Pendulum Flow

Wealth management. Markets since 2003. Current research S&P holdings to 1957 & Polymarket database growing @ 2.3 billion rows / day. Data psychosis.

[Pendulum Flow @PendulumFlow]

Attention Polymarket Data Geeks. I have built military grade orderbook capture and provided it for free. 🪖

Growing at 2.3 Billion rows per day. Check it out and pls share.

[Pendulum Flow @PendulumFlow]

Common Vulnerabilities and Exposures. AI is simply better at coding than humans now.

[d1namit @0xd1namit]

If you’re using Polymarket’s native ref system, you’re leaving a lot of money on the table

Just look at how it compares to the program on Dub

On top of that, they’ve started running bonus programs, like a $10k prize pool for the top-performing Perps affiliates

DM me and I’ll onboard you

[Pendulum Flow @PendulumFlow]

Polymarket Orderbook Archive Update

New London, bare metal dedicated server is unparalleled. In v3.1 I expect this to be the primary witness. Potentially room for another ~0.3% capture improvement with tuning.

Rolling out some enhancements to the other witnesses that should make them more stable. It has only been thanks to the 5 witness VPS merge that the capture has stayed > 99.85% at times.

[Pendulum Flow @PendulumFlow]

Attention Polymarket Data Geeks. I have built military grade orderbook capture and provided it for free. 🪖

Growing at 2.3 Billion rows per day. Check it out and pls share.

[Pendulum Flow @PendulumFlow]

@PredictTerminal Nice :) Great tip. Thanks!

[Pendulum Flow @PendulumFlow]

Worked some of these out:
0x8a80356b6304a08c24da30b0cf0d85b6907824ee (and its predecessor 0x1510565e…): referral rewards. Daily payments to referrers on their referees trading, which is why most recipients have no trades of their own. (Happy to be corrected.)
0x71a7b22d671c93a7a40edff2ffed52dd64ed3235, 8–9 Jul: partly solved.

- One list is the taker level-up bonuses ($10–$25,000 for reaching the 3–50% tiers).

- The rest are taker rebates paid by tier: 3% in one list, 8–18% in another, 32–44% on 8 Jul.

Still open, if anyone knows:

What period and formula the 8–9 Jul taker-rebate amounts covered. They aren't a simple tier × fees, and the regular taker rebates from 0x520bf77d… for 20 Jun – 3 Jul don't follow tier × fees either.
0xc0126535cbc6f70fc71d41bbc266d76321159efe: weekly ~$12.5k, later ~$25k, 9 Nov 2025 – 5 Jan 2026, to a fixed group of 11–35 wallets.
0x53d111557bc0d48ec540fbdca14a1f148cabdb40, 22 Sep: one Disperse tx (0xb1d7b542…8bd1), $149.40 to 54 wallets. Looks like a refund, but can't tie it to a market.

[Pendulum Flow @PendulumFlow]

More data psychosis...

Anyone tell me what there Polymarket payments were for?

1. 0x71a7b22d671c93a7a40edff2ffed52dd64ed3235, 8–9 Jul 2026 ~$4.2M to ~34.5k wallets in a few lists. One list is clearly the taker level-up bonuses (10–25,000 by tier), and the others go to wallets by taker-rebate tier (3%, 8–18%, 32–44%). What period and formula did the non-bonus amounts cover?

Related: the taker-rebate payments from 0x520bf77d9d34c34a6a9723f50e1dcb887ed238c5 for 20 Jun – 3 Jul don't match tier × fees the way later days do (and a few days were skipped). Was there a different rule in that first fortnight?

2. 0x8a80356b6304a08c24da30b0cf0d85b6907824ee Daily pUSD payments since 25 Jun 2026 (~$2.5M total), to a stable group of ~3.7k wallets, many with no trading activity. Most were previously paid by 0x1510565e. What program is this?

3. 0xc0126535cbc6f70fc71d41bbc266d76321159efe Weekly payments of ~$12.5k, later ~$25k, from 9 Nov 2025 to 5 Jan 2026, to a fixed group of 11–35 wallets. What program was this?

4. 0x53d111557bc0d48ec540fbdca14a1f148cabdb40, 22 Sep 2026 One Disperse tx (0xb1d7b542a7b8ccd9c04f777dbe05eb701b8b62240199d99b1d61528cf1fe8bd1), $149.40 to 54 wallets. This address's other batches look like market refunds, but I couldn't link this one to a market.

[Pendulum Flow @PendulumFlow]

Very impressive thread of analysis from @PredictTerminal 💪🧠🫡

[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.

[Pendulum Flow @PendulumFlow]

Perfect, that's the missing piece, thank you. I'll recompute every wallet's daily tier from fills and post the cent-match rate for 0x520b and the Jul 8/9 lists.

Anything I can do to return the favor?

Where do you read a market's fee category from? The Gamma market fields (feeType / category), or the fee schedule at fill time? Some markets carry several tags, and I want the 1.0/1.3/1.7/2.3 weights to match yours.

[Pendulum Flow @PendulumFlow]

@polyfees checked: no. For four of the 54 recipients, Polymarket's /activity feed around the payment (22 Sep 22:57 UTC) lists their rewards, maker/taker rebates and yield, but not the 0x53d1 refund. It only exists as a plain pUSD transfer on-chain, so anything built on the activity feed misses it.

@convexlyapp I re-ran it from chain capture: 48/54 match the wallet's 19 Sep fees on the four markets to the cent. The FeeCharged event already holds the combined (platform + builder) charge, e.g. 5.848330 on 0xf7f23e94. One extra detail: refunds stop at a cut-off. 0x233b's later fill that day (fee 1.68) wasn't refunded, only the 5.85 one.

[Pendulum Flow @PendulumFlow]

This is brilliant, thank you. Seriously. It fits what I'd found (the Jul 8/9 lists line up with the 3/8/18 and 32/44 % tiers). I'll check the formula against every wallet's fills and post the cent-match rate. One question: for the tier, how is "weighted volume" weighted? By price, by min(p, 1−p), or something else?

[Bala 🧤 @BalaiBB]

a critical vulnerability was just discovered in PHP. the language that runs 80% of the internet

severity score: 9.2 out of 10. that’s about as bad as it gets

what can an attacker do with this? take full admin control of a server. steal entire databases. passwords, credit cards, payment histories. plant malware and use the server to attack other systems

but here’s the terrifying part that most people missed

the moment the patch drops tonight attackers are expected to use AI to reverse engineer what was fixed within 15 minutes

15 minutes. that’s the window between the fix going live and automated attacks launching worldwide

every online store, booking site, and corporate system running PHP that doesn’t update in those 15 minutes becomes a target

this isn’t just a developer problem. if you’ve ever signed up to a website, entered your card details, or created an account online your data is sitting on servers that might be vulnerable right now

and someone just publicly dropped the working exploit on twitter before the patch was even ready. gave every attacker on earth a head start

engineers worldwide are pulling all-nighters tonight to patch their systems before the attacks begin

change your passwords on any service that matters to you. enable 2FA everywhere. and watch out for fake “maintenance” or “password reset” emails over the next few days

the internet is about to have a very long night

[Pendulum Flow @PendulumFlow]

OK, not happening. Getting the gap filled back to 18th Aug though 🥲

[Pendulum Flow @PendulumFlow]

@beniduboss What a day to be alive!

[Pendulum Flow @PendulumFlow]

@convexlyapp Nicely done 🫡

[Pendulum Flow @PendulumFlow]

This is extreamly impressive work!

[Convexly @convexlyapp]

On the 22 Sep batch (0xb1d7b542...8bd1), we reconstructed all 54 payments totaling 149.40 pUSD. The strongest explanation is mixed reimbursement of selected fees and purchase costs, rather than one market.

52 recipients received 146.97 pUSD. Across 107 selected fills, the recorded fee components total 146.985350 before per-wallet rounding, on four markets:
• Russia military action against an EU country by 30 Sep and 31 Dec 2026.
• EU country blames Russia for a new sabotage attack by 31 Oct and 31 Dec 2026.
One combined builder charge is included, so this is not entirely platform fees.

Archived Polymarket event pages show those exact conditions with fee coefficient 0.04 on 19 Sep and zero on 20 Sep. That is a fee-curve coefficient, not a flat 4% charge. The captures do not pin down the effective change or prove an incorrect charge.

The remaining payments:
• 2.13 pUSD matches 42.54736 Corey Craig match-winner shares × 0.05 = 2.127368, in Bobo/Craig.
• 0.30 pUSD has two matching candidates: Craig first-set (5 × 0.06) or Maltsev vs Pytsko (10 × 0.03). The first-set market settled after the payment; Maltsev settled losing before it.

The combined final ITF main and qualifying draws contain both Bobo and Craig but no pairing between them. That supports investigating the fixture, not declaring it canceled or wrongly resolved.

Rounding matters: decimal half-up matches 53/54. One 0.495 basis was paid as 0.49. Tested half-down and binary-rounding variants match 54/54, but do not identify the payer's implementation. The 0.30 attribution and payment purpose remain unconfirmed.

We have the 54-wallet comparison and transaction/log references prepared. This analysis is specific to the September batch; it does not explain the July or weekly payments.

Military-market archives, before and after:
https://t.co/ride8UOzZu
https://t.co/Ixx5VPgXA5
ITF final main and qualifying draws:
https://t.co/XtjZzxX3j8
https://t.co/5BGsmIBFKk

[Pendulum Flow @PendulumFlow]

This is excellent, thank you. I checked it and it holds. GOAT! The fees on those four markets were only charged on Sep 19, and 45 of the 54 payments match each wallet's Sep 19 fees on them to the cent.

Curious how you found it: was it the archived event pages that pointed you to the fee change, and how did you pick the 107 fills?

[Pendulum Flow @PendulumFlow]

Agreed, haven't gone deep but my scan says market maker, not signal bot:

- ~83% of its buying is resting maker orders.
- never sells, merges matched Up+Down pairs instead.
- trades ~700 up/down markets a day across every asset and timeframe.
- maker rebates roughly cover its fees.

Spread + rebate, not %K/%D.

[Pendulum Flow @PendulumFlow]

@0xpinner Free egress yes :)

[Pendulum Flow @PendulumFlow]

Orbook archive is getting hit by distributed scraping swarms from Vietnam, Taiwan, and China spreading requests across thousands of IPs to dodge the per-IP rate limit I had to implement. FU very much.

[Pendulum Flow @PendulumFlow]

@crime_xbt Very nice of you say. Can do it without funding :)