bangertrades 0.1.4

Banger: write, test, and run prediction-market strategies that hit.

banger (Python SDK)

Strategies that hit.

The Python SDK for Banger. Write a strategy as a Python class, ship it with banger deploy strategy.py --paper, watch it run.

Install

uv pip install bangertrades

The comprehensive distribution includes the SDK, CLI, runtime, and paper-mode venue adapters.

For local development on this monorepo:

uv pip install -e packages/sdk-python
uv pip install -e packages/adapters/polymarket
uv pip install -e packages/adapters/kalshi

The 30-second tour

from decimal import Decimal
from banger import Strategy, signals
from banger.venues.kalshi import Kalshi
from banger.risk import RiskEnvelope


class MyStrategy(Strategy):
    name = "my_strategy"
    venues = [Kalshi()]
    universe = signals.kalshi_markets(category="NFL", time_to_resolution_lt="24h")
    risk = RiskEnvelope(
        max_position_usd=Decimal("200"),
        max_daily_loss_usd=Decimal("500"),
        max_open_positions=5,
    )

    async def on_market_tick(self, market, tick, ctx):
        if tick.yes_price and tick.yes_price < Decimal("0.45") and ctx.can_open_position():
            await ctx.place(Kalshi(), market, side="yes", size=ctx.kelly(edge=0.05))

banger deploy my_strategy.py --paper

That's it. The runtime will:

1. Resolve universe against Kalshi to get the live market list.
2. Subscribe to ticks for all matching markets.
3. Call on_market_tick for each update.
4. Risk-gate every simulated paper order before recording its fill.
5. Show you P&L, drawdown, and per-position attribution in real time.

Core concepts

• Strategy — base class. Subclass and declare name, venues, universe, risk.
• Venue — abstract execution venue. Provided by adapter packages.
• Universe — declarative spec of which markets to subscribe to (signals.kalshi_markets(...)).
• RiskEnvelope — declarative risk limits enforced before orders ship.
• Context — passed to event handlers; provides place(), reduce-only close(), cancel(), kelly(), log(), positions, and can_open_position(). place(size=...) accepts USD notional while close(contracts=...) accepts held contract quantity; contracts=None closes the full position.

Examples

See examples/:

• kalshi_nfl_arb.py — Kalshi vs DraftKings spread arb (illustrative)
• polymarket_simple_fade.py — Fade overconfident Polymarket markets

Status

Public paper beta. Historical venue datasets and live execution are not shipped. The interface is still settling; expect reviewed breaking changes.