The transition from paper trading to live trading is where most algorithmic strategies fail. Not because the code was wrong. Not because the backtest was flawed. Because the trader overrides the first signal that feels uncomfortable, and the systematic trading process stops being systematic at exactly the moment it costs something real.

What Paper Trading Actually Tests

Paper trading tests code correctness and execution mechanics. It validates that orders are placed in the right format, fills are received and logged accurately, position tracking matches expected state, and the strategy runs end-to-end without runtime errors.

These are not small things. Getting them right matters. But they are also the floor, not the ceiling.

Paper trading does not test your emotional reaction to watching a real loss accumulate in real time. It does not test the temptation to override a rule when the market "feels wrong." It does not surface the slippage that occurs in illiquid conditions — paper fills assume the midpoint; live fills do not. It does not expose partial fills that alter position sizing mid-trade.

These are not minor gaps between paper and live. They are the gaps where most systematic traders fail when transitioning to real capital. A strategy that ran 50 paper trades without a single override will encounter its first override opportunity within the first week of live trading — typically during the strategy's first meaningful drawdown.

That moment is what paper trading cannot prepare you for. Only live trading can test whether you will hold discipline when it costs something.

What Live Trading Actually Tests

The most important test live trading runs is not the strategy. It is the trader.

The critical failure mode is not technical. It sounds like this: "The market feels wrong right now. I'll skip this signal." That sentence, once spoken aloud or silently, ends the systematic approach. You now have a rules-based strategy with a human override layer. The override layer will be wrong at the worst possible moment — because traders override when anxiety peaks, which is statistically correlated with the strategy being closest to a recovery point.

The override turns a temporary drawdown into a permanent missed recovery. The strategy would have recovered. You prevented it from doing so. The drawdown becomes a loss. The loss becomes evidence the strategy doesn't work. The strategy gets abandoned. The cycle repeats with the next one.

Beyond psychology, live trading also surfaces real execution quality differences that paper trading obscures. Slippage — the difference between the signal price and the actual fill price — can shift a marginally profitable strategy into unprofitable territory. Partial fills change position sizing in ways that backtest code never modeled. Network latency introduces timing differences between signal generation and order submission. None of these appear in paper mode. All of them appear in live mode on day one.

Metrics That Must Be Stable Before Going Live

Stability is not the same as profitability. A strategy does not need to be profitable in paper trading to be worth going live — paper fills are too clean to trust P&L signals directly. But it does need to be stable.

Over 50 or more paper trades, examine three metrics:

Win rate variance should stay below 5 percentage points across rolling 10-trade windows. If your win rate swings from 30% one week to 70% the next with no change in market conditions or parameters, the strategy behavior is unpredictable. Unpredictable behavior in paper becomes dangerous behavior in live.

Holding period should stay within 25% of the backtest average. If your backtest shows a mean hold of 4 hours and your paper trades are averaging 6 hours, something changed between backtest and execution — signal timing, order logic, or fill simulation. Find it before going live.

Max adverse excursion per trade should stay below your stop loss threshold 95% or more of the time. If trades regularly breach the stop level before recovering, your stop placement logic has a bug in the execution layer that paper trading's clean fills may have masked.

If any of these are unstable after 50 trades, the strategy is not mature. Deploy more capital into a mature paper run before moving to live. The cost of patience here is zero. The cost of impatience is real money.

The Readiness Checklist

Before the first live trade, verify every item:

  • [ ] 50+ paper trades completed and logged with timestamps and fill prices
  • [ ] Worst paper drawdown is within your acceptable live drawdown tolerance
  • [ ] Kill switch implemented: one command closes all open positions
  • [ ] Kill switch tested: verified it actually executes in paper mode, not assumed
  • [ ] Position size is fixed and written down before trade 1 — not adjustable mid-run
  • [ ] Monitoring configured: alerts on runtime errors, drawdown breach, and missed signals
  • [ ] "If this strategy loses 8 trades in a row, I will let it keep running" — your answer must be yes

The last item deserves attention. At a 55% win rate — a reasonable edge — 8 consecutive losses occur approximately twice in every 1,000 trades. The math is not intuitive: 0.45^8 ≈ 0.0017, which means it happens roughly 1.7 times per thousand trades, or roughly twice. It will happen. The question is whether you will override at trade 6 or 7 and turn a statistical fluctuation into a strategy abandonment.

If you cannot answer yes to that final item, the strategy is not ready. You are not ready. More paper time is the right answer.

The Oyamori Approach

Oyamori tracks live versus paper performance in parallel, showing the execution delta in real time — exactly where fills, slippage, and timing diverge from the paper simulation. The gaps are visible, not hidden inside aggregate P&L numbers.

The paper-to-live switch is a configuration change, not a platform migration. The same command structure, the same log format, the same monitoring configuration — the only difference is that orders route to a live brokerage account instead of a simulation layer. This removes one class of failure: the "I had to set up a whole new system to go live" problem that introduces bugs at the worst possible time.

The readiness decision, though, still belongs to the trader. No platform can answer the checklist for you. No dashboard can confirm that you will hold discipline through the first real drawdown. That answer comes from the 50-trade paper run and the honest conversation you have with yourself about item 7.

Next: The Oyamori CLI Workflow →