The moment a strategy moves from paper to live is the moment real capital is at stake. The platform treats that as a deliberate decision, not a toggle you can flip by accident. The paper-to-live gate is a confirmation checkpoint the bot refuses to skip, and clearing it well is a matter of working through a short, calm checklist rather than rushing.
Why a gate exists at all
A backtest can pass and a paper run can look healthy, and the market can still behave differently when live orders touch it. The gate forces a pause at exactly the point where optimism is highest and the cost of a mistake is real. It is friction on purpose.
What you confirm before you cross
- Paper results that earned trust — a sustained run that cleared costs and tracked its backtest, not a single good session.
- Keys scoped to trade only — the live Kraken API key grants querying and trading, with withdrawals firmly disabled.
- A loaded risk policy — per-trade limits, the daily cap, cooldowns, and the kill switch are all configured before the first live order can fire.
A worked example: reading a paper run before you cross
"Paper results that earned trust" needs a definition you can point at, not a feeling. Suppose a strategy has run in paper mode and the audit trail summarises like this:
| What you check | This run | Reading |
|---|---|---|
| Paper duration | 6 weeks | Long enough to span calm and volatile sessions |
| Trades taken | 74 | A sample, not a handful |
| Signals rejected on cost | 211 | Cost rule is filtering, as designed |
| Net after modelled costs | positive | Cleared its own friction |
| Paper vs backtest gap | −4% | Tracking, not diverging |
| Risk events triggered | 2 cooldowns, 0 kill-switch | Limits fired and behaved |
Every line above points the same way, so the paper criterion is genuinely met. Had the paper-vs-backtest gap been −30%, or the net negative once costs were modelled, or the trade count only a dozen, the disciplined answer is to stay in paper — no single green box overrides a red one. The gate is a conjunction: all conditions, not most.
Crossing it is a choice
The bot will not promote itself to live on your behalf. Clearing the gate is an explicit action you take once the checklist above is genuinely satisfied. If any item is uncertain, the disciplined move is to stay in paper and resolve it first — the gate is there to be respected, not beaten.
Common questions
How long should a paper run last before I consider the gate?
Long enough to cover more than one kind of market — calm and volatile — and to accumulate a real sample of trades rather than a lucky session. The point is that the run tracked its backtest with costs applied, not that it hit an arbitrary day count.
What if I clear the gate and the bot stops trading anyway?
That is the risk policy doing its job: a daily cap, a cooldown, or the kill switch can halt a freshly live bot. Read what fired in how to read the execution audit trail, and if the kill switch tripped, follow how to recover the bot after a kill switch.
Is there a single list I can run before going live?
Yes — a pre-launch checklist for a Kraken bot walks keys, paper proof, risk policy, and audit trail in order, so nothing is left to memory.
For the differences on each side of the gate, read paper mode vs live mode. For the key setup it depends on, see how to create and scope Kraken API keys. The limits that must be loaded first live in the risk pillar. Going live never guarantees a profitable trade.