Trading Tools

Paper Trading

8 min read

What it is

Paper Trading runs YOUR saved Strategy Builder strategies against LIVE exchange candles with virtual capital. No real orders fire — but the worker reads real Coinbase / Binance US candles every 5 minutes and evaluates your strategy rules tick-by-tick. Trades execute at the candle's close price with 10 bps slippage + 40 bps fee, matching realistic retail fills. Sessions persist server-side, so you can close the browser and they keep running. Cap: 3 concurrent sessions.

How to think about it

Backtesting tells you a strategy worked on past data. Paper trading tells you it works on TODAY'S data without the lookahead bias backtests can sneak in. Run it for at least 7 days — ideally 30 — before assigning the same strategy to a real bot. If a strategy can't print on a paper session, it won't print live either.

Step-by-step

  1. Prerequisite: save a strategy + connect an exchange

    Paper Trading runs YOUR saved strategies, not generic presets. Build one in Strategy Builder first and save it. Also connect at least one exchange in Settings → Connections — we need real candle data, but no API write scope is required (read-only is enough since no orders fire).

  2. Click "New session"

    Top-right amber button. Opens the session creation modal. Disabled if you have 3 running sessions already, or no saved strategies, or no connected exchanges — the warning banner tells you which one to fix.

  3. Pick strategy + exchange + pair + capital

    Strategy dropdown lists all your saved Strategy Builder strategies. Exchange dropdown lists only your connected ones. Pair list is exchange-specific and searchable (type "ETH" to filter). Initial capital defaults to $10,000 — use whatever amount you'd actually trade with to get meaningful % returns. Timeframe is fixed at 1h.

  4. Create — session goes live immediately

    Click Create. The new session appears in the sessions grid with RUNNING status. The worker evaluates it every 5 minutes against the latest candle from your chosen exchange. First fill may take up to 5-10 minutes depending on when the next worker tick lands.

  5. Watch the 4 KPI cards on each session

    Equity (current virtual portfolio value) vs Initial. Return % (signed). Trades (total count + win rate). Last price + 📈 Open indicator if a position is currently open. Last tick timestamp shows when the worker last evaluated the session.

  6. Pause, Resume, or Stop

    Pause = freeze the session, no new trades, equity locked. Resume = unfreeze, worker picks back up on next tick. Stop = end the session permanently (status STOPPED, no resume possible — but trade history stays). Delete = remove the session entirely (history lost).

  7. Read the result the right way

    GOOD: 30+ trades over 7+ days, positive return, win rate doesn't have to be >50% if average win > average loss. BAD: under 10 trades (not enough signal — wrong pair or wrong strategy logic), large drawdown vs market in the same period, all losses cluster in the same window. Compare the same strategy on 2-3 different pairs to find where it works.

  8. Promote to live when ready

    After 7-30 days of acceptable paper results, the SAME strategyId is what you assign to a live bot. No translation needed — paper and live use the same engine, just with real vs simulated fills. Start the live bot with the smallest viable position size for 30 more days before scaling up.

Tips & pitfalls

  • Server-side persistence: sessions keep running even if you close the browser, log out, or your computer sleeps. The worker ticks every 5 minutes regardless of who is watching. Come back next week, the data is there.
  • Cap of 3 concurrent sessions is intentional — testing 3 different strategies on the same pair, or the same strategy on 3 different pairs, both give meaningful comparisons. Beyond 3 you start chasing noise.
  • Fee model is realistic: 40 bps fee + 10 bps slippage = effective 50 bps round-trip cost. A strategy that needs 0 fee to be profitable is not actually profitable on a real exchange — paper trading exposes this honestly.
  • Timeframe is locked at 1h — the worker runs every 5 minutes, so anything shorter would have evaluation lag. 1h is the sweet spot for retail swing/trend strategies. For scalping (5m/15m), the Strategy Builder backtest is the right validation tool.
  • Open Position indicator (📈) on a card means the strategy currently holds the asset. If you Stop the session while a position is open, the worker closes the position at the next tick's price — your final equity reflects that close.
  • 7 days = bare minimum. 14-30 days = meaningful. Strategies that print for 3 days then collapse are common — short windows lie. Run longer than feels necessary.
  • Same strategy on 3 different pairs (e.g., BTC/USDT, ETH/USDT, SOL/USDT) is one of the highest-value tests. Strategies that work on ALL three are robust. Strategies that only work on BTC may just be over-fit to BTC's historical regime.
  • No bridge to your real exchange balance. The $10,000 you pick is virtual — your real Coinbase/Binance US account is untouched. The exchange connection is read-only candle data.