A backtest catches
A rule with no edge, a rule that only worked in one regime, parameters that will not hold, and — with a walk-forward and an audit — overfitting and look-ahead. It is fast, repeatable, and covers years.
Method
They answerdifferent questions
A backtest asks whether the rule would have worked. Paper asks whether the rule, this platform, that venue and your own hand can execute it now. Passing the first tells you nothing about the second.
What each one catches
The overlap is smaller than people expect, and the gaps are where accounts get lost.
A backtest cannot catch
Latency between signal and fill, a venue rejecting your size, a stream that drops, a stop that was never actually placed, or you deciding at 3am to override it.
Paper catches
The whole path: rounding to the instrument’s filters, the risk gate refusing something you thought was fine, capture at your size, protective exits arming, reconciliation after a disconnect.
Paper cannot catch
Regimes not happening this month, and the emotional half of live trading. A simulated book also never rejects you for reasons a real one will — which is why a certified venue rehearsal is still not the same as the venue.
The order that works
Backtest first, because it is free and fast
Most ideas die here, and they should — it takes minutes and no market has to cooperate. Our own strategy search kills most candidates on the first run, which is the system working rather than failing.
Then paper, at the size you actually intend
Not a token size. Capture and rounding are size-dependent, and a rehearsal at a tenth of your intended position proves the wrong thing. Paper is where you find out that the majors round a third of a position away at your size while the mid-caps do not.
Then live, deliberately, and small
Going live needs your own venue credential and a human promotion with a scope — it is not a toggle, on purpose. The first live size should be smaller than the paper size, because the one thing neither test measured is you.
Questions about the two
How long should I paper trade?
Long enough to have taken the trades the strategy takes, including a losing streak. A week of a strategy that trades twice a month proves nothing except that the plumbing works — which is worth knowing, and is not the same as the strategy working.
Do paper results transfer to live?
The execution mechanics do, closely: the same rounding, the same gate, the same order path. The fills do not transfer exactly — a simulated book is a model of depth, not the book itself.
Can I run both at once?
Yes, and it is the most useful comparison available: the same strategy paper-trading forward while you backtest variations behind it, on the same series.
Do the second half for free
A paper account is the whole system minus the exchange — which is the half a backtest never touches.