frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

Open in hackernews

Risk Management and Hedging in a Polymarket Dual-Side Arbitrage Bot

1•Benjamin-Cup•1h ago
Dual-side arbitrage in prediction markets like Polymarket looks simple:

Buy both outcomes below $1 → hold → guaranteed profit.

But in real trading, this breaks without one key component:

Risk management.

The real risk is not price direction — it’s execution imbalance.

The Real Risk: Unpaired Exposure

In theory:

Buy UP at 0.47

Buy DOWN at 0.46

Total = 0.93 → locked profit.

In practice:

One side fills

The other doesn’t

Time runs out

If the market resolves before completing the pair, you face a full loss.

So the focus must be:

Resolve imbalance, not just find price edges.

Core Risk Management Principles 1. Limit Unpaired Inventory

Never allow unlimited exposure on one side.

Track unpaired tokens and stop new buys when imbalance exceeds a threshold.

Result: Prevents runaway risk.

2. Time-Based Forced Hedging

The most important safeguard.

If the opposite side isn’t filled within a set time:

Force-buy it at a more aggressive price.

This reduces profit slightly but guarantees completion.

3. Progressive Hedging

Don’t wait for perfect fills.

As imbalance grows:

Hedge small portions early

Increase hedge size over time

Result: Avoids costly last-second execution.

4. Liquidity-Aware Execution

Not all opportunities are real.

Check:

Orderbook depth

Spread

Fill probability

If liquidity is weak, reduce size or skip.

5. Time Window Control

Markets become unstable near resolution.

Best practice:

Trade only between defined times (e.g., 300s → 90s)

Avoid late entries

Complete all pairs before final window

6. Emergency Hedging

Near market close:

Close all unpaired positions immediately.

Even if it means crossing the spread and accepting worse prices.

7. Order Escalation

Limit orders improve profit but reduce fill rate.

Use a layered approach:

Passive limit

Retry with better pricing

Aggressive fill if needed

Result: Higher completion rate.

8. Edge Validation

Not every “cheap pair” is profitable.

Before entering:

Include slippage and execution cost

Reject weak setups

9. Inventory Skew Control

When imbalance appears:

Prioritize buying the missing side

Reduce buying the dominant side

This naturally rebalances positions.

The Role of the Hedge Manager

A proper system separates hedging logic.

The hedge_manager handles:

Forced hedging

Partial hedging

Emergency closing

Execution escalation

Its goal:

Ensure every position becomes a complete pair.

Final Insight

Dual-side arbitrage is not truly risk-free.

In reality:

Fills are uneven

Liquidity disappears

Time is limited

The real edge is not finding opportunities —

It’s how effectively you neutralize risk.

A successful bot:

avoids getting stuck

resolves imbalance fast

exits fully hedged

That’s what makes the strategy work.

Contact

Email: benjamin.bigdev@gmail.com

Telegram: https://t.me/BenjaminCup

X: https://x.com/benjaminccup

GitHub: https://github.com/Gabagool2-2/polymarket-trading-bot-python

The Human Skill That Eludes AI

https://www.theatlantic.com/technology/2026/03/ai-creative-writing/686418/
1•droidjj•24s ago•0 comments

Bare: Small and modular JavaScript runtime for desktop and mobile

https://github.com/holepunchto/bare
1•nateb2022•44s ago•0 comments

Enslaved potter David Drake's work still speaks for him today

https://www.csmonitor.com/The-Culture/Arts/2026/0302/reparations-slavery-dave-the-potter
1•Tomte•2m ago•0 comments

Who Profits from Prediction Markets? Execution, Not Information

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6191618
1•paulpauper•2m ago•0 comments

Om Malik: Lobster Boil

https://om.co/2026/03/16/lobster-boil/
1•rmason•2m ago•0 comments

I Built Six Simulations to Find the Optimal Engineering Team Size in the AI Era

https://fastnotenough.com/
1•kucukkanat•4m ago•0 comments

Elbowing In

https://www.cjr.org/feature/olympia-washington-press-corps-statehouse-credentials-lawsuit-brandi-...
1•Tomte•5m ago•0 comments

Openpilot 0.11 - first robotics agent fully trained in a learned simulation

https://blog.comma.ai/011release/
1•LorenDB•6m ago•0 comments

The Small Web Is Tricky to Find

https://matduggan.com/the-small-web-is-tricky-to-find/
1•speckx•6m ago•0 comments

Show HN: AI-native memory, recall and reminder on CLI – 100% local with Ollama

https://github.com/KunalSin9h/yaad
2•knlsn•7m ago•2 comments

How to lose AI arms race

https://www.thefp.com/p/how-to-lose-the-ai-arms-race
1•paulpauper•8m ago•0 comments

Video Encoding and Decoding with Vulkan Compute Shaders in FFmpeg

https://www.khronos.org/blog/video-encoding-and-decoding-with-vulkan-compute-shaders-in-ffmpeg
2•AshleysBrain•8m ago•0 comments

Twenty years of Amazon S3 and building what's next

https://aws.amazon.com/blogs/aws/twenty-years-of-amazon-s3-and-building-whats-next/
2•mooreds•8m ago•0 comments

Reynard: A simple Gecko-based web browser for iOS 14+

https://github.com/minh-ton/reynard-browser
1•tech234a•9m ago•0 comments

"Feel" as a Determinant of College Choice: Evidence from Campus Tour Weather

https://www.nber.org/papers/w34944
1•paulpauper•9m ago•0 comments

Claude Is Having an Outage

9•theahura•9m ago•1 comments

We Wasted 4 Weeks on a $1k/Month AI Agent

https://blog.kilo.ai/p/we-wasted-4-weeks-on-a-1000month
2•mmcclure•12m ago•0 comments

Welcome to the Block Universe

https://nautil.us/welcome-to-the-block-universe-1278973
1•samizdis•12m ago•0 comments

Mac Neo and my afternoon of reflection and melancholy

https://twitter.com/stevesi/status/2031842797838614548
1•CharlesW•12m ago•0 comments

Taste for Makers [February 2002]

https://paulgraham.com/taste.html
1•dnw•13m ago•0 comments

Knowledge Beacons: Semantic Compression (early prototype)

https://github.com/baevmikhail/mishx50-cyber
1•mishx50•13m ago•0 comments

Vibe-Code vs. Buy for B2B SaaS in 2026

https://forstarters.substack.com/p/for-starters-72-vibe-code-vs-buy
1•garrickvanburen•13m ago•0 comments

Solid-state EV batteries that can achieve 800 miles of range becoming a reality

https://electrek.co/2026/03/17/solid-state-ev-batteries-with-800-miles-range-become-reality/
2•breve•15m ago•0 comments

Engram: Persistent memory system for AI coding agents

https://github.com/Gentleman-Programming/engram
1•nateb2022•15m ago•0 comments

Karpathy's Autoresearch Is Two Theses Running in a Loop

https://sameernanda.com/autoresearch-verification-search/
1•sameernanda•16m ago•0 comments

ProtonOS – Bare-metal operating system written in C# (Claude-assisted)

https://github.com/ProtonOS/ProtonOS
1•olvy0•17m ago•0 comments

React Alternatives for Front-End Development in 2026

https://acropolium.com/blog/react-alternatives/
1•stmw•17m ago•0 comments

`Mr AI' Peter Thiel Lectures about Antichrist in Rome

https://www.ft.com/content/fc1e7e9a-9d5d-4217-b9b2-38069eb1197b
1•1vuio0pswjnm7•18m ago•1 comments

Show HN: Kita (YC W26) – Automate credit review in emerging markets

https://www.usekita.com/
1•rheamalhotra1•18m ago•0 comments

How I built a minimal-knowledge sync for WorkLedger

https://bastiangruber.ca/posts/how-i-built-a-minimal-knowledge-sync-for-workledger/
1•PaulHoule•18m ago•0 comments