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
Parts 1–4 defined the production contract, the slot log, the failure catalog, and the retune order.
Part 5 is the other half of production: saying no.
After TWAP60 and the 150ms taker delay, the tempting move is to add products until the 5-minute book feels easy again. That is how a specialist becomes a junk drawer.
Wallet: 0x12b712029efeee7efc4066e98e4e1b6beb8b1b0b
1. The Thing I Am Protecting
One loop:
BTC 5m
→ 60s Chainlink TWAP vs open
→ 45–55¢ only
→ re-score
→ scratch or redeem
Public shell: twapFeed.ts, signal.ts, engine.ts, clob.ts.
Private layer: calibrated P(up).
Every extra product competes with that loop for attention, logs, and risk budget.
2. I Will Not Add Combos to engine.ts
Combos are RFQ multi-leg parlays.
5m BTC Up/Down is a single-clock CLOB.
Shared logger: fine.
Shared trigger: no.
If Combos ever get research time, they get a separate adapter and a separate cap. They do not inherit 5m scratch timers or TWAP path features.
3. I Will Not Let an Agent Connector Become the Trigger
Exa Connect and similar tooling can query the venue. That is a client.
Production decisions still require:
- official TWAP freshness
- delay-aware execution
- per-slot caps
- skip-on-stale
An agent that “buys Up if BTC is green” is the old CEX-last bot with a new coat of paint.
4. I Will Not Spray Assets to Replace a Weaker BTC Read
ETH / SOL / XRP 5m markets exist. They are not a patch for BTC calibration decay.
Different books. Different depth. Different lead-lag.
Same official-sounding word (“TWAP”) does not make the weights portable.
The public footprint is ~99% BTC for a reason.
5. I Will Not Share 5m Weights With 15m or Kalshi
15-minute Kalshi CF-close and Polymarket 5m TWAP60 are cousins, not clones.
Reuse:
- risk caps
- stale-data hard-fail
- structured logs
Do not reuse:
-
P(up)weights - 47s / 226s timing constants
- scratch thresholds
- open-pin code
Part 4’s rule still holds: change one system at a time.
6. I Will Not Revive Snapshot Logic
No CEX-close labels.
No “use spot until RTDS returns.”
No 30s leftovers on new 5m markets.
If the official series is missing, the slot is skipped. That skip is a feature.
7. I Will Not Treat July Stats as Live Setpoints
The README sample is style evidence:
- mid-band fills
- later sells than buys
- filtered resolved book around 60.7% on 150 markets
Useful to audit identity.
Useless as a frozen production KPI after the delay change.
Copy the shape. Re-estimate on post-Sept 4, 60s-TWAP data.
8. What I Will Keep Adding
Only things that make the existing loop auditable:
- better freshness checks
- signal / submit / fill clocks
- rate-limit backoff
- scratch-spread logging
- Data API reconciliation
- hard tags for
STALE_TWAP,FILL_TAX,DELAY_ROT
That is production work. It is not a new product.
9. Series Recap
- Part 1: Production means the 5m contract, not a longer README
- Part 2: A slot log has to separate signal death from venue friction
- Part 3: Mid-band bots still die from stale pins, fill tax, thin scratches, quota storms
- Part 4: Retune feed → delay/fills → scratch → threshold → weights, on the right calendar
- Part 5: Do not dilute the specialist loop with Combos, agents, or extra coins
One line:
Forecast the official TWAP path, buy it only near fair value, refuse weak tickets — and refuse every extra product that would make that loop un-auditable.
Links
- Repo: https://github.com/abrownfox0/abrownfox001-twap60-prediction-trigger-system
- Video: https://www.youtube.com/watch?v=XzhugRL6BV4
- Live activity: https://polymarket.com/@abrownfox001?tab=activity
Not financial advice. Paper or micro-size first. Saying no is part of the system.
If you have more questions, please feel free to contact me at any time: https://t.me/abrownfox001
My Polymarket Activity: https://polymarket.com/@abrownfox001?tab=activity
#Polymarket #TWAP #BTC #TradingBot #AlgoTrading #PredictionMarkets #QuantTrading #Chainlink
Top comments (0)