DEV Community

ABROWNFOX001
ABROWNFOX001

Posted on

Inside My Live BTC 5m Polymarket Bot — Architecture, Mistakes & Edge

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

I’ve been running a live automated BTC 5-minute Up/Down engine on Polymarket under @abrownfox001 (wallet 0x12b7…1b0b).

This is not a 94–99¢ convergence farm.

It is a directional system built for the coin-flip band (~45–55¢): enter only when there is a real edge, hold only high-conviction legs, and scratch everything else near breakeven.

The edge is in the stack — signal construction, entry timing, and real-time risk culling.

What Runs Every Slot

One engine instance per active btc-updown-5m-{ts} market:

1. TWAP feed

Continuous Chainlink TWAP stream (60-second window — the current settlement reference for 5-minute crypto markets). No reliance on a single CEX last price or end-of-slot snapshot.

2. Open reference

TWAP value locked at slot open. This is the price-to-beat.

3. Private signal

Calibrated P(up) from:

  • live TWAP path vs open reference
  • short-horizon CEX lead-lag
  • flow / microstructure features
  • time remaining in the slot

4. Entry filter

Trade only when:

  • |P(up) – 0.5| clears the edge threshold
  • ask is inside the fair 45–55¢ zone

Clips stay small ($50–175). Entries are staggered (median ~47s after open) so the bot does not lift a thin early book.

5. Re-score loop

Every 10–30 seconds the position is re-evaluated.

6. Hybrid exit

  • Conviction decays → taker scratch back to flat (observed buy ~0.504 / sell ~0.503)
  • Conviction holds → ride to resolution and redeem at $1

The Six Mistakes This Design Was Built to Kill

1. Using raw CEX direction as the signal

“Binance is up → buy Up” is mostly noise on a market that settles on Chainlink TWAP.

Design response: Every probability is anchored to the live TWAP path versus the exact open reference. The model targets the settlement rule, not a random mid.

2. Aggressive early size

Buying the first 5–10 seconds often means paying 53–55¢ for something worth ~50¢.

Design response: Staggered entries. Wait for liquidity near fair value. Observed average entry on held legs sits near 0.503.

3. Holding every ticket to resolution

A raw 55–58% signal still loses money if the weak losers are never cut.

Design response: Continuous re-scoring + near-zero-cost scratches. Only the filtered high-conviction book reaches settlement. That is how held-side win rate can approach ~60.7% even when the raw signal is lower.

4. Ignoring latency and adverse selection

Slow loops buy the move after it has already been priced in.

Design response: Tight TWAP → signal → order path. Fill quality is treated as part of the edge, not an afterthought.

5. Spreading across too many markets

Multi-asset, multi-duration bots accumulate hidden regime risk.

Design response: Deliberate specialization — almost exclusively BTC 5-minute. Depth over breadth.

6. Entry engine with no real exit engine

Most “directional” bots are just triggers with hope as the exit.

Design response: Scratch logic is a first-class edge. Weak capital is recycled immediately; only quality exposure is allowed to mature.

Sample Results (Public Resolved Set)

Metric Value
Resolved markets (sample) 150
Held-side win rate 60.7% (91W / 59L)
Avg entry (held legs) ~0.50
ROI on capital at resolution ≈ +24.5% (sample)
Fills in 45–55¢ band ~90%

These numbers describe the filtered book that actually reached resolution, not every temporary position.

High-Level Control Loop

60s Chainlink TWAP
        ↓
P(up) vs slot-open reference
        ↓
Enter only in 45–55¢ if edge is real
        ↓
Re-score every 10–30s
        ↓
Scratch near flat    or    Hold → Redeem
Enter fullscreen mode Exit fullscreen mode

What Is Public vs Private

Public

  • Architecture and execution pattern
  • Wallet and on-chain footprint
  • TypeScript reference for discovery, routing, state machine, and scratch flow
  • Live profile activity

Private

  • The calibrated 5-minute signal (the actual moat)

Repo: https://github.com/abrownfoxbot/abrownfox-trading-system

Why This Style Can Work at ~50¢

At 99¢, fees and tiny edges dominate.

At ~50¢, the binding constraints are different:

  • Signal accuracy
  • Latency
  • Fill quality
  • Discipline to cut weak tickets

This system is optimized for exactly those constraints.

Final Notes

  • Settlement rules evolved (5m markets now use 60s TWAP). Any bot must stay aligned with the official window.
  • Short-horizon markets remain high-risk.
  • Past sample results are not a guarantee of future performance.
  • Paper trade or micro-size before scaling.

Live profile: @abrownfox001

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

Not financial advice.

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 #BTC #TWAP #TradingBot #AlgoTrading #PredictionMarkets #Chainlink #QuantTrading
Enter fullscreen mode Exit fullscreen mode

Top comments (0)