Most people who want to become profitable traders do the same thing. They learn a strategy from a YouTube video, open a live account, and start pressing buttons. Then they lose money for a few years and conclude that trading does not work.
But trading does work. Taking shortcuts, skipping steps in the process and not putting in the work do not.
There are five stages, and they run in a fixed order:
- Learn a trading approach until you can see it on a chart.
- Backtest it until you have a model with a real edge.
- Forward test it on live data without trading.
- Paper trade it until your results match your backtest.
- Go live, and start the real work of managing yourself.
The order is not arbitrary. Each stage produces the thing the next stage needs. You cannot backtest an approach you cannot identify. You cannot forward test a model you have not defined. You cannot paper trade rules you have not written down. Skip a stage and you are not moving faster, you are just arriving at the next one without the input it requires.
David Nowak, who is building BacktestingLab, skipped most of them. He went straight from learning to live trading, and it cost him roughly five years, a lot of money and a lot of mental capital. You can read about his hard trading lessons over those five years.
This framework comes from ICT (Michael J. Huddleston), who teaches it as education, backtesting, forward testing, demo trading, and then live. It is the clearest articulation of the process out there, and it works whether or not you trade his concepts.
Stage 1Learn
The goal of this stage is not to make money. It is to be able to look at a chart and see something specific.
Pick one approach
There are more ways to trade than anyone can learn, and they are not equally suited to everyone. Broadly, the approaches worth knowing about:
- Smart money concepts (ICT). Liquidity, market structure, order blocks, fair value gaps. Discretionary, deep, and popular. A long learning curve and a lot of free material.
- Classic price action. Support and resistance, trend lines, chart patterns, candlestick reading. The most widely taught, and the most widely regurgitated.
- Supply and demand. Zone-based, mechanically simpler than smart money concepts, and it overlaps with them more than either camp admits.
- Indicator-based systems. Moving averages, VWAP, RSI, MACD and the rest. Easy to learn and easy to automate, but indicators are derived from price and therefore lag it.
- Mechanical and algorithmic systems. Fully rule-based, testable at scale, and the natural fit if you can write code. Little room for discretion, which is either the appeal or the problem depending on who you are.
Pick one and go deep. The most common way to waste a year is to sample five approaches and master none of them, because every time results get uncomfortable there is a new method promising it will be different.
It will not be different. Every approach has losing streaks. Switching during one just resets your progress.
Where to learn it
- Free material is genuinely good now. Full mentorships sit on YouTube for nothing.
- Paid courses can be worth it, and can be fraud. The tell is whether the teacher trades live.
- Traders who stream live are the single most useful resource in this stage, and the most underused. Watching someone identify a setup in real time, take the trade, and handle a loss teaches you things no recorded lesson can, because nothing is edited and nothing is known in advance.
- Hindsight screenshot posters teach you nothing. Anyone can mark up a chart after the move. If someone only ever shows you completed trades, you are watching marketing.
- Communities help with motivation and blind spots. They also produce herd behaviour and signal-chasing. Use them for questions, not for entries.
How long this takes
Months. Not weeks.
You are not learning definitions, you are training pattern recognition, and that only comes from screen time. Expect the middle of this stage to feel like you are not progressing, because progress here is invisible right up until the point where a chart suddenly reads as obvious.
You are ready for stage 2 when
- You have picked one approach.
- You know its concepts well enough to explain them.
- You can identify them on a chart yourself, without someone pointing them out.
- You feel the itch. You want to stop learning and start doing.
That last one is the real signal, and it is also the trap. The itch is exactly what sends people to a live account. Don't do that. It only means you are ready to backtest.
Stage 2Backtest
This is the stage that traders who never make it tend to skip. It is also the single most expensive mistake David made in five years of trading, which is a large part of why he is currently building his own backtesting software.
Don't take my word for it. Go and backtest it.
ICT (Michael J. Huddleston)
First, two different things are called "backtesting"
Almost all of the confusion about this stage comes from the fact that the word means two completely different activities, and doing the wrong one wastes months.
1. Strategy development. You do not have a model. You are trying to build or validate one, and find out whether it has an edge at all. The output is a set of rules with statistics attached.
2. Practice reps. You already have a model. You jump to a random historical date, replay the chart forward, mark it up, take your setup when it appears, and record the result. The output is repetitions and an equity curve.
Both are useful. They are not interchangeable.
If you do not have a strategy, strategy development is the only thing you should be doing. Practice reps on a model you have not validated just teaches you to execute something that may not work.
Where a strategy comes from
Two options.
Invent your own from the concepts you learned in stage 1. This is the romantic option and it is not the recommended one. You are combining ideas you have never tested into a system nobody has ever traded, and you will spend a long time discovering why it does not work.
Take a model from a trader who trades live and shares it, backtest it yourself, and then add your own flavour once you understand its behaviour. This is faster and far more likely to work, because you are starting from something that has already survived contact with a real market.
Note the qualifier: a trader who trades live. Not someone who posts winning screenshots.
The three ways to do it
Manual. You scrub through historical charts by hand, wait for your setup, mark it, record it. This is how most discretionary traders start, and it works, but understand the cost. You need at least 100 to 200 tests executed identically, it is slow and tedious, and there is a real bias problem: you tend to already know what the price did next. It is very easy to award yourself entries you would never have taken live.
Fully programmatic. You write the strategy as code and run it over years of data. Excellent if you are building a trading bot or a fully mechanical system. You can test one year or ten, and see which market conditions your strategy thrives in and which ones break it. The limitation is that it only tests what you can express in code, which rules out most discretionary judgment.
Semi-automated. The software finds every occurrence where your conditions are met, and you review them one at a time. For each, you look at the chart and decide: I would take this, or I would pass on this because there was no divergence, or the liquidity sweep was not convincing enough.
That third method exists because it matches how discretionary traders actually trade. Most discretionary strategies are 60 to 90 percent mechanical, with the rest being human judgment. A fully automated test cannot capture that last part, and a manual test cannot give you the sample size.
Match the model to yourself
A strategy with a 50 percent win rate is a bad strategy for someone who takes losses badly, no matter how good its numbers are.
If a run of losses puts you on tilt, you may be better served by a model with a 70 or 80 percent win rate even though each winner is smaller. The profitability can be similar. The experience of trading it is completely different.
Know yourself before you decide which statistics you are optimizing for. The best model on paper is worthless if you cannot personally execute it.
You are ready for stage 3 when
- You have a defined model with a good profit factor.
- It matches your personality and your tolerance for losing streaks.
- You have tested it over at least three digits of trades. A hundred is the bare minimum and more is better. There is deliberately no exact number here, because the right sample size depends on how often your setup occurs.
- You can state your win rate, average winner, average loser and drawdown from memory. If you cannot, you do not know your model yet.
A note on tooling: use software you can see into. A number at the end of a black box tells you that a strategy worked. It does not tell you why, or when, or what the losing trades had in common. That last part is where the improvement actually comes from.
Stage 3Forward test
Now you have a model with statistics. Forward testing is where you find out whether you can spot it in real time, on a chart that has not finished forming.
You do not press any buttons in this stage. That is the entire discipline.
What you actually do
Show up during the times your backtest says your setups occur. Watch. When you see your setup form, write down where you would have entered, where your stop loss would go, and where you would take profit.
Then keep watching, and record what happened. Stopped out, or target hit.
That is it. One to three months of it.
Why this stage exists
The data is a secondary benefit. The real purpose is you.
Forward testing builds the ability to sit in front of a market and do nothing for hours, which is the single hardest skill in discretionary trading and the one nobody practises. Most traders cannot do it. They came to trade, so they trade, and a strategy that produces two setups a week becomes a strategy that takes fifteen.
Doing this with nothing at stake, day after day, is how patience stops being a personality trait you either have or lack and becomes something you have actually trained.
You will also find out something uncomfortable: your setup is much harder to see when the candles are still moving. In a backtest the pattern is complete. Live, you are looking at something half-formed and deciding whether it will become your setup or not. That skill is what this stage buys.
One thing to decide here: partials
Do you take partial profit, or run the full position to target?
ICT recommends taking partials. Others argue that partials cut your winners and quietly destroy your risk-to-reward ratio. A third approach takes most of the position off at target and leaves a runner in case the move continues.
There is no universal answer, and you should not accept one from anyone. This is the stage to test which version you can actually execute, since it changes the statistics you carry into paper trading.
You are ready for stage 4 when
- You see your setup and you are certain.
- No second-guessing. None of "maybe this counts, I am not sure".
- You are not talking yourself into something that is nearly your setup.
- When your conditions are met you know it, and when they are not you know that too.
Uncertainty here becomes hesitation later, and hesitation with money on the line becomes a missed entry or a chased one.
Stage 4Paper trade
Forward testing taught you to see it. Paper trading is where you press the button.
Treat the account as real money
This is not a soft recommendation. It is the requirement that makes the stage work at all.
If the balance is not real to you, you will take trades you would never take, size positions you would never size, and learn nothing transferable. The results will be meaningless and, worse, they will be encouraging.
If that is you, skip this stage. Some people genuinely cannot make fake money feel real, and there is no point pretending otherwise. Go to live trading with the smallest position size your broker allows, an amount you truly do not care about losing. Real money at trivial size beats fake money you do not respect.
Make the conditions realistic
This is where most paper trading quietly fails.
- Do not paper trade at 500:1 leverage if you will go live at 20:1.
- Do not trade 100 contracts in a simulator if you will trade one.
- Do not paper trade a $100,000 balance if you are funding a $2,000 account.
Every one of these produces statistics that will not survive the transition, and a psychological baseline that is completely wrong. The purpose of this stage is a smooth transition to live, and unrealistic settings guarantee a rough one.
You are ready for stage 5 when both halves are true
The statistical half. Your paper trading results are comparable to your backtested results, within a reasonable tolerance. Compare all of it:
- Profit factor
- Win rate
- Number of consecutive losses
- Average drawdown
- Average winner and average loser
And do it for at least three consecutive profitable months. One good month is variance.
If the numbers diverge sharply from your backtest, that is information, not failure. Either you are not executing the model you tested, or the model does not hold up in live conditions. Find out which before real money is involved.
The emotional half. This one matters just as much and gets ignored.
- No spike of emotion when you enter.
- You wait until every condition is met, then execute without hesitation.
- You follow your rules when following them is inconvenient.
- Five losses in a row does not shake you, because you know what your model does over hundreds of trades.
ICT's marker for this is that trading has become boring. That is the correct target. Boredom means the outcome of any single trade has stopped feeling significant, which is precisely the state you need before scaling up.
When both halves are true, you have earned genuine trust in your system, and going live is a change of account rather than a change of behaviour.
Stage 5Go live
There is no exit criteria for this stage. This is the part that does not end.
You know you are ready to be here when every condition from the previous four stages is met. Not when you feel like it, not when you are bored of paper trading, and not when you see a setup you cannot stand to miss.
The strategy is solved. Now the problem is you
By this point your edge is defined and tested. Almost everything that goes wrong from here is you, which means the work changes completely:
- Risk management. How much of the account is at risk per trade, and per day.
- Position sizing. Covered below, because it deserves its own treatment.
- Emotional control. Trading while angry, anxious, or distracted by something outside the market.
- Tilt. Trading to recover a loss instead of trading your model.
- FOMO. Entering because the move is happening, not because your conditions were met.
- Revenge trading. The most expensive of all of them.
- Rule breaking. Widening a stop. Removing a stop. Cutting a winner early. Adding to a loser.
Every one of these is a version of the same failure: overriding a tested system with an untested impulse.
Scaling up is its own gauntlet
Every increase in position size introduces problems that did not exist at the previous size.
At a few dollars a trade, nothing bothers you. At tens, you notice. At hundreds, you start watching the profit and loss instead of the chart. At thousands, everything you thought you had solved comes back.
The test is simple. Can you sit through a $9,000 loss and feel nothing, because it is your model doing exactly what your model does? Then that is your size.
If you are getting antsy, closing early, moving your stop, or checking the position every few seconds, you are trading too big. That is not a discipline problem to push through. It is a sizing problem, and the fix is to size down, get comfortable, and step up again more slowly.
Keep a trading journal
Record every trade, and record your state going into it.
A journal is the only way to find out what you actually do, as opposed to what you believe you do. Patterns that are invisible day to day become obvious across fifty entries: that you lose on Fridays, that your losers cluster after a big winner, that your rule breaking always follows the same trigger.
You cannot fix a pattern you cannot see.
Adapt, but slowly
Market conditions change, and a model that worked for two years can stop working. You need to notice this without confusing it for a normal losing streak, which is exactly what it looks like from the inside.
This is what your backtested statistics are for. A drawdown inside your historical range is normal. A drawdown well outside it, sustained, is a signal to go back and test what changed. Your statistics are what tell the difference between a bad month and a broken model.
Stay in the game
- Cash out profits regularly. Money withdrawn cannot be given back to the market.
- Take a break after a very profitable run. Overconfidence costs more than losses do.
- Stay humble. The market has removed better traders than you.
- Do not overtrade. Most of the job is waiting.
The aim is to still be trading in ten years, not to get rich this quarter.
The contrarian take: this might really be three stages
Now that you have the five, here is where they are worth arguing with.
Proper backtesting already contains a lot of forward testing. If you are replaying historical candles, going setup by setup, marking up charts and watching your model succeed and fail in sequence, you are doing something close to what stage 3 asks for. Some backtesting tools blur into paper trading as well, so you can end up getting three kinds of reps out of one activity.
So if you already have live trading experience, and you have put genuine time into backtesting a model, you can compress the middle:
Learn → Backtest → Go live at the smallest size you can trade.
The logic is that live trading at trivial size is forward testing and paper trading, except the money is real, which fixes the one weakness paper trading has. You are proving that your backtested statistics hold up under live conditions. Use the smallest contract available and an amount you genuinely do not care about losing. If your live numbers match your backtested numbers, the account grows, you are already in the groove, and the only thing left is scaling up.
And scaling up is where your psychology gets tested anyway, so you may as well get there sooner.
Two things stay non-negotiable.
You must learn the concepts properly. And you must backtest the strategy, which means real strategy development, not curve-fitting a model until the historical equity curve looks beautiful.
People will tell you that you can skip backtesting. For the overwhelming majority who try, it does not work. It did not work for David.
The process is the edge
Everyone looking for the secret to profitable trading is looking one level too high. There is no indicator, no mentor, and no strategy that substitutes for having tested the thing you are about to risk money on.
Work through these stages properly and profitability becomes a realistic expectation. People fail because they take shortcuts. They skip the backtest, they skip the reps, they go live because live is the exciting part, and then they sit at a screen risking money on something they never verified and cannot describe.
That feeling of not quite knowing what you are doing is not a phase you trade your way out of. It is the direct result of skipping the stages that would have told you.
Start where you actually are, not where you would like to be.
Not sure which approach to learn first? Browse the guides →