How Long Should You Paper Trade Before Going Live?

Jay Sharma · Founder, Botsfolio

Published September 21, 2026

Search this question and you get a number: thirty days, ninety days, six months. Or a trade count: fifty trades, a hundred trades. Then the same caveat every time, "it depends, be consistent."

None of that is wrong, exactly. It is just measuring the wrong thing. Calendar time and raw trade counts do not tell you whether you have an edge. And an edge is the only thing that makes going live a decision instead of a gamble. You can paper trade flawlessly for six months and still be trading a strategy that loses money over a large enough sample. The clock has nothing to say about that.

Here is the version of the answer that actually protects your account.

Why "trade for X weeks" is the wrong answer

Say you paper trade for two months and finish up 8%, with a 68% win rate over 40 trades. Are you ready?

You cannot tell from those numbers. Forty trades is a small sample. A 68% win rate over 40 trades has a confidence interval wide enough to include "no edge at all." If those 40 trades all happened in one steady uptrend, you have learned that your strategy works in an uptrend, which every strategy does. You have not learned whether it survives a chop or a news week.

This is the core problem with time-based and count-based rules: they measure your persistence, not your strategy's edge. They answer "did you show up?" when the question is "does this actually work?"

The traders who blow up after a great paper run almost always cleared the folklore bar. They put in the ninety days. They logged the hundred trades. What they never did was separate two very different questions:

1. Does this strategy have a real, repeatable edge?
2. Can I personally execute it without breaking my own rules?

Paper trading is how you answer #2. It is a terrible way to answer #1 on its own, because a small live sample in one market regime cannot distinguish a real edge from luck. You need to answer #1 with data first.

What "ready" actually means: two gates

Think of going live as a door with two locks. Both have to open.

Gate 1: the strategy has a measured edge

Before you spend a single week paper trading an idea, you should know whether the idea has ever worked. Not "it looks good on this chart", every strategy looks good on the chart you picked it from. You want the honest historical answer across a lot of instances, and it has to clear three tests that most "backtests" quietly skip:

• No look-ahead. The most common way a backtest lies is by confirming a setup using candles that had not closed yet at the moment you would have entered. It inflates every number. When we removed look-ahead from our own detectors, most setups that had looked strong collapsed to the lowest grade. That is the honest picture, and it is the one worth trading against. (Full detail on our SPI methodology.)
• Out-of-sample. Split the history by time. Measure the edge on the earlier slice, then re-check it on a later slice the strategy never saw. If the edge does not survive the unseen slice, it was curve-fitting.
• A second market. Re-run the same rules on a coin you did not tune them on. A real edge tends to show up on more than one market. One that only works on the exact coin you built it for is usually noise wearing a costume.

If a strategy cannot clear those three, no amount of paper trading will save it. You would just be collecting live evidence of a losing system. Fail Gate 1, and the answer to "how long should I paper trade?" is "don't, fix the strategy first."

Gate 2: the forward paper record confirms it

Once a strategy has a measured edge, now paper trading earns its keep. This is where you find out whether the edge that existed in history still shows up in the live market, and whether you can actually trade it.

A forward paper record that means something has four properties:

1. Enough trades. Not a fixed number, but enough that a couple of lucky or unlucky trades cannot swing the verdict. Twenty is a floor for a first read; the more the better. A 90% win rate over ten trades tells you almost nothing.
2. Across regimes. Your sample should include a trend, a chop, and at least one high-volatility or news-driven stretch. A strategy that only survives calm uptrends is a bull-market strategy, and you should know that before you find out with real money.
3. Realistic costs. Fees and slippage are not rounding errors on an active strategy. A forward record computed net of round-trip costs is the only one worth trusting. (More on why in Why Your Paper Trading Results Lie.)
4. Rules followed. If you moved your stop, sized up after a loss, or skipped the setups that scared you, your paper record is measuring a different strategy than the one you backtested. Clean execution is part of the test.

When those four hold and the forward record still shows the edge, the door is open. The number of weeks it took to get there is just how long it took, not a milestone you were aiming for.

Paper trading is not practice. It is how you find your edge.

Here is the reframe that changes everything about this question.

Most guides treat paper trading as a rehearsal, a way to "get comfortable" before the real thing. That framing is why people paper trade for months and learn nothing: they are just going through the motions, waiting for a calendar to give them permission.

The useful way to think about paper trading is as a discovery loop. Every paper trade is a real, timestamped data point about what actually works for you, in this market, right now. Reviewed properly, that record does not just tell you whether to go live. It tells you what your edge is and how to sharpen it:

• Which setups carried your results, and which quietly bled them.
• Which timeframe and which coins your edge actually lives on.
• Whether your winners are big enough relative to your losers to survive a rough patch.
• Where your own execution is adding or subtracting from the strategy's raw edge.

That is the difference between "I paper traded for 90 days" and "I paper traded until I could see, in the data, exactly where my edge comes from and confirmed it forward." The second person is ready. The first is just tired of waiting.

This is the loop Botsfolio is built around, and it is why we log every paper trade automatically. You do not journal by hand and hope you remember your reasoning. The record writes itself with the structural context attached, and the companion reads it back to you: here is what is working, here is what is dragging, here is the change that would have helped. You watch the number move as you improve, which is the whole point. (We go deep on the auto-journal in What a Companion Sees That You Miss.)

A worked example: one idea through both gates

Abstract gates are easy to nod along to and hard to act on, so here is one idea taken all the way through. The idea: on the 4-hour, enter the order block that forms right after a liquidity sweep, stop below the sweep, first target at 1R. We picked it because it has a real edge to show, but the process is identical for any idea you type in.

Gate 1, prove the edge. Backtested with no look-ahead across the coins we track, here is what that confluence has actually produced, live from our backtest engine:

Grade it the way you would grade anything: is expectancy positive after fees, is the sample large enough to trust (n of at least 20, ideally 100-plus), and does the edge show up on more than one coin? Where it does, the idea clears Gate 1. Now contrast that with a typical hand-picked idea: run most "this looks great" setups through the same no-look-ahead test and the edge usually evaporates, the grade lands at D and expectancy sits at or below zero. That is the strategy telling you not to bother paper trading it, and it is far cheaper to hear it from the backtest than from your account. The final Gate 1 check is out-of-sample and cross-coin: measure on older history, confirm it holds on a later slice it never saw, and re-run it on a coin you did not tune it for.

Gate 2, confirm it forward. Now, and only now, paper trade it. A forward record worth acting on might look like this (illustrative): over six weeks the setup fires 24 times, 15 winners and 9 losers, +6.2R gross and about +4.1R after costs, and, crucially, those trades span a clean trend, a two-week chop, and a CPI-week spike. The edge that existed in history showed up live, across conditions, net of costs, and you followed your rules on all 24. That record, not the calendar, is your green light. If instead the forward record drifts, say the win rate holds but the average loss balloons, it has told you something specific: the edge is intact but your execution or your fills are not. You fix that before real money, not after.

What "not ready" looks like

Most advice tells you what "ready" looks like and stops there. The failure shapes are more useful, because they are what you will actually see.

Gate 1 failures all say the same thing, fix the strategy, do not paper trade it yet:

• The honest, no-look-ahead backtest grades it D, or expectancy is at or below zero. The idea has no edge. Paper trading it just buys a slower, more expensive version of the same conclusion.
• It only works on the coin you built it for. That is a fit to one market's quirks, not an edge.
• It is strong on the old data and falls apart on the held-out slice. That is curve-fitting, and it will not survive live.

Gate 2 failures say the edge may be real, but this record cannot confirm it yet:

• Every trade happened in one market regime. You have proven it works in a trend, not that it works.
• It is positive before costs and negative after. That is a fee generator, not an edge.
• Your actual exits do not match your planned stops. You are measuring a looser strategy than the one you backtested, and the live version will be looser still.

In every case the fix is specific, and it is almost never "keep paper trading unchanged for another month."

Why more time can make you worse, not better

There is a real cost to paper trading too long, and almost nobody warns you about it.

The first is false confidence from a single regime. Paper trade through a three-month bull run and you will finish convinced you have an edge, when what you have is a long position in a rising market. The longer that regime lasts, the more certain, and more wrong, you become.

The second is drift. After a few hundred simulated trades, discipline erodes precisely because nothing is at stake. You start taking marginal setups and nudging stops because it does not matter, so your paper record quietly starts measuring a sloppier trader than the one who built the plan.

The third is recency curve-fitting. Trade one idea long enough and you will unconsciously tune it to whatever the market just did. That is overfitting by hand, and it breaks the moment the market does something new.

This is why the answer is gates, not a clock. Once a strategy has a measured edge and a forward record across regimes, more paper trading adds noise, not confidence. The next honest test is a small amount of real money.

A readiness checklist that beats "count the weeks"

Trade this list, not the calendar. You are ready to put real money on a strategy when you can check every box:

1. Edge, measured. The strategy is positive expectancy on a large historical sample, computed with no look-ahead.
2. Edge, out of sample. It stayed positive on a held-out slice of history it was never tuned on.
3. Edge, portable. It held up on at least one coin you did not build it for.
4. Forward sample. You have paper traded it live over enough trades to trust the read, ideally 20-plus.
5. Regime coverage. That forward sample spans a trend, a chop, and a volatile or news-driven stretch.
6. Costs included. Your forward numbers are net of realistic fees and slippage, not perfect fills.
7. Rules kept. You followed your own entry, stop, and sizing rules on the large majority of trades, and you can prove it from the log.
8. A plan you could hand off. You can write the strategy down clearly enough that someone else could trade it without asking you a single question.

Boxes 1-3 are Gate 1. Boxes 4-8 are Gate 2. If any box is blank, that is your answer for what to do next, and it is almost never "wait longer."

Going live: start small, scale on the record

Clearing both gates is permission to start, not permission to size up. The forward record does not stop when real money begins, it just gets more honest, because now the emotional variable is switched on.

Go live at a size where a normal losing streak cannot hurt you, small enough that fear does not change your decisions. Keep the log running. Compare your live behavior to your paper behavior: if your win rate holds but your average loss suddenly grows, that is the emotional gap showing up as moved stops, and it is exactly what to fix before you scale. Increase size only after the live record confirms you are trading the same way you did on paper. The strategy already earned the size; you are just proving you can hold it.

Concretely: start at a risk-per-trade small enough that a normal losing streak is boring. If a five-loss run at your chosen size would make you change how you trade, the size is too big, cut it. Run the first 20 to 30 live trades and compare them to your paper record. If your live expectancy tracks your paper expectancy and your average loss has not grown, step risk up one increment, for example 0.5% to 0.75% to 1% per trade, one step per profitable block of trades. If your live average loss is suddenly bigger than your paper average loss, hold size flat and fix that first, because scaling a leak just scales the damage.

FAQ

Educational analysis, not financial advice. Past performance does not predict future results.