Philosophy of Science: Science advances by testing ideas against evidence, not by defending beliefs.
Replace prediction and opinion with observation, evidence and repeatable decisions
Here is Chapter 1 from my book: How to Trade Like a Scientist: Use the Principles of Science to Build, Test, and Continuously Improve a Profitable Trading System
Two traders opened the same chart on the same morning. The first one says the stock looks strong and is probably beginning a new uptrend, so they buy it. The second pulls up a spreadsheet of 74 prior bullish 5-day / 20-day EMA crossovers, finds that the setup produced about a 60 percent win rate with an average 2:1 reward-to-risk ratio over a large sample of trades, and buys it too. Same trade, same entry price, same afternoon of uncertainty while the stock pulls back fifty cents after entry. Only one of those traders can tell you what they expect to happen over the next hundred trades, and only one of them knows they are executing a strategy with positive expectancy instead of acting on a hunch.
That difference is the whole subject of this book, and it starts with a procedure borrowed from the laboratory.
Observe something specific.
Propose an explanation that could be wrong.
Build a test capable of disproving that explanation.
Gather enough data so that luck isn't driving the answer.
Compare the result against what would have happened.
Then run the idea again on data that you have never seen.
Six steps. None of them asks you to predict anything.
Prediction is the habit that makes most retail traders unprofitable, and can be a slow leak over the years, while the trader feels intelligent the entire time. A trader who trades like a scientist exchanges a bad question for a much better one. Instead of asking what the market will do next, they ask how often a defined situation has produced a defined outcome, by how much, and measured against what. You will never know what the next trade does. You can know, within a stated margin of error, what one hundred trades of the same kind have done. That is a different kind of knowledge. Only the second one pays.
Observation Comes First
Every scientific discovery begins with someone writing what actually happened. For traders, that means a journal, though probably not the kind you have kept before. The emotional diary, where you record that you felt anxious and jumped in early, has some value for behavior, but it produces no data you can sort, filter, or test. What you want is a structured dataset with identical data fields on every row.
Record the date, ticker, setup name, entry price, initial stop, exit price, position size, and result in R multiples. R is your unit of risk, defined as the distance between your entry price and your original stop loss.
For example, suppose you buy a stock at $100 with an initial stop at $95. Your risk is $5 per share, so 1R equals $5. If you exit at $110, you earn $10 per share, or +2R. If you exit at $95, you lose $5.00 per share, or -1R.
Measuring every trade in R standardizes your results regardless of the stock's price or the number of shares traded. It allows you to compare trades objectively and evaluate the actual performance of your trading system, rather than being misled by raw dollar gains and losses.
Then record the context based on the market parameters you want to analyze. Was the S&P 500 above or below its 50-day moving average that day? What was the stock's Average True Range as a percentage of price, what was the volume against its 20-day average, which sector, and what time of day did you enter? A lot of columns of what feels like clerical work, but gives you valuable data to analyze. What is in those columns depends on which signals you find most valuable.
For the first few dozen trades, keeping detailed records can feel like busywork. However, with a much larger sample, patterns emerge. You may discover that your biggest losses consistently occur under certain market conditions, that one setup significantly outperforms another, or that trades taken outside your written rules rarely succeed.
Those insights don't come from memory. They come from data. A few minutes spent recording each trade creates a database that can reveal weaknesses to eliminate and strengths to exploit. Over the course of a trading career, those small daily habits can become one of the highest-return investments a trader ever makes. Today, many software platforms automate this.
Turning a Belief Into Something That Can Fail
Write an observation, and your brain immediately hands you a story explaining it. That story is a hypothesis. It only has value if it is specific enough to be proven wrong.
"Moving average crossovers make money" is an opinion, not a hypothesis. There is nothing specific to test.
Now compare it to this statement: when a specific stock's 5-day EMA crosses above its 20-day EMA, the trade produces a 60 percent win rate with an average reward-to-risk ratio of 2:1 over the next twenty trading days, when the same cross-under is used as an exit signal.
Every part of that statement can be tested. Another trader using the same rules should get nearly the same results. Most importantly, the hypothesis can be proven wrong. If the backtest shows only a 45 percent win rate with a 1:1 reward-to-risk ratio, you have discovered that the idea doesn't provide a profitable edge before risking real money.
That is why every trading idea should begin as a testable hypothesis, not a belief. A hypothesis tells you exactly what to measure and gives the market a chance to prove you wrong through live price action, backtesting, or historical chart studies. If a signal survives repeated testing across different stocks and market environments, you may have found a genuine edge. If it doesn't, you've saved yourself from trading a strategy that only sounded good in theory.
Measuring the Edge
Once you have a few hundred rows of data, one number matters more than anything else on your screen: the average result per trade in R units. Traders call it expectancy.
Expectancy is the average amount you can expect to make or lose on each trade over a large sample.
The formula is:
Expectancy = (Win Rate × Average Win in R) − (Loss Rate × Average Loss in R)
Suppose you analyze 100 trades and find a 40 percent win rate. Your average winning trade is 2R, and your average losing trade is 1R.
The math is simple:
(0.40 × 2) − (0.60 × 1) = 0.80 − 0.60 = +0.20R
Your expectancy is +0.20R per trade. If you risk $100 on every trade, your expected profit is $20 per trade over the long run. After 500 trades, your statistical expectation is about $10,000 in profits before commissions and slippage.
Now compare that to another system with a 70 percent win rate, an average winner of 1R, and an average loser of 3R.
(0.70 × 1) − (0.30 × 3) = 0.70 − 0.90 = -0.20R
Even though this trader wins seven out of every ten trades, the system loses an average of 0.20R per trade because the losses are much larger than the gains.
This is one of the most important lessons in trading. A high win rate does not guarantee profitability. What matters is expectancy. It combines both your win rate and your reward-to-risk ratio into a single number that tells you whether your trading strategy has a positive edge. Almost no trading strategy will be profitable long-term if it has big losses.
Deciding How Much Data Is Enough
One of the biggest mistakes a trader makes is believing a small sample of trades proves they have found an edge. It doesn't. Small samples are heavily influenced by luck.
Imagine you test a new trading strategy for just 20 trades and it wins 14 of them. A 70 percent win rate sounds impressive, but after only 20 trades, it could simply be a lucky streak. The next 20 trades might win only 8, dropping the overall win rate to 55 percent.
Now suppose you test the same strategy over 200 trades and it still wins about 60 percent of the time. That result is much more believable because random luck has much less influence over a large sample.
The larger your sample size, the more confidence you can have that your results reflect the actual performance of the strategy instead of chance.
As a practical guideline:
Fewer than 30 trades tell you very little.
Around 100 trades give you a reasonable first estimate.
Around 200 trades provide much greater confidence.
Three hundred or more trades give you a much more reliable picture of whether a trading edge actually exists.
Professional traders understand that every new trading system is an experiment. They begin by risking small amounts of capital while collecting enough data to determine whether the strategy has a genuine positive expectancy in real trading. Only after the evidence consistently supports the strategy do they consider increasing position size.
The goal isn't to prove yourself right quickly. The goal is to gather enough evidence that you can trust the results. In trading, confidence should come from data, not from a handful of winning trades.
Position Size Is the Price of Uncertainty
Finding a profitable trading strategy is only half the job. The other half is deciding how much money to risk on each trade.
Many traders think they should risk more money once they find a strategy with a positive expectancy. The problem is that no backtest or historical study can tell you exactly how the strategy will perform in the future. Every estimate contains uncertainty.
That's why professional traders risk only a small percentage of their account on any single trade.
A common guideline is to risk between 1% and 2% of your account on each trade. This keeps a normal losing streak from causing serious damage while giving your edge enough trades to play out.
For example, suppose you have a $100,000 trading account and decide to risk 1 percent per trade. Your maximum loss is $1,000.
If your entry is $50 and your stop loss is $45, you are risking $5 per share. Divide your maximum dollar risk by your risk per share:
$1,000 ÷ $5 = 200 shares
Now suppose the same trade requires a wider stop at $40 instead of $45. Your risk is now $10 per share.
$1,000 ÷ $10 = 100 shares
Notice what changed. The stop-loss widened to twice its size, so your position size was cut in half. Yet your maximum loss remains $1,000.
This is how professional traders think about position sizing. They don't decide how many shares to buy first. They decide how much money they will lose if the trade is wrong, then calculate the correct position size based on that.
By keeping your dollar risk consistent from trade to trade, you give your trading edge the time it needs to work while protecting your account from the inevitable losing streaks that every profitable system experiences.
Change One Thing at a Time
Amateur traders often change several parts of their strategy at the same time after a losing streak. They might adjust their entry signal, move their stop loss, change their profit target, and increase or decrease their position size all at once. If the results improve, they have no way of knowing which change actually made the difference. If the results get worse, they don't know what caused that either. Good experiments change only one variable at a time.
For example, suppose you want to know whether a trailing stop performs better than a fixed profit target. Keep everything else exactly the same. Use the same 300 historical trades, the same entry rules, the same stop losses, and the same position sizing. The only difference is the exit method. Your original strategy produced an expectancy of 0.20R per trade. The trailing-stop version yields an expectancy of 0.30R per trade. Because only one rule changed, you can confidently conclude that the new exit method improved the strategy by 0.10R per trade. That improvement is measurable, repeatable, and gives you a solid foundation for testing the next idea.
Define System Failure in Advance
The habit that separates a scientific trader from an optimistic one is written into the trading plan before the first live order goes in, and it defines the conditions under which the system is declared invalid.
Something like this works: if the realized expectancy across any rolling 100 trades turns negative, or if drawdown exceeds 1.5 times the worst drawdown observed in testing, trading stops and the system goes back for review. A backtested maximum drawdown of 10R gives you a live shutdown level of 15R. Written in advance, dated, and treated as non-negotiable when the moment arrives.
That single paragraph in your plan defends you against the most expensive behavior in this business, which is continuing to trade a failed strategy idea with actual capital because quitting would feel like an admission that your research was wrong. A researcher whose hypothesis fails has not failed at research. They have removed one wrong path from their journey and moved on with their trading capital still intact.
What Intuition Is Actually For
None of this makes your instincts worthless. Instinct is a superb generator of hypotheses but usually a disastrous primary decision maker, and the mistake most traders make is assigning it the second job. It takes many years of market and trading experience to develop intuition, and that usually comes from seeing patterns in the markets repeatedly until it trains your subconscious. Everyone must use quantified data even after reaching that level.
For veteran traders, when something on a chart makes the hairs on your arms stand up, that reaction stems from thousands of hours of pattern exposure and often points to something real. Honor it by writing it down. Screenshot the chart, date it, describe the condition in numbers you could code, and drop it into a file of untested ideas. Then keep trading your tested rules. Once a month, pull three items out of that file and put them through the same process you just read about.
Most of them will probably fail. One might add to your R expectancy, which can add up. The errors are in missing your trade management ideas so you can't research them, either by ignoring the instinct or, far worse, betting on it live with no research on their viability.
Build the Scientific Method Into Your Trading Plan
Build a structured trade log with fixed columns (date, ticker, setup name, entry, initial stop loss, trailing stop, profit target, share count, R multiple, relative volume, sector, entry time, etc.) and complete every column on every trade without exception. Gather the data you will need to analyze.
Convert each trading idea into a written hypothesis containing a defined entry condition, a defined exit condition, a defined measurement window, and a defined baseline it has to beat before it earns capital.
Calculate monthly expectancy as (Win Rate x Avg Win in R) minus (Loss Rate x Avg Loss in R), and track it separately for each distinct setup and for the account.
Refuse to draw conclusions before 30 trades; treat 100 to 200 as the confidence threshold, and always state your win rate as a range using the square root of p(1 minus p) divided by n rather than as a single number.
Compute t = (Mean R divided by Std Dev of R) x the square root of n at every review, and hold risk at or below 1 percent of equity per trade until t clears 2.0.
Set position size from the stop distance and a fixed fractional risk of 1 percent to 2 percent of equity, and calculate share count as risk dollars divided by stop distance per share.
Test exactly one rule change per cycle against an unchanged trade list, and record the expectancy difference in R so every improvement has a documented cause.
Write criteria that invalidate a strategy before going live (negative expectancy across a rolling 100 trades, or drawdown beyond 1.5 times the worst tested drawdown) and treat the trigger as automatic.
Keep a dated file of untested instincts, with screenshots and numeric conditions, and promote three of them to formal testing on a fixed monthly schedule, rather than acting on them in real time.
If you liked this chapter sample, you can get a copy of my book on Amazon worldwide: How to Trade Like a Scientist: Use the Principles of Science to Build, Test, and Continuously Improve a Profitable Trading System
#WhatWillDriveCryptoandStocks’NextRebound? #USPPIMissMayEaseBTCPressure #TradFiForBeginners #BitcoinBreaksBelow200WeekMovingAverage #BitcoinCoreDisclosesTorTrafficVulnerability