From Backtest to Paper Trading: Forward-Test Polymarket & Kalshi Strategies Live

Why a backtest alone isn't enough

A backtest replays markets that have already settled. Done honestly — against the real book, with depth-aware fills — it is the best cheap filter there is for a bad idea. But it shares one unavoidable weakness with every backtest: it is fit, however lightly, to the past. The only way to know whether an edge is real is to run it forward on markets whose outcome nobody knows yet.

That is what paper trading is for. It takes the rule that survived your backtest and runs it in real time on live markets, marking and settling it exactly as the live book and the real resolution dictate. Nothing about the outcome is known in advance, so the track record it builds is forward evidence a backtest cannot provide.

Two ways to feed it: a deployed rule, or your own bot

The first path is the one that chains off the Backtest Lab. When a rule survives a backtest, deploy it to paper trading with a stake, take-profit/stop-loss, and a max-open cap. It then runs server-side against live quotes — DepthFeed evaluates the rule, opens and closes positions, and marks the book for you, with no infrastructure on your end.

The second path is for people who already have a system. Each strategy gets a secret webhook URL; your bot, cron job, or notebook POSTs buy/sell/close signals with a market id and a dollar size, and DepthFeed fills them at the live book at signal time. A payload can never set its own price, so the track record is honest by construction. Idempotency keys make retries safe, the token rotates on demand, and the rate limit is 120 signals per minute per strategy.

How fills and settlement work

Every fill is priced from the live order book at the instant the signal arrives: a buy pays the best ask, a sell hits the best bid — the same prices the data API and the WebSocket serve. Positions are marked from the live book on roughly a 90-second cadence, so the equity curve moves as the market does, and each position settles to $0 or $1 automatically when the market resolves.

One honest caveat, stated up front: paper fills use real displayed liquidity for pricing but don't consume it. A large live order could see more slippage than the top-of-book paper fill, so paper results are a fair test of a signal's timing and direction, and a slight upper bound on a size-sensitive strategy's real execution.

A real track record, not a spreadsheet

Each strategy holds virtual cash — anywhere from $100 to $1,000,000, defaulting to $10,000 — and reports a live equity curve, realized versus unrealized P&L, open positions, and a per-trade log. It is the same reporting a real book would give you, on money you are not risking, so you can watch how a strategy behaves through markets it has never seen.

How many strategies you can run at once scales with your plan: one on the free tier, five on Quant, fifteen on Desk Lite, and forty on Desk — enough to paper-trade a whole portfolio of ideas in parallel and let the equity curves decide which ones graduate.

The full loop

Put the pieces together and the workflow is a loop: form an idea, kill it or keep it cheaply in the Backtest Lab, deploy the survivor to paper trading, and watch it earn a forward track record on live books. If it holds up, you have a strategy tested both on history and on markets that hadn't happened yet — and the same DepthFeed data and JSON shape run underneath all of it, so nothing is rewritten between stages.