Paper Trading vs Live Trading, and When to Switch

Paper trading runs your strategy against live market prices with simulated money: real signals, real timing, pretend fills. Live trading is the same pipeline with real orders. The difference sounds small and is not, and knowing exactly what each stage proves is what keeps the switch from being a leap of faith.

The short answer to "when do I switch?" is: after a few weeks of paper trading whose results broadly match what your backtest predicted for the same period. Everything else in this lesson unpacks that sentence.

What paper trading actually tests

A backtest replays history. Paper trading meets the present, and the present contains an entire class of problems history cannot show you:

• Operational failures. Data feeds stall, connections drop, orders would have been rejected. None of this exists in a backtest; all of it exists in production.
• Signals on live data. Backtests read clean, finalized bars. Live data arrives tick by tick, occasionally glitches, and sometimes revises. A rule that behaves on polished history can misfire on the raw feed.
• Current market conditions. History contained its own regimes. The present may be one history never produced, and paper trading is a free look at how the strategy handles it.
• You. Paper trading previews the strategy's rhythm: how often it trades, how long drawdowns last, how it feels to watch. If the pace is intolerable simulated, it will be worse funded.

What paper trading hides

Paper trading is a rehearsal, not the performance, and its fills are polite fictions. Three gaps matter:

Fills are optimistic. A simulated order fills at the quoted price. A real order competes for liquidity: it can fill partially, fill worse, or move the price against you. For strategies trading liquid instruments at modest size the gap is small; for anything fast or thin it is decisive.

Slippage and fees round down to zero. Simulated results often skip the spread, the commissions, and the funding costs that real accounts pay on every round trip. A strategy whose edge is thinner than its costs looks fine on paper in both senses.

Nothing is at stake. The psychological load of watching real money draw down cannot be simulated. The discipline that felt easy for six simulated weeks gets its first real test the day a live position goes red.

The switch test

The decision to go live should be a comparison, not a feeling. Over the paper period, line the results up against what the backtest expected for those same weeks:

1. Trade count. Did the strategy take roughly the number of trades the backtest implied for that stretch of market?
2. Win rate and average trade. Within the backtest's normal variation, not matching it exactly.
3. Fills. Are simulated entry and exit prices close to the prices the backtest assumed, or is every fill a little worse in the same direction?
4. Drawdown. Inside the tested envelope.

Broad agreement means the pipeline works and the strategy is behaving like its tested self: go live small. Sharp disagreement means something specific is wrong, and every one of the checks above points at where. Diverging trade counts implicate the data or the rules; diverging fills implicate liquidity assumptions; a drawdown outside the envelope implicates the test itself.

Going live without betting the account

The switch is not one moment but a ramp:

1. Start at a fraction of intended size, commonly a quarter or less. The first live weeks are still a test, now with the last simulated assumptions removed.
2. Compare live fills to paper fills. This is the gap paper trading could not measure, and it is now measurable. Persistent shortfall beyond modeled slippage means resizing or rethinking.
3. Scale up on evidence. After a sustained stretch of live results tracking expectations, step the size up. Two good days are not a stretch; think weeks.
4. Keep the risk rules from day one. The stop-loss distances, the 1 to 2 percent per-trade risk, and the drawdown circuit breaker from Risk Management for Algorithmic Trading apply at quarter size exactly as they will at full size.

On Horizon the same deployment pipeline serves both stages: a strategy that passed its backtest can run against live data before real capital is attached, then deploy to a connected broker with its risk guardrails in place.

Common mistakes at this stage

• Skipping paper trading entirely because the backtest was beautiful. The backtest tested the idea; nothing yet has tested the pipeline.
• Paper trading forever. After the comparison test passes, more simulated weeks add little information. The remaining unknowns are only measurable live and small.
• Judging the strategy on the paper period's profit instead of on agreement with expectations. A lucky simulated month proves less than an ordinary month that matched the model.
• Going live at full size the day the paper test passes, which converts the one remaining unknown into a maximum-stakes experiment.

Frequently asked questions

How long should I paper trade a strategy?

A few weeks at minimum, and ideally long enough to collect a meaningful sample of trades for your strategy's pace. A day trading system can earn its comparison in two or three weeks; a strategy that trades weekly needs months, or a lower bar and a smaller first live size.

Is paper trading accurate?

Directionally yes, precisely no. Signals and timing are real, but fills are optimistic and costs are often understated, so paper results are a gentle upper bound on live results. The more liquid the instrument and the slower the strategy, the smaller the gap.

Why does my strategy work on paper but lose live?

The usual suspects, in order: execution costs the simulation ignored, fills consistently worse than assumed, an overfit backtest finally meeting unseen data, or a market regime change. Compare live fills to simulated fills first; that check isolates execution problems from strategy problems.