The Final Check

Three months in a test account has a different job from proving the edge again. It is where execution, timing, behaviour, and code safety get their final check before live capital.

Someone asked me a couple of weeks ago how long I sit a strategy in a test account before it goes anywhere near my live book. The answer surprised them: I said at least three months, no matter how good the numbers look on the backtest.

They assumed paper trading was a formality. Prove the edge one more time, tick the box, go live. That is not what those three months are for. The backtest already argued the edge exists. Paper trading is where I find out whether the edge survives contact with everything the backtest never had to deal with.

So what is paper trading actually testing?

Three things, and none of them is "is the edge real".

Execution. My backtest fills me at a price. Live, the fill is a negotiation between my order, the broker's feed, and whatever the spread is doing in that exact second. On paper, in real time, I get to watch the gap between the price my model wanted and the price the market gave me. If that gap is wider than my costed assumption, the edge I validated is thinner than I think, sometimes to the point of not existing. I can also find my capacity on that strategy by watching the live fills, much easier than on a backtest.

Behaviour under a live clock. A backtest reacts the instant a bar closes. My VPS could have dropped or the platform might have dropped its connection. Paper trading in real conditions shows me how much of the edge depends on being filled at a moment I cannot actually guarantee I will be present for. If a strategy only works when the fill is instantaneous, I want to know that now.

Me. I still feel the pull to interfere when a paper strategy is two weeks into a drawdown, even though I know it's normal. Watching that happen on a demo, where nothing is at stake, is far easier. We're just practicing. By the time real money is on it, I have already sat through the discomfort once and learned that the correct action was to do nothing.

Why three months and not three weeks?

Because a fortnight of paper trading only ever shows you one market. If those three weeks happened to be a calm, trending stretch, the strategy looks immaculate and you have learned nothing about how it handles a chop, a gap, a news shock, or a slow grind where it takes trade after trade and none of them work. Three months has no magic in it. It is usually long enough to contain more than one condition, which is the whole point. I want to see the strategy have a bad week on paper before it has one live.

The paper-trading period is a filter, and filters only work if you actually let them run.

This is the slow, boring part of the process, which is exactly why it is the easiest to cut. Cutting it is what puts fragile systems on live capital. The paper-trading period is a filter, and filters only work if you actually let them run.

When Good Backtests Still Fail

Here is the uncomfortable part: sometimes a strategy that passed every robustness test I could throw at it, walk-forward, parameter stability, a second broker's data, still falls apart on paper. The edge may still have existed. The execution reality ate it. That quarter still earns its keep. The process did its job away from my live book.

The bar for going live

The bar for going live is simple in my book. The paper period has to reproduce something close to the backtest's behaviour, through at least one stretch where the strategy was uncomfortable, with an execution cost that matches what I modelled. Clear that, and it earns a small live allocation. Miss it, and it goes back to the research pipeline or the graveyard.

But if you've made it this far, then the secret is that 3 months is generally long enough to see any "security" related issues. Are you getting filled when spread blows out? Is it executing through news releases? Is it clustering trades? Performance is one half of it. I also want proof that the code is safe.

Three months feels slow when you are itching to trade something new. It is a lot faster than finding out live that your fills were fiction.