N
NOSTRABOTUSProductLive StatsPricingDocs
LIVE PAPER: 57 pairs • $0.040 avg profit • 1.19% ROI net
Prediction markets are inefficient. We print the difference in USD.
Nostrabotus monitors 500k+ markets across Kalshi & Polymarket in USD, finds hidden equivalences with AI embeddings, and executes arbitrage. Real ROI math: profit $ diff, cost $ 1-diff, ROI = profit/cost, fees Kalshi 7% winning leg, Poly 2%. Single brain in Ohio, orders routed via Dublin t3.small 84ms WireGuard — private key stays in US.
$12400
Volume arb'd paper
57
Pairs tracked real 57 approved
84ms
US→EU WG proxy (only on execution)
1-4%
Net ROI after fees USD
Example $0.62 vs $0.58: diff $0.04 cost $0.96 gross ROI 4.17% | fee Kalshi 7% of $0.38 win = $0.0266 worst, Poly 2% of $0.58 = $0.0116 | net worst $0.0114 ROI net worst 1.19% on $10 cost = 10.4 contracts → $0.114 paper profit. No bps, all USD.
US-OHIO brain 3.16.140.129
EU-DUBLIN router 54.170.20.163 • t3.small • CLOB 200 OK
KALSHI $0.62POLY $0.58
Profit USD$0.04 on $0.96 cost → 4.17% gross ROI
Net after fees (7% Kalshi, 2% Poly worst)$0.0114 → 1.19% net ROI
PAPER TRADE EXECUTED · 84ms WG US→EU + 15ms CLOB · $0.0114 P&L net • via dublin-router
US brain only • detection 4ms p50 • Poly via Dublin 84ms only on executionLIVE PAPER
PROBLEM
Markets fragmented. Same events different prices. $0.62 vs $0.58 = $0.04 edge but hidden cost $0.96 = 4.17% ROI gross, net 1.19% after fees. Closes in 120ms.
Kalshi: TRUMP YES $0.62 ↔ Poly YES $0.58 diff $0.04 cost $0.96 ROI 4.17% gross
Fee worst $0.0266 → net $0.0134 ROI 1.40% → passes MIN_NET_ROI 1%
Human can't compute fees+slippage+W G 84ms fast enough
SOLUTION USD ACCURATE
US brain only: Ohio does 20Hz detection in USD (diff, cost, ROI, fees worst/best). Signs Polymarket EIP-712 locally, routes via Dublin router t3.small stateless FastAPI — key never leaves US. 7% Kalshi, 2% Poly fees accurate.
Discover 426k→57
Track $ ROI 4ms
Profit $ net 84ms proxy
HOW IT WORKS • ROI USD • PROXY NOT DUAL
From 490k markets to USD-profit via Dublin router.
Discover $
490k markets Ohio DB, 57 approved EQUIV/INVERSE 0.85 conf, 1024-dim Qwen3 HNSW, DeepSeek Pro. Pairs cached USD.
Track ROI $
US brain only: Kalshi WS direct Ohio 18ms, Poly WS via WG 10.10.0.2 or direct, calc diff $ cost $ 1-diff ROI diff/cost gross, fee worst max(7%*(1-win),2%*(1-win)), net worst diff-fee_worst, ROI net worst = net/cost. Thresholds MIN_GROSS_ROI 2% MIN_NET_ROI 1% MIN_PROFIT_USD $0.01. Depth $20 min VWAP slippage.
Profit $ via Dublin router
Paper: kalshi_dry_run_fill 20ms + poly_dry_run_routed_via_dublin 87ms measured live WG 84ms. Live: sign EIP-712 US, POST signed order via http://10.10.0.2:8000/proxy/polymarket/order → CLOB 15ms. Key stays US, Dublin only egress, stateless, t3.small 2vcpu 2GiB sufficient.
See it in action — live ROI USD
Explore Live Pipeline Paper →Real-time ROI net % (last 60s) — 1.19% net worst $0.040 diff
$0.040 diff cost $0.96 ROI 1.19% net worst • EQUIV
GET /api/stats/public → 57 pairs, 1.19% ROI net, 84ms WG US→EU, $10 cost target
ARCHITECTURE • PROXY NOT DUAL • USD
Kalshi $ direct 18ms→US Ohio brain 10.10.0.1
p50 detection USD ROI calc 4ms • MIN_GROSS 2% MIN_NET 1%
WG 84ms only on Poly submitEU Dublin router t3.small 10.10.0.2
Router POST /proxy/polymarket/order → CLOB 15ms • key stays US • stateless
Poly CLOB $ direct or via WG→Fill ~100ms
Arby 426k → 57 approved → Polybot backend ROI USD accurate fees 7%/2% → paper fills via Dublin 87ms → reports/daily-YYYY-MM-DD.md • No bps
ROI USD Accurate
diff $ gross profit per $1 payout, cost $ 1-diff, gross ROI diff/cost, fees USD 7% Kalshi win leg + 2% Poly win leg, net worst/best, ROI net = net/cost. All integer bps removed, USD only.
Proxy Not Dual
US 3.16.140.129 brain does 20Hz detection $ ROI 4ms p50, no RTT. Poly orders signed US, routed via 10.10.0.2:8000 Dublin t3.small stateless router → CLOB 15ms. Private key stays US, 84ms only when edge present.
Dublin t3.small
Was t3.medium 4GiB 2GB disk 93% full build failed. Now t3.small 2GiB mem 1.1Gi free 20GB gp3, sufficient for router ~50MiB, Caddy TLS poly-eu 200 OK, WG handshake 7s, CLOB 200 OK.
Obsessive Logging USD
Every trade edge $ cost $ ROI % fee $ worst/best net $ ROI net %, trace_id, book snapshot depth 5, latency breakdown us_to_dublin 84ms, execution 100ms, exportable CSV/JSON.
Self-Improving USD
Daily AI 00:00 UTC cron: 24h attempts/orders/failures/latency ROI %, suggests thresholds MIN_GROSS_ROI MIN_NET_ROI per category, blacklist, roadmap auto POST Arby.
Battle-Tested Safety
Circuit breaker 10% 5m failure rate pause pair, naked leg auto-hedge 60s market, daily loss $500 paper, reconciliation loop 30s, dry_run=true enforced, paper fills $ via Dublin 87ms.
Start arbitraging in $ — paper mode live.
Live at polybot.nostrabotus.com paper USD ROI: 0.040 diff 1.19% net, 57 pairs, 84ms WG US→EU, Dublin router CLOB 200 OK. Join waitlist.
Your keys US only • Dublin router stateless no keys • Audit logs USD
Nostrabotus is not a broker-dealer. Trading involves risk. Past performance not indicative. All P&L USD paper, ROI net worst after 7%/2% fees, no bps. Educational only.