Portfolio Value
$1,000
+0.00%
Balance Breakdown
Starting
$1,000.00
Cash
$1,000.00
Deployed
$0.00
Total P&L
$0.00
Realized: $0.00
Win Rate
—
0 closed trades
AI Cost Today
$0.00
0 cycles run
Sizing Health
1.00×
—
Gate Recommendations
Checking replay outcomes...
Missed profit and avoided loss will appear here once rejected trades are replay-priced.
Collecting replay samples...
No replay pricing run yet
net $0.00
Portfolio Value Over Time
Cycle Activity
Calibration Brier (per category)
Brier Legend & Status
Waiting for calibration data…
Open Positions
| Market | Side | Size | Unr. P&L |
|---|---|---|---|
| No open positions | |||
Latest Signals
Trade History
| Market | Side | Type | Entry | Current | Peak | Size | Unr. P&L | Edge | Est.Prob | Volume | Liquidity | Hold | Ends | Opened | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| No open positions | |||||||||||||||
Signal Feed
Total Signals
0
Critical/High
0
Sources Active
0
Active Markets
| Question | YES Price | NO Price | Volume | Liquidity | Tags | End Date |
|---|---|---|---|---|---|---|
| Waiting for bot cycle... | ||||||
Cross-Platform Arbitrage
True Arb
0
Guaranteed profit opportunities
Directional
0
Large cross-platform spreads
Last Scan
—
How this works:
Each bot cycle fetches matching markets from Kalshi and Manifold and compares prices against Polymarket.
True Arb = YESpoly + NOother < $1.00 — guaranteed profit if you execute both legs manually.
Directional = large price gap suggesting Polymarket may be mispriced.
⚠ Live execution is manual — place orders on both platforms yourself.
US users: Kalshi is available; Polymarket is geoblocked.
Opportunities
| Market | Type | Platform | Poly YES | Ext YES | Spread | Combined Cost | Profit % | Action | Confidence | Match | Volume | Liquidity | Ends | Tags |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| No opportunities detected yet — start the bot to begin scanning | ||||||||||||||
Rejected Opportunity Log
Skipped candidates and counterfactual EV filtered by the gates
Rejected
0
Shadow EV
$0.00
Top Reason
—
Missed Good
0
Correctly Avoided
0
Replay P&L
$0.00
| Market | Reason | Stage | Strategy | Side | Shadow EV | Replay | Time |
|---|---|---|---|---|---|---|---|
| No rejected opportunities recorded yet | |||||||
Cross-Platform Signal Feed
Cross-platform signals appear here when divergences are detected during bot cycles.
Configuration
Trading Parameters
Maximum USD per single market position
Skip trades whose Kelly size computes below this (dust floor — not rounded up)
Minimum probability edge required to trade (e.g. 0.08 = 8%)
Fraction of full Kelly bet size (0.25 = conservative)
Bot halts trading if daily loss exceeds this
Max positions opened per calendar day (resets at midnight)
Minimum AI confidence score to allow a trade (e.g. 0.30 = 30%)
Skip markets with less total volume than this
Skip markets with less available liquidity than this (0 = disabled)
Limits AI cost per cycle — top-scored markets are chosen first
Max fraction of total exposure in one tag cluster (e.g. 0.40 = 40%)
Auto-close when price rises ≥ this fraction above entry (e.g. 0.15 = +15%). Requires position held ≥ 60 s.
Auto-close when position is down this % from entry (e.g. 0.20 = close at −20% loss)
Lock in gains: close if price drops this % below its highest point. Leave blank to disable. Only activates once a position is in profit. (e.g. 0.10 = close if price falls 10% from peak)
After a stop-loss or trailing-stop fires, block re-entry into the same market for this many minutes. Set to 0 to disable.
AI Model Pipeline
Select provider then model for each layer. Changes apply immediately via hot-swap.
For persistence add to .env: AI_LAYER1_MODEL=provider/model-id
FALLBACK CHAIN:
FALLBACK CHAIN:
FALLBACK CHAIN:
Only invoke Layer 3 for trades larger than this amount
PINNED MODELS — COST REFERENCE (per 1M input tokens)
Loading…
Try AI first; fall back to rule-based algorithm when all AI quotas are exhausted.
☆ Model Library — pin models to your shortlist
★ pinned models appear in the layer selectors and cost grid above. ☆ unpinned models are hidden.
Loading…
Signal Sources
Broad English news — free tier has 24-32h delay
Real-time financial news — requires Polygon.io API key (POLYGON_API_KEY in .env)
Arbitrage Auto-Execute
When enabled, the bot automatically paper-trades the Polymarket leg of detected arb opportunities each cycle. You still need to place the external platform leg manually.
Bot will open a paper position on the Polymarket side of qualifying arb opportunities
Trade gate. A detected arb only auto-executes if true-arb guaranteed profit (or directional spread) is ≥ this. Separate from the detection thresholds in the Arb Detection panel — a true arb between the detect bar and this bar is shown but never traded.
A cross-platform spread this wide is more likely a resolution-criteria mismatch than a real arb. Above it, the position is sized down (below) instead of bet bigger.
Size multiplier for arbs above the suspicious-spread threshold (0.30 = 30% of normal Kelly size).
Hard gate before auto-execution. The external market title and URL must match the Polymarket question at or above this score.
Arb Detection Thresholds
Controls what qualifies as an arbitrage signal. Changes apply immediately.
Minimum % gap vs Kalshi/Manifold to generate a cross-platform signal. Also gates what appears in the Arb tab.
Detection bar: min guaranteed profit % (after fees) to classify a pair as true arbitrage and show it. It still only trades if it also clears "Min Profit to Execute" in Auto-Execute.
Detection bar: min price gap to surface a directional opportunity. (Note: this also acts as the execute bar for directional, since it's stricter than the Auto-Execute %.)
How long to reuse a cached Kalshi/Manifold price before re-fetching. Lower = fresher data, more API calls. Default 300s (5 min).
How many active Polymarket markets to check for arb opportunities each scan cycle. Higher = more coverage, slower scan.
Skip markets expiring within this many hours — short-lived markets expire before the spread closes, causing breakeven exits. Set to 0 to disable.
Cycle Intervals
How often the dedicated arbitrage scanner runs — independent of the main cycle (default 60s)
🐋 Whale Signal
Whale signal thresholds and position sizing multipliers. Changes take effect on next bot start.
Detection Thresholds
Whale bets ≥ this amount skip the AI edge gate entirely (default $50,000)
Single order ≥ this in book = whale wall detected (default $2,000)
Wall ≥ this = CRITICAL severity (default $10,000)
Markets with liquidity below this are treated as thin (default $20,000)
Price change ≥ this in ~1h = fast move (default 8%)
Price change ≥ this in ~1h = critical move (default 20%)
Probability Nudges (Follow)
Probability boost when whale and AI agree on direction (default 0.03)
Smaller nudge in thin markets where whale impact is less reliable (default 0.01)
Probability Nudges (Fade)
Probability shift away from whale move direction for fade signals (default 0.04)
Larger fade nudge in thin markets (more likely overreaction) (default 0.06)
Position Size Multipliers
Scale up position when whale and AI agree (default 1.15 = +15%)
Scale down when whale and AI point opposite ways (default 0.50 = −50%)
Scale down for fade trades — less conviction, more protection (default 0.60 = −40%)
Anti-churn. A whale order wall must persist on the same side for this many consecutive evaluations before a bypass position opens. 1 = enter on first sighting; 2+ filters transient/spoofed walls.
Live Model Swap
Hot-swap the AI model for any layer without restarting the bot. Requires the provider's API key to be set in .env.
API Key Status
Configure keys in your .env file. Restart the server after changes.