presence does not mean money is paid.

Measured counter-example: on crypto up/down, live 1h and 15m boards return empty from /rewards/markets, and 5m returns a placeholder (rate_per_day: 0.001). In other words market making up/down earns essentially none of ③ — that lane's only subsidy is ①.

6① and ③ pull against each other

① Maker rebate③ LP rewards
Triggermust be filledpays without a fill
You wanttight quotes, get hitqualifying quotes, ideally not hit
Main riskadverse selectionbad fills when swept
Availableany board with a taker feeonly boards in the reward programme

So "market making" here is two different businesses depending on which you are chasing — and why copying someone's approach so often fails: they are collecting ③ while you think you are collecting ①.

7④ Referral: rewritten on 2026-05-28, so most guides are stale

Most important first: current terms date from 2026-05-28. Before that it was 30% for 180 days; it was cut to 10% for 30 days. Large numbers of guides in both languages still quote the old figures. As checked in 2026-08:

ItemCurrent
Who is paidthe referrer receives 10% of the recruit's net taker fees; indirect referrals pay 5%
The recruit receivesnothing. "Signup bonus" claims on third-party sites are marketing
"Net fee"venue receipts = taker fee − the recruit's own ladder rebate (so ② erodes ④)
Duration30 days from signup, or until they reach Platinum
Eligibilitythe referrer needs $10k lifetime volume before earning
Binding?r=<code> → 30-day cookie → locked at web signup. New accounts only
⚠ Self-referral is prohibited in writing

The terms explicitly forbid self-referrals, referring accounts you control, and inauthentic trading, and reserve the right to disqualify and claw back.

Worse, the exposure is collective: the terms permit removing an offender from every reward programme — so chasing ④ can cost you ① ② and ③ as well. And detection is close to free: transfers between wallets one person controls are public on-chain. Limited upside, unlimited downside.

If you are opening an account

Polymarket supports connecting a wallet or signing up by email — the email route generates a custodial wallet, so no prior crypto experience is needed.

If you are going to sign up anyway, you can use my referral link: 👉注册Polymarket. Per §7 above, the recruit gets neither a discount nor a bonus, so it makes no difference to your costs either way. Consider it encouragement to keep writing these. Not using it is entirely fine — nothing here gets shorter because of it.

Practical notethe binding happens at the moment of web signup and only for new accounts. Existing accounts cannot be attached retroactively.

8Add them up: what is it worth

Measured on one basis — effective uplift, restated as points off each trade. The fee is about 1.75pp at p≈0.5; ② contributes its tier rate × 1.75pp, and ④ roughly 10%:

② + ④ fully maxed ≈ 0.17–0.18pp against a fee line of 1.75pp

Both maxed only moves the required gross edge from about 1.75pp to about 1.5pp. Rebates are a multiplier, not an addend — they improve a positive strategy and cannot rescue a negative one.

I walked into this wall with real money

A real-money up/down taker line: about 496 windows, roughly 41 hours, net −$496.51, of which fees were $145 and ladder rebates actually received were $0.

Basisthe ledger recorded the $0 and not the reason. It may relate to the line running only ~41 hours (the ladder is a 30-day rolling measure, paid daily, $1 minimum), but I did not verify that, so I am not stating it as a conclusion.

Either way it does not change what matters: that line's problem was not "just short, rebates would fix it". The edge was absent. Rebates cover about 10%; the gap was 100% — an order of magnitude. Maxed out, it stays negative.

9Four payout times

TimeWhat lands
00:00Z② taker ladder · ④ referral (pUSD, straight to balance)
~00:45Z① maker rebate · ③ LP rewards
paid as batched transfers, roughly 400 per batch
⚠ None of these four appear in "trading profit"

They are transfers, not trade P&L. The leaderboard's /profit field therefore contains no subsidy income at all. Real net income requires trading P&L + total received, queried separately and summed.

This makes you systematically underrate subsidy-driven players — and systematically overrate your own strategy, if you counted subsidies as strategy income.

10One line to keep

Next time you see "Polymarket rebates", ask which one: paid on fills (①), tiered on volume (②), paid for resting (③), or paid for recruiting (④). Four answers, four completely different businesses.

EvidenceCheck it yourself

Fee formula predicted a day of maker flow with 0.07·p(1−p)·shares against next-day receipts: $45.23 predicted / $46.58 received (97%).
Ladder and weights official taker-rebates documentation, tier by tier, checked 2026-08-12.
Referral rules official referral-programme page (terms effective 2026-05-28) plus help-centre entries, checked 2026-08-12; note there is a separate perps referral programme with different terms.
③ is zero on up/down measured returns from /rewards/markets for live 1h / 15m / 5m boards.
−$496.51 with $0 rebates my own live ledger. The ledger did not record why rebates were zero.
Checked 2026-08. These rules have been revised before and will be again. Verify the current terms before betting.

NextWhere to go

B · REBATES
The taker ladder in full
B · REBATES
Adverse selection: the real cost of quoting
F · METHOD
Rebates are a multiplier, not an addend
E · UP/DOWN
The line that bet on rebates, and got none
This is an educational and research record. It is not investment advice, promises no returns, and offers no personalised trading recommendations. Rules and API behaviour are per the official documentation; this page states when it was checked and both can change without notice. Prediction markets are restricted or unavailable in some jurisdictions — confirm your own before taking part.