Skip to main content
← ← Back to Guides
Advanced📖 20 min

Building Your First Bot

Predite's Bot Builder lets you automate any strategy that can be expressed as a rule. No code. You pick a template, configure parameters, and the bot runs every 15 minutes scanning for opportunities. This guide walks you through building the most common starter bot — an EV Follower — from scratch.

Why Automate?

The biggest edge in prediction markets is consistency. Markets move fast. Edges appear and disappear within minutes when news drops. If you're not at your screen 24/7 (and you shouldn't be), you'll miss opportunities. A bot doesn't miss. It scans every 15 minutes, evaluates every active market against your criteria, and executes whenever the conditions match.

The second edge is discipline. Humans abandon strategies after losing streaks. Bots don't. As long as your strategy is +EV and your sizing is sane, the bot will keep doing the right thing through every drawdown.

Prerequisite: Bot Plan

Bots require the Bot plan ($99/mo). The plan gates:

  • •Up to 5 active bots simultaneously
  • •Live CLOB execution on Polymarket
  • •TWAP order splitting
  • •Full backtesting with Monte Carlo replay
  • •Bot Builder access

If you're on Starter or Pro, you can upgrade from Settings → Subscription. The Bot plan is the only tier with live execution; Starter and Pro can build paper bots for testing but can't go live.

Step 1: Choose a Strategy Template

We ship six templates, each tuned to a different market behavior:

  1. EV Follower — buys when AI edge exceeds your threshold. Best for: catching mispriced markets. Recommended starter.
  2. ARB Hunter — executes both sides of a cross-platform spread. Best for: risk-free profit from price discrepancies.
  3. Whale Copier — mirrors trades from wallets you follow. Best for: piggy-backing smart money.
  4. Mean Reversion — fades extreme price movements. Best for: markets that overreact to news.
  5. News Reactor — trades on sentiment shifts from the news feed. Best for: fast-moving events.
  6. Calendar Bot — positions before scheduled events (Fed meetings, earnings, elections). Best for: predictable catalysts.

For your first bot, start with EV Follower. It's the simplest, has the most predictable performance, and is the easiest to reason about. Once you've run it for a month and understand its behavior, branch out.

Step 2: Configure Parameters

Go to Dashboard → Bot Builder → New Bot → EV Follower. You'll see a configuration form:

  • •Bot name. Anything descriptive: "EV Follower 10pp Conservative". Future you will thank you.
  • •Mode. Paper or Live. Always start in Paper. Run for at least 100 trades before considering Live.
  • •Min edge (pp). The minimum AI edge to trigger a trade. Default 5pp. Recommended starter: 10pp. Higher = fewer but better-quality trades. Lower = more frequent, more variance.
  • •Max position size ($). USD cap per trade. Start at $25-50 for your first bot. You can scale up later.
  • •Daily trade limit. Maximum number of trades per day. Default 5. This is a circuit breaker against runaway bots.
  • •Stop-loss (%). Close the position if it moves against you by this much. Default 25%. Tighter (15-20%) catches losses sooner but exits good trades; looser (35-50%) rides through volatility.
  • •Take-profit (%). Close the position if it moves in your favor by this much. Default 50%. Conservative: 30-40%. Aggressive: 60-100%.
  • •Platforms. Polymarket only (default), Kalshi only, or both. Bot evaluates each platform independently.
  • •Categories. Filter to specific market types (politics, economics, sports, etc.). Default: all. Use this to specialize.

A good first config:

  • •Min edge: 10pp
  • •Max position: $50
  • •Daily limit: 5
  • •Stop-loss: 25%
  • •Take-profit: 50%
  • •Platforms: Polymarket
  • •Categories: all

This is conservative enough that one bad day won't hurt much, but liberal enough to actually trade.

Step 3: Backtest Before Going Live

Predite's backtesting engine replays your strategy against historical market data. It tells you:

  • •How many trades the bot would have taken in the last 90 days
  • •Total P&L
  • •Win rate
  • •Max drawdown
  • •Sharpe ratio (risk-adjusted return)
  • •Distribution of outcomes via Monte Carlo simulation

To backtest:

  • •In Bot Builder, click "Backtest" before saving.
  • •Choose your lookback period (30, 90, or 180 days).
  • •Wait 10-30 seconds for results.
  • •Review the metrics.

What to look for:

  • •Win rate ≥ 50%. If your bot is winning under 50% of trades with a 10pp edge requirement, something's off. The edge requirement should produce 55-65% winners.
  • •Max drawdown under 30%. Drawdowns that exceed this are a sign of insufficient diversification or excessive sizing.
  • •At least 20-30 trades in 90 days. Fewer than that and your sample size is too small to trust.

If the backtest looks bad, adjust parameters and re-run. Don't go live on a bot that doesn't backtest well.

Step 4: Run in Paper Mode

After saving the bot, it's in Paper mode by default. The bot will execute simulated trades every 15 minutes, but no real money moves. Watch it for at least two weeks:

  • •Check the trades tab daily.
  • •Compare actual outcomes to backtest projections.
  • •Look for surprises — markets the bot traded that you wouldn't have, or vice versa.
  • •Note any trades where the resolution was ambiguous or the bot got "stuck" with a position it couldn't exit.

If after two weeks the paper performance matches backtest within reasonable variance, you're ready for Live.

Step 5: Go Live

Before flipping the switch, make sure:

  • •You've connected Polymarket via Settings → Trading API (CLOB credentials).
  • •Your Polymarket wallet has USDC equal to at least 2x your max bot position size.
  • •Your kill-switch is configured (Settings → Trading → Kill Switch).
  • •You have alerts wired up so you know when the bot trades.

Then in the bot detail page, change "Mode" from Paper to Live and save. The bot will execute its next eligible trade at the next 15-minute cron tick.

Start small in Live. Cut your max position size by 50% for the first week. Watch every trade. Make sure executions match expectations. Then ramp back up.

What to Monitor After Launch

Daily for the first week, then weekly:

  • •Win rate trend. If it's drifting below your backtest by more than 5pp, investigate.
  • •Average slippage. Live executions are rarely at the exact price the bot expected. If slippage is over 2pp on average, your edge threshold needs to be higher.
  • •Position diversification. Is the bot concentrating in one category? Adjust filters or add caps.
  • •Resolution surprises. Did any market resolve in a way you didn't expect (rule lawyering, edge cases)? Document these and refine your filters.
  • •Drawdown recovery. If the bot is down 15-20% from peak, pause and review. Don't just hope it'll come back.

Common First-Bot Mistakes

  • •Going live with too-aggressive parameters. Min edge of 3pp and max position of $500 will blow up. Be conservative for the first 100 trades.
  • •Skipping backtesting. Backtest reveals failure modes you can't see in theory.
  • •Running multiple bots before understanding one. Master one strategy before adding more. Multiple bots interact in non-obvious ways.
  • •Disabling stop-loss to "let it ride". This is the fastest way to turn a bad trade into a catastrophic one.
  • •Not checking the bot for weeks. Markets change. Strategies that worked in Q1 may not work in Q3. Review monthly minimum.

Next Steps After Your First Bot Is Stable

  • •Read the Bot Strategy Templates doc to learn the other 5 templates.
  • •Try a second template (ARB Hunter is a good second bot — orthogonal to EV Follower).
  • •Read the Bot Execution Engine doc to understand the cron + audit log mechanics.
  • •Join our community to discuss strategies with other Bot-plan users.
← Previous
Whale Tracking & Copy Trading
Next →
Cross-Platform Arbitrage
← ← All guides
Building Your First Bot | Predite