Evidence-based trading

What you will learn

The numbers nobody shows you

The European regulator (ESMA) forces brokers to publish it: between 74% and 89% of retail CFD accounts LOSE money. And the largest day-trading study (Brazilian futures, everyone persisting 300+ sessions in 2013-2015): 97% lost, and only 0.4% earned more than a bank teller. This isn't to discourage you: it's your real starting point. Whoever knows the base rates plays a different game.

The law of large numbers (the casino model)

The casino doesn't know the next spin, yet it ALWAYS wins by year-end: it has a small edge and repeats it millions of times. That's the law of large numbers: with repetitions, your actual result converges to your expectancy. Translated to trading: stop thinking about THE trade and think about the NEXT 100. Your job isn't to guess; it's to execute a small edge many times, like an insurer or a market maker.

Expectancy: your only compass

Expectancy = (win% × average win) − (loss% × average loss). If it's positive and you repeat it enough times, you win; if it's negative, NO psychology saves you. That's why you can profit winning only 40% of the time (if your wins are twice your losses) or go broke winning 70%. Professionals don't brag about win rates: they measure expectancy per trade and track it across hundreds of logged trades — exactly what the journal is for.

The index beats the professionals (SPIVA)

S&P measures professional managers against their index every year (the SPIVA report). The 15-year result: 89.5% of US large-cap equity funds lose to the S&P 500, and NOT ONE of the 22 categories analyzed has a majority that wins. The cause is mathematical: fees and costs compound against you. Double lesson: buying the index cheaply for the long run is the most proven strategy in existence, and beating the market is so hard it demands method, not opinions.

The rewarded factors: value, momentum, quality, low volatility

Academic research (Fama and French, Carhart) and quant managers like AQR have documented 'factor premia' that persist across decades and dozens of countries: VALUE (buying cheap relative to fundamentals), MOMENTUM (what's rising tends to keep rising), QUALITY (profitable, solid companies) and LOW VOLATILITY. They don't work every year — they go through long bad stretches — but they're among the few things that have survived serious statistics. Much of institutional 'smart beta' is exactly this.

Momentum: the most robust anomaly

Jegadeesh and Titman (1993) measured it: the stocks that rose most over the past 3-12 months kept beating the ones that fell most, with close to 1% monthly excess return in 1965-1989. Since then it has been confirmed in 30 years of out-of-sample data, in some 40 countries and in more than a dozen asset classes. It's the statistical validation of 'trade with the trend'. Mind its Achilles heel: sharp market reversals ('momentum crashes').

Trend following: a century of evidence

AQR reconstructed trend following back to 1880 across 67 markets and 4 asset classes ('A Century of Evidence on Trend-Following'): positive returns in EVERY decade from 1880 to 2016, through wars and peace, inflation and deflation. And it did well in 8 of the 10 largest crises of a 60/40 portfolio. It's what banks' and managers' 'managed futures' funds do: buy what rises, sell what falls, with rules and stops — no guessing tops and bottoms.

Size is (almost) everything: Kelly and fixed risk

Kelly's math (1956), brought to markets by Ed Thorp, proves something brutal: WITH the same edge, betting too much per trade still leads to ruin — variance kills you before the law of large numbers can save you. That's why professionals risk a small, FIXED fraction of capital (the famous ~1% per trade is a conservative version of this) and never full Kelly. Survival is the precondition of any statistics.

How professional desks manage risk

At a bank or a fund, discipline isn't optional: it's an imposed LIMIT. A daily loss limit (cross it and your trading is cut off), per-position and per-strategy limits, VaR limits, and diversification across weakly correlated strategies. The trader doesn't decide how much to risk: the risk department does. Copy the system in miniature: set your maximum daily and weekly loss BEFORE trading, and respect it as if your boss imposed it — because statistically it's what keeps you in the game.

Your biases, measured in money

Barber and Odean analyzed 66,465 real accounts (1991-1996): those who traded MOST earned 11.4% a year versus the market's 17.9% — overtrading cost ~6.5 points a year, almost all in costs. And Odean measured the disposition effect: we sell winners too early and hold losers, and the sold winners kept rising more than the held losers. Two data-backed conclusions: trade LESS, and cut your losses / let your winners run isn't a slogan — it's statistics.