Paper trading, and how long to do it

Paper trading means the strategy runs for real — real prices, real timing, real decisions, recorded as they happen — but no order ever reaches a brokerage account.

It’s the one test that can’t be gamed. A backtest can be tuned until it flatters you. Forward time cannot: the data genuinely hasn’t happened yet, so there’s nothing to fit to.

In BotStock it needs nothing to set up. Every account starts with its own paper account: a simulated brokerage with practice money, filled against the same live quotes your strategies read. You don’t need a brokerage account to begin, and when you do want one, the same strategies switch to it.

How long

Longer than feels necessary. The useful measure isn’t days, it’s trades.

A strategy that trades twice a month gives you 24 data points a year — nowhere near enough to distinguish skill from luck. One that trades daily gives you a usable picture in a couple of months. So the honest guidance is: paper trade until the strategy has been wrong a few times and you’ve seen how it behaves when it is.

Thirty to fifty trades is a reasonable minimum before you start taking the numbers seriously. Even then, treat them gently.

What to actually watch

Not the profit. Three other things:

Did it do what you expected? If you’re regularly surprised by a trade, you don’t understand your own rule yet. That’s worth fixing before money is involved.
How did the losing stretches feel? There will be a week where it’s down and you want to switch it off. Noticing that now — for free — is the most valuable thing paper trading gives you.
Does the paper result resemble the backtest? A big gap usually means the backtest was overfitted or was peeking.

How the paper account fills, and why it’s a little unkind

Every paper account anywhere is a simulator: it marks your orders against real prices, but nothing goes into the real order book. The differences are in the fill rules, and a simulator that fills you better than a real broker would is teaching you the wrong lesson at the worst possible time.

So BotStock’s paper account errs the other way, on purpose:

You always pay the spread, plus a little slippage. Buy at the ask, sell at the bid, and a few hundredths of a percent worse than that. Buy 100 shares quoted at $200.00 and you’ll pay about $200.10 — the ten cents a real order usually loses.
A limit order needs the market to trade through it. Reaching your price isn’t enough, because a real order at that price is at the back of a queue and often doesn’t fill.
A stop fills where the market is, not where the stop was. A stop at $37 in a stock that gaps to $34 overnight fills near $34. That’s exactly what a real stop does, and it’s the thing paper most needs to be honest about.
Two account rules a real brokerage enforces. In a cash account, today’s sale proceeds can’t be spent until tomorrow. And four day trades in five sessions on an account under $25,000 gets the next new position refused — the single most common surprise for a bot that trades intraday. The paper account can simulate both, and will tell you how close you are.

What it can’t simulate is you: a real order moves the market a little and misses occasionally, and none of that shows up on paper. On liquid stocks the difference is small. On thin ones it can eat the whole edge. So expect live results to be a bit worse than paper, and if a strategy is only barely profitable on paper, treat it as probably not profitable at all.

Then what

Going live is a few deliberate steps, none of which can happen by accident:

Connect a brokerage. Under Settings, with a sandbox token first. The sandbox is a rehearsal of the connection — your token, your account number, the order plumbing — with the brokerage’s own practice money.
Switch the account to it. Refused while anything is open on paper, and it stops the engine. Your strategies, settings and track record all carry across; positions don’t.
Switch that brokerage to production. A separate control, behind a typed confirmation and the live-trading gate.

There’s no rush — a strategy worth trading next month will still be worth trading next month.