X is full of Polymarket bot success stories:
- “I gave a bot $100 and it made $347 overnight.”
- “$900 per day exact formula.”
- “$75k in a month from market making.”
- “$500 to $800 daily from arbitrage alerts.”
We reviewed hundreds of Polymarket bot repositories (~860 on GitHub for “polymarket bot,” ~3,500 Polymarket-related repos total), official SDKs, community tools and academic research.
Most of this is fake, misleading or affiliate marketing. Most people who run automated bots on prediction markets lose money.
The two biggest scams in Polymarket bot content
1. Viral profit claims
Most viral bot posts are designed for engagement and referral fees, not profitability. If you see links with ?ref= or ?via= parameters, that’s usually the real business model.
70% of all Polymarket traders have realized losses. Only 0.04% of addresses captured over 70% of total profits.
Survivorship bias makes this worse. The bots that survive have 85%+ win rates, but most attempts at running automated trading bots result in losses. The winners have custom infrastructure, significant capital and years of experience.
Profitable traders usually have:
- Years of market intuition
- Significant starting capital
- Custom infrastructure and low-latency execution
- Ruthless risk management
- Comfort with being wrong often
There is no bot that reliably turns $100 into $347 overnight. Claims like these trace back to single unverified anecdotes, often involving general-purpose AI tools that aren’t even trading bots. The CFTC has explicitly warned that claims of huge returns from AI trading bots are a hallmark of fraud.
2. Keyword-stuffed GitHub repos
After reviewing hundreds of Polymarket repositories, the pattern is obvious. Many “popular” bots have descriptions like:
polymarket bot, polymarket trading bot, polymarket copy trading bot, polymarket arbitrage bot, polymarket trading bot…
This is SEO spam, not documentation.
Red flags:
- Extremely high star counts
- Almost no real explanation of strategy
- Repeated keywords instead of real descriptions
- Forks of forks with no meaningful changes
Before running any third-party bot:
- Read every line of code
- Check for hard-coded wallet addresses
- Verify how API keys are used
- Never give write access unless you fully understand it
If you can’t explain what the code does, don’t let it touch your money.
How Polymarket actually works (and why this matters)
Polymarket uses a central limit order book (CLOB). That means:
- Limit orders, not just market orders
- Visible bid/ask spreads
- Order book depth is public
- Markets resolve to $1.00 (YES) or $0.00 (NO)
This matters because bots don’t “predict outcomes.” They:
- Read order books
- Manage inventory
- React faster than humans
- Execute predefined rules
Speed, execution quality and discipline matter more than “intelligence.”
5 bot strategies, from simple to dangerous
1. Copy trading
What it is: Automatically mirrors trades from “smart money” wallets.
Why people like it: It feels easy. Someone else does the thinking.
Why it fails:
- You enter later at worse prices
- You don’t know the trader’s risk tolerance
- Past performance doesn’t persist
- Position sizing rarely matches your bankroll
Reality: Start with alerts. Don’t auto-execute.
2. Arbitrage
What it is: Buying mispriced YES/NO outcomes for guaranteed profit.
Example: YES @ $0.48 + NO @ $0.49 = $0.97 cost → $1.00 payout.
Why people like it: It seems “risk-free.”
Why it fails:
- Opportunities last milliseconds
- Fees and slippage erase gains
- Liquidity is thin
- Professionals get there first
Claims of “$900/day arbitrage” for retail traders are not credible.
3. Market making
What it is: Providing liquidity on both sides and earning the spread.
Why people like it: It looks steady and systematic.
Why it fails:
- Informed traders trade against you
- Inventory risk accumulates quickly
- Requires constant rebalancing
- Needs meaningful capital to matter
This is hard even for professionals.
4. Sentiment-driven trading
What it is: Using X (Twitter) sentiment to anticipate price movement.
How it works:
- Scan posts by keyword
- Score sentiment
- Compare against implied odds
- Trade large divergences
Why it fails:
- Social media is noisy
- Markets often price news faster than sentiment tools
- Small sample sizes mislead
Useful for alerts, not blind execution.
5. AI-assisted trading
What it actually is: A self-hosted AI assistant, not a money printer.
What it can realistically do:
- Cross-market research
- Sentiment divergence alerts
- Liquidity monitoring
- Trade journaling and post-mortems
- Wallet tracking
- Performance dashboards
What it cannot do:
- Guarantee profits
- Replace judgment
- Turn $100 into $347 overnight
- Beat the market autonomously
AI helps you think better. It doesn’t think for you.
The only path that doesn’t end in a blow-up
Step 1: Learn the API (no trading yet)
Use the official SDK. Just read data.
pip install py-clob-client
- Fetch market data, read order books, understand the data structures
- Query historical trades and resolutions
- Spend at least a week just reading data before writing any trading logic
Step 2: Build analysis before execution
Before placing a single trade:
- Track your own manual trades with entry thesis and outcome
- Monitor specific markets you’re interested in
- Set up alerts for price movements or unusual activity
- Calculate your own calibration (how often your predictions are right)
- Review this data weekly
If you can’t explain why you’re winning or losing, automation will amplify mistakes.
Step 3: Paper trade (minimum 4 weeks)
- Log what you would have bought/sold
- Track simulated P&L with realistic assumptions about slippage and fees
- Run for a few days/weeks before using real money
- Identify whether your strategy has an actual edge (>55% hit rate after costs)
Step 4: Start small and conservative
- Use an amount you’re genuinely comfortable losing entirely
- Set strict position sizing rules (e.g. never more than 2-5% of bankroll on one market)
- Implement stop-losses or maximum loss limits in code
- Log every trade with full reasoning
- Review losses
Step 5: Add automation gradually
Don’t go from 0 to fully autonomous overnight.
- Manual trading with bot-generated alerts and analysis
- Bot suggests trades, you approve each one manually
- Bot executes within tight parameters (small size, high-confidence only)
- Expand parameters slowly as you build confidence and data
Risk rules every bot must have
- Max position size
- Daily loss limit
- Liquidity thresholds
- No leverage
- Market time-to-resolution limits
- Full audit logs
- One-command kill switch
- Weekly reviews (especially losses)
If your bot doesn’t enforce discipline, it will enforce losses.
Official Polymarket repos
If you’re going to build, start from the source. These are maintained by Polymarket Engineering under github.com/Polymarket.
CLOB clients:
| Language | Repository |
|---|---|
| Python | py-clob-client |
| TypeScript | clob-client |
| Rust | rs-clob-client |
Order utilities:
| Language | Repository |
|---|---|
| TypeScript | clob-order-utils |
| Go | go-order-utils |
Smart contracts:
- ctf-exchange: core on-chain exchange contract
- neg-risk-ctf-adapter: negative risk adapter
- uma-ctf-adapter: market resolution via UMA’s Optimistic Oracle
Real-time data:
- real-time-data-client: WebSocket feeds for live orderbook and trade updates
Examples:
- examples: runnable code for placing orders, fetching markets, etc.
No shortcuts, just better tools
Polymarket bots are tools, not cheat codes.
The people making money from “bot strategies” on X are often selling referrals, farming engagement, or both.
If you treat this as a learning project, you might build something useful. If you treat it as a get-rich-quick scheme, you will lose money.
There are no shortcuts. Just better tools, slower decisions and honest accounting of what works.