The Updates That Actually Break Bots

GitHub: https://github.com/abrownfox0/abrownfox001-twap60-prediction-trigger-system

YouTube walkthrough: https://www.youtube.com/watch?v=XzhugRL6BV4

Live profile: https://polymarket.com/@abrownfox001

This 5-part series is about the current Polymarket stack as of early September 2026 — not the announcement-era story from July.

If you run a 5-minute crypto bot, three layers moved at once:

Settlement
Market access / rate limits
Client surface (CLOB, SDKs, data APIs)

Miss any one and the strategy post still looks fine while production silently rot.

Part 1 — Map of the live platform changes (this post)
Part 2 — TWAP60 settlement and what bots must rewire
Part 3 — CLOB V2, volume-tiered rate limits, unified SDKs
Part 4 — Data, Combos, infra: the operational changes
Part 5 — What I am changing in the abrownfox001 engine next

The Change Map That Matters

Layer	What shipped	Why a bot cares
Crypto Up/Down settlement	Snapshot → Chainlink TWAP (Aug 7)	Last-tick models die
5-minute TWAP window	30s → 60s (markets since Aug 14)	Open pin + labels must move with it
TWAP delivery	RTDS + SDK TWAP helpers	Official path is streamable
CLOB	V2 only; V1 clients are dead	Signing, fees, collateral changed
Rate limits	Volume-tiered token buckets (Aug 6)	Spam loops get clipped
SDKs	Unified TS/Python DeFi SDKs still moving (updated Sep 7)	Old packages go stale fast
Data / Gamma	Faster positions/PnL, approvals endpoint, serving-layer fixes	Reconciliation and charts drift if you cache old shapes
Combos	Collateral Return live; combo status now in SDK models	Capital and position state get more complex
Perps	Separate stack, WS concurrency + overload shedding in Sept	Do not mix perps assumptions into 5m CLOB code

The useful sentence:

Polymarket did not ship one feature. It changed the price the market pays, the pipe you use to hear that price, and the quota you get for acting on it.

Why This Series Starts With a Map

Most “latest update” posts pick one headline: TWAP, or rate limits, or Combos.

A live 5-minute engine has to absorb all three.

My public system is a BTC 5m specialist:

official TWAP path versus slot open
mid-band 45–55¢ entries
scratch decaying tickets
redeem the filtered book

That design only stays coherent if:

the feed is the current 60s series
orders are CLOB V2
the bot respects volume-tiered limits
logs still reconcile after Data API / indexer changes

What Is Already Live vs What Is Still Moving

Already live and binding

TWAP resolution on short crypto Up/Down
60s lookback on current 5m crypto markets
RTDS TWAP stream
volume-based CLOB rate limits
CLOB V2 as the only production order path

Still moving

unified SDK identifiers / protocol-version fields
Combo status exposure in client models
Data API serving performance
Perps gateway concurrency rules
docs versus older README comments that still say 30s

If your code matches a July blog post and not the live market’s resolution source, you are researching a retired contract.

The Bot Question for the Rest of the Series

Do not start with “is Polymarket better now?”

Start with:

What number settles the market I trade?
Can I stream that number with freshness checks?
Can I submit and cancel without burning the rate bucket?
Does my client still speak V2?
Can I reconcile fills after the data-layer changes?

Parts 2–4 answer those in order.

Part 5 maps them onto the public abrownfox001 repo.

What Part 2 Covers

Settlement is the first break.

Part 2 goes deep on:

30s → 60s for 5-minute markets
why open and close must use the same TWAP family
RTDS vs snapshot leftovers
the exact state a 5m engine has to store now

Polymarket #TWAP #TradingBot #BTC #AlgoTrading #CLOB #PredictionMarkets #Chainlink

DEV Community