Prediction Terminal

Analytics for prediction markets.
Every market, every trader, every trade.

[Prediction Terminal @PredictTerminal]

https://t.co/KlxiXDQngB

[Prediction Terminal @PredictTerminal]

Use the market’s feeType on Gamma (category is empty on anything recent). One per market, so tags don’t matter.

One trap: Gamma only shows the current feeType, and later waivers flip markets to zero_fees. If a fill actually paid a fee but the market reads zero_fees now, it still earned wV, so use the class of the rate actually charged.

[Prediction Terminal @PredictTerminal]

p × (1 − p) per share.

Polymarket’s formula is:

wV = $ size × (1 − entry price) × category weight

so per share it’s p × (1 − p): biggest at 50¢, close to zero near 0 or 1.

I tested it against the actual rebate payouts, and a few things the docs don’t say:
• sells count too, the same way (selling at p = buying the other side at 1 − p)
• the weight follows the market’s fee category (Sports 1.0, Politics/Finance/Tech 1.3, Econ/Culture/Weather 1.7, Crypto 2.3); sports stayed 1.0 even after its fee went up on Jul 10
• fee-free markets, combos and maker fills earn nothing
• your tier uses the 30 full UTC days before today

[Prediction Terminal @PredictTerminal]

Oh, and the exact formulas:

Jun 1 – Jul 2 (what 0x520b paid):
rebate = tier % × (sell fees + each buy fee × that buy’s price)

From Jul 3 (correct):
rebate = tier % × all taker fees

Jul 8–9 top-up (0x71a7):
top-up = Σ (tier % that day × taker fees that day) for Jun 1 – Jul 8, skipping Jun 5–6, minus what 0x520b already paid for those days

Tier for a day = your weighted volume over the previous 30 days (for Jun 1–4, counted from the May 28 launch). Combos excluded.

Example: a buy at 5.6¢ with a $470.24 fee at the 8% tier paid 0.08 × 470.24 × 0.056 = $2.11 instead of $37.62.

[Prediction Terminal @PredictTerminal]

I see some were already answered, so here are the ones still open 🫡

July taker rebates: your rebate should be your tier % of the taker fees you paid. For the first five weeks (fee days Jun 1 – Jul 2), 0x520b undercounted fees on buys: each buy fee only counted at the share price. Buy at 20¢ and only 20% of that fee counted toward your rebate; sell fees counted in full. They fixed it from Jul 3.

The Jul 8–9 payment from 0x71a7 is the top-up for that mistake. For each wallet it’s the rebate you should have got for Jun 1 – Jul 8, minus what 0x520b had already paid. Your tier each day is based on your previous 30 days of weighted volume, combos don’t count, and Jun 5–6 are skipped. For most wallets in the 3%, 8% and 18% lists, it matches to the cent.

That also explains the gaps in the early payments. The Jun 23 payout also covered Jun 1–4 and Jun 7–18, and the Jul 1 payout covered Jun 25, 27 and 29. Rebates for May 28–31 and Jun 5–6 were never paid at all.

The last $0.30 from the Sep 22 batch (0x53d1) is a refund on “Set 1 Winner: Dakotah Bobo vs Corey Craig”. Polymarket posted an on-chain clarification that evening resolving it to Bobo and refunding net losing positions, so all 54 payments are now explained.

Bonus: 0x71a7’s round-dollar payouts on Sep 28 (72 wallets) and Oct 2 (22 wallets) look like payments from the Sep 19 trading-disruption.

[Prediction Terminal @PredictTerminal]

Dug into this. Short version: neither signer is trading profit. 0x71a7 pays outage goodwill credits; 0x53d1 refunds net losses on specific markets Polymarket voided or re-resolved. Both are funded batch-by-batch, to the cent, by the same treasury wallet.

0xd152 isn't a Polymarket contract, it's https://t.co/2jiAPyJgiu (a public multisend), so the signer is what tells you the program:
https://t.co/i6nboue7q0

0x71a7b22d…:
$1,365,350 to 6,743 wallets, 5 days:
• Sep 10: 3,061 wallets, $546K
• Sep 12: 1,335 wallets, $155K
• Sep 20: 3,281 wallets, $554K, the day after the Sep 19 16:54–18:58 UTC incident (https://t.co/25hDpOoEdt). 90% of recipients traded in that window vs ~60% on the same hours Sep 15–18.
• Sep 28 / Oct 2: 72 + 22 wallets, round-dollar amounts, which look like manually reviewed claims
No single market covers more than 13% of recipients, so it's outage compensation, not a market refund.
https://t.co/LX3268NAyQ

0x53d11155…:
$536,449 to 1,693 wallets, 7 days (you missed Sep 12). Each batch clusters on a few markets, and payouts track the wallets' losses on them:
• Sep 11, $332K: ATP Genoa Seyboth Wild vs Ferrari. On-chain clarification: match not played on the listed date, all markets resolve 50-50, "total net losing positions… will be refunded". Payout vs loss on it: r=0.98.
https://t.co/sLQeKXWnTU
• Sep 18, $36K: FIBA Women's World Cup USA vs France (UMA disputed, re-proposed). 35 of 44 recipients who traded it are paid within 1% of their loss.
https://t.co/sVHQzmafIE
• Oct 2, $120K: WTA China Open qualifying Wang vs Bai, hand-resolved 50-50 after repeated disputes. $32K loss-matched; ~$88K not yet attributed (incl. 208 wallets paid exactly $1.03).
https://t.co/yIwoZWXhqH
https://t.co/zloHdEYeKh

Still open: 0x53d1's Sep 12 ($27K, mostly 5/15-min crypto Up/Down traders), Oct 6 ($21K, exact matches on many unrelated markets, e.g. NFL season props) and two tiny batches (Sep 22 $149, Oct 9 $238), which line up in time with on-chain "refunded" clarifications.

0x9653915a is maker rebates, as you said: $127 to 26 wallets on Sep 14

[Prediction Terminal @PredictTerminal]

@mrjeffu Kalshi was right to fade him. Anne Carson just won, and she last traded at 13¢ before the announcement. Only Can Xue was higher

[Prediction Terminal @PredictTerminal]

The two months Kalshi priced parlays closest to fair were the first two months anyone could buy one.

Combos opened to everyone on 15 December. Through December and January the markup over fair value nearly vanishes, and from February it sits near 7% and hasn't left.

[Prediction Terminal @PredictTerminal]

Buyers of ordinary Kalshi contracts paid about half that, because the contracts they buy cost more.

Read full article: https://t.co/nYeBoDcsdH

[Prediction Terminal @PredictTerminal]

https://t.co/KlxiXDQngB

[Prediction Terminal @PredictTerminal]

Buy a contract on Kalshi at 95 cents and the fee costs you 0.35% of your money. Buy one at 2 cents and it costs 6.9%.

The fee is worked out from the contract itself, so the less you paid for it, the bigger a share the fee takes.

Parlays are cheap by design. That's what stacking legs does to the price. The average parlay dollar goes in at 31 cents, where the fee takes 4.8%. Across every combo Kalshi has sold that comes to $140m.

[Prediction Terminal @PredictTerminal]

We can't say why. Quotes on Kalshi are private between the requester and each maker, and by Kalshi's own docs the makers can't see each other's, so there's no way to count who was bidding at 4am.

Read full article: https://t.co/qmy9cngdHr

[Prediction Terminal @PredictTerminal]

https://t.co/KlxiXDQngB

[Prediction Terminal @PredictTerminal]

The worst time to buy a parlay on Kalshi is the evening. The best is mid-morning.

In July it ran 4.9% over fair value at 10am and 19.2% at 10pm. That holds when you compare like with like: inside matched cells of the same series, leg count and price band, 45 of 47 rise from morning to evening.

Overnight, when almost nothing is trading, it sits around 7%.

[Prediction Terminal @PredictTerminal]

That markup plus Kalshi's taker fee comes to 10.6% of everything a parlay buyer stakes. Sportsbook parlays hold about 17.8% in the six states that break parlays out separately. Kalshi is still the cheaper place to buy one.

Method and caveats: https://t.co/qmy9cngdHr

[Prediction Terminal @PredictTerminal]

https://t.co/KlxiXDQngB

[Prediction Terminal @PredictTerminal]

Every leg you add to a parlay makes the price worse, and it compounds.

On Kalshi a two-leg parlay costs about what its legs cost. Four legs, 2% over. Ten legs, 14% over. Twenty-five legs, 55% over.

We checked 55 million parlays against the price of their own legs at the moment each one traded. Eleven months, all from Kalshi's public data.

[Prediction Terminal @PredictTerminal]

Polymarket can get much bigger from here

- Perp dex launch
- Own chain with native yield
- Perp dex also mean smart wallet under the hood so people store their assets on chain means Spot built in DEX also has sense.
- DeFi Layer on their chain is also possible because it unlocks capital efficiency
- Leverage trading for prediction markets
- Cross positions so you can hedge your perp position with prediction markets

I can keep going but even with these all implemented Polymarket with its user base and marketing / distribution could be worth more

But agree here acquisition is very unlikely

[Prediction Terminal @PredictTerminal]

👨🏾‍💻 We are cooking and preparing for the LAUNCH

We’ll be revealing the key features of our platform

FOLLOW US so you don’t miss what’s coming

👉🏼 And join the WAITLIST to get PREMIUM access at launch - https://t.co/mZ13sbFXlq

Prediction Markets Will Win 🏆

[Prediction Terminal @PredictTerminal]

Steam Log In on @Polymarket 🤔

Сapturing the gamer market, nice move btw

[Prediction Terminal @PredictTerminal]

@Nostroah @realDonaldTrump crime season

[Prediction Terminal @PredictTerminal]

@trylimitless Sounds fair