Polythemes · Trading automation

Your tested strategy, on autopilot — with a human hand on the switch

The ruleset that survived the backtest goes live unchanged. The bot scans and proposes around the clock; you decide how much it may execute on its own.

strategy: boring-v3
SEMI-AUTOFULL
Scan
Signals
Propose
Human sign-off
Execute
orderbook depth$41k bid @ 0.96
resolution window18 days
emotion score2 · single-actor
news scan (3D)no contradicting headlines
proposed trade
NO · “Rate hike before October?” $250 @ 0.96
FILLED · 260.4 NO @ $0.9598tx 0x7c…91 · Polygon

01 · Backtest-integrated

What you simulated is what trades

Automation isn't a separate product to configure from scratch. A validated backtest run exports as the live strategy — same filters, same exits, same sizing.

  • 01Same ruleset, same code paththe parameters you swept and validated export directly as the live trading configuration
  • 02No slippage between sim and prodentry filters, exit ladders, and sizing behave in production exactly as they scored in the backtest
  • 03Live performance vs. expectationrealized fills are tracked against the simulated curve, so drift shows up immediately
Start in the backtester →
semi-auto
full auto
Scan & score markets
bot
bot
Build trade proposal
bot
bot
Approve execution
you
bot (within limits)
Place & monitor orders
bot
bot
Ladder exits / stops
bot
bot
Raise limits, new markets
you
you

Move between modes any time. Most funds run semi-automated for the first weeks, then graduate the routine trades to full auto and keep sign-off on the outliers.

02 · Guardrails

Autonomy you can grant — and revoke — by degrees

Every execution path assumes the bot could be wrong or compromised. Approvals, key segregation, and hard limits keep the blast radius near zero.

  • 01Semi-automatedthe bot scans, scores, and proposes; nothing executes until you approve — ideal while trust builds
  • 02Fully automatedhands-off execution inside hard limits: position size, max exposure, category caps, drawdown stops
  • 03Kill switchone action pauses all execution instantly — exits keep running, entries stop
  • 04Human-factor authenticationbot executions are proposals until a human signs off — approvals are explicit, auditable actions
  • 05Segregated keysthe operator key can trade but can never withdraw — custody stays with the fund owner
  • 06Hard risk limits on-chain pathsspending caps and exchange approvals are scoped so even a compromised bot can't drain funds
approval request · push + signature
Awaiting your sign-offexpires in 29:41
Buy 260 NO · “Rate hike before October?” · $250 @ 0.96
within limits: size ✓ · exposure ✓ · category ✓
Approve & signReject
  • Operator key trades, never withdraws — funds sit behind contract-scoped approvals
  • Every approval is logged — who signed, when, against which limits

03 · Real-time inflows

The 3D-strategy layer runs live

The same real-time intelligence you backtested — news flow, feeds, live orderbooks — streams into production decisions. Before the bot proposes a trade, the analyst layer reads what the world looked like a minute ago, not last month.

Will the Fed cut rates in September?
NO @ $0.91 · 14d to resolve · passed all 30+ filters
news as it stood at entry
Fed officials signal patience as inflation cools further
Reuters · 2d before entry
Markets price just 8% odds of a September cut
Bloomberg · 1d before entry
AI · TAKE

News flow agrees with the market's calm — no catalyst found that the price hasn't already absorbed.

Will Iran strike back within a month?
NO @ $0.86 · 22d to resolve · passed all 30+ filters
news as it stood at entry
Iran vows 'decisive response' after overnight strike
AP · 3h before entry
Regional powers brace for weeks of escalation
FT · 1d before entry
AI · SKIP

Live escalation risk — the headlines contradict the calm this price implies. Not worth the yield.

See a strategy go from sweep to live fill

We'll walk you through the whole loop — backtest, export, approve, execute — on a live account.