Skip to main content
← Back to Docs
🤖Bot Builder

Bot Strategy Templates

EV Follower, ARB Hunter, Whale Copier, and more

Predite ships with 6 pre-built strategy templates, each tuned to a different market behavior. This doc explains what each one does, when it works best, and the key parameters to configure.

1. EV Follower (Signal)

What it does: Scans for markets where the AI edge exceeds your threshold and buys the favored direction.

Best for: Catching mispriced markets early. This is the bread-and-butter strategy and our recommended starter.

How it works:

  • •Every 15 minutes, the bot checks all active markets matching your category and platform filters.
  • •For each market, it computes the current AI edge.
  • •If edge ≥ min_edge AND confidence ≥ min_confidence AND liquidity is sufficient, the bot places a market order.
  • •Position size = min(max_position, Kelly_size × bankroll, available_liquidity).
  • •Stop-loss and take-profit trigger close orders.

Parameters:

  • •Min edge (pp): 5-20. Higher = fewer but better-quality trades.
  • •Min confidence: 50-95. Higher = more selective.
  • •Max position ($): the per-order default for automation is $25. This field only tightens it further — to go above the default, raise Bots+copy / trade in Settings → Trading first, otherwise the order is rejected at the executor.
  • •Stop-loss (%): 15-50. Tighter = exits losses sooner but cuts good trades.
  • •Take-profit (%): 30-100. How much profit before locking in.
  • •Daily trade limit: 1-50.

Expected performance: 55-65% win rate, 8-15% monthly P&L on well-tuned settings (historical backtest).

2. ARB Hunter (Arbitrage)

What it does: Executes real single-market arbitrage on Polymarket, and alerts you to cross-platform spreads against Kalshi.

Best for: Markets where YES + NO briefly sum to less than $1.00, and matched Polymarket/Kalshi pairs with a price gap.

How it works — execution (Polymarket, automatic):

  • •When YES + NO on the same Polymarket market sum below $1.00 minus your min spread, the bot buys BOTH legs with equal shares. Held to resolution, the pair pays exactly $1.00 per share — the gap is locked-in profit.
  • •Both legs go through the guarded executor with all the standard caps and switches.

How it works — cross-platform (Kalshi, alert only):

  • •The scanner matches markets representing the same event across Polymarket and Kalshi and persists opportunities with the spread and both legs.
  • •When a matched pair clears your min spread, the bot sends you an alert spelling out which side to buy where.
  • •The Kalshi leg is YOURS to execute manually: Predite does not place orders on Kalshi yet. The bot never opens only the Polymarket leg of a cross-platform pair — a one-legged "arb" is a directional bet, not arbitrage.

Parameters:

  • •Min spread (%): 1-10. The minimum locked-in gap. 2-3% is realistic; 5%+ is rare.
  • •Max position per arb ($): bounded by your Bots+copy / trade cap ($25 by default). The value here only matters below that cap.

Expected performance: Low volume, high reliability on the executed single-market arbs. Cross-platform alerts are opportunities, not fills — your execution speed on the Kalshi side decides how much of the spread you capture.

Risks: Execution timing on the two Polymarket legs — if one fills and the other does not, you hold a directional position until the second leg fills. On cross-platform alerts, the spread can close before you execute the manual leg. ## 3. Whale Copier (Copy)

What it does: Mirrors trades from wallets you specify in real-time.

Best for: Following established profitable traders without manually monitoring their activity.

How it works:

  • •You provide a list of wallet addresses to follow.
  • •The bot watches Polymarket on-chain activity for those wallets.
  • •When a followed wallet trades, the bot replicates the trade in your account, scaled by your copy %.
  • •Stop-losses fire based on your config (not the followed wallet's).

Parameters:

  • •Wallets: list of addresses. 1-10 wallets recommended.
  • •Copy %: 5-100. What fraction of their size you replicate.
  • •Max size per trade ($): hard cap regardless of their size.
  • •Min size per trade ($): skip their tiny exploratory positions.

Expected performance: Mirrors the followed wallet's P&L minus slippage. If they make 20%/month, you'll make ~15-18% after fees and delays.

4. Mean Reversion (Quant)

What it does: Fades extreme recent price moves, betting the overreaction reverts toward the recent mean.

Best for: Liquid markets that swing 10+ points on a single headline or viral post.

How it works:

  • •Predite snapshots every tracked market's price and volume every 15 minutes (market_history). The bot uses the last hour of snapshots.
  • •It computes the 1-hour mean price (excluding the newest snapshot) and the typical per-interval volume.
  • •A trade triggers only when ALL of these hold: the current price has moved at least deviation_threshold_pp from the 1-hour mean; the last interval's volume exceeds volume_spike_mult x the typical interval (a move without volume is book drift, not an overreaction); and the AI estimate sits closer to the mean than to the new price, with confidence at or above min_ai_confidence. If the AI moved with the price, the news may simply be true - the bot does not fade it.
  • •Direction is against the move: spike up means buy NO, crash down means buy YES. One fade per cycle.
  • •If the market has no snapshot history (or the snapshot cron is down), the bot does not trade at all.

Parameters:

  • •Deviation threshold (pp): 3-25, default 12. How far price must sit from the 1-hour mean.
  • •Volume spike multiplier: 1.5-10x, default 3. Last interval volume vs the typical interval.
  • •AI confidence threshold: default 60. Below it, you are just guessing.
  • •Max hold (hours): closes the position after N hours regardless of P&L (max_hold_hours). A fade is a fast-reversion thesis - if it has not reverted, the thesis is dead.

Expected performance: Moderate win rate with above-average winners. Quiet markets produce zero trades by design - no catalyst, no fade.

5. News Reactor (News)

What it does: Uses fresh headlines from the integrated news feed as the trigger, and the AI's edge as the direction.

Best for: Event-driven markets where a catalyst moves probabilities before the market fully adjusts.

How it works:

  • •The bot reads real headlines (Finnhub and other sources) with a sentiment label and topic keywords.
  • •A headline counts as a catalyst only if it is fresh (within max_news_age_min), non-neutral, and - when you set news_categories - inside your chosen categories.
  • •Markets are matched to headlines by keyword overlap between the headline and the market question.
  • •Direction comes from the AI signal, not from the headline's sentiment: positive news does not map to YES for an arbitrary question (positive for whom?). The headline decides WHEN, the AI decides WHICH WAY - and only with at least 12pp of edge.
  • •No fresh matching headline means no trade. That is the difference from EV Follower: this bot stays flat in quiet news periods.

Parameters:

  • •Max headline age (minutes): default 60. Bots run on a 5-minute cycle, so the reaction window is minutes, not seconds.
  • •News categories: which feed categories count (economics, crypto, tech...). Empty = all.
  • •Max trade size ($): standard per-bot sizing applies.

Expected performance: Bursty. Active during news-heavy periods, intentionally silent during quiet ones.

6. Calendar Bot (Macro)

What it does: Enters positions ahead of scheduled events from the integrated economic calendar.

Best for: Predictable catalysts - Fed decisions, CPI, jobs reports, earnings.

How it works:

  • •The bot reads the real economic calendar (Finnhub) - event time, impact level and topic.
  • •An event is a trigger only when it starts within lead_time_hours (2-48h ahead), has impact at or above min_event_impact, and - when set - matches your event_categories.
  • •Markets are matched to events by keyword overlap between the event and the market question.
  • •Direction and conviction come from the AI: the bot only enters when the AI edge on the matched market clears your minimum. It does not predict the event's outcome from historical patterns - no such dataset exists, and the bot does not pretend otherwise.
  • •Exits use the standard time triggers: exit_minutes_before_end to be flat before resolution, or max_hold_hours to cap the hold.

Parameters:

  • •Lead time (hours): 2-48, default 24. How far before the event to enter.
  • •Min event impact: low / medium / high, default medium.
  • •Event categories: economics, politics, crypto, tech, earnings, ipo. Empty = all.
  • •Min edge (pp): default 8. The AI edge required on the matched market.

Expected performance: Steady but modest, concentrated around the macro calendar. No scheduled events in window means no trades.

Combining Strategies

The Bot plan allows up to 10 bots (a paused bot still holds its slot). A reasonable diversified portfolio:

  • •1 × EV Follower (broad signal capture)
  • •1 × ARB Hunter (low-correlation income)
  • •1 × Whale Copier following 1-2 trusted wallets
  • •1 × Mean Reversion in highly liquid markets
  • •1 × Calendar Bot for scheduled events

This gives you exposure to multiple alpha sources. Avoid running 3 EV Followers with slightly different parameters — they'll fight for the same trades.

Related Docs

← All Documentation