Market Making & Trading
P&L, Sharpe Ratio & Risk
Overview
P&L analysis and risk metrics are how you evaluate whether a strategy actually works. The Sharpe ratio is the most important single number in systematic trading — it tells you how much return you get per unit of risk. The t-statistic tests whether the Sharpe is statistically significantly different from zero. In practice, quant firms require t-stat > 3 (much stricter than the academic 1.96 threshold) because they test many strategies.
How to Recognize
- →'Is this strategy statistically significant?'
- →Given returns, compute performance metrics
- →Sizing a position based on a signal's strength
- →Questions about when to stop trading a losing strategy
Step-by-Step Approach
- 1.Sharpe = (mean return − risk-free rate) / std of returns × √252 (annualized)
- 2.t-stat of Sharpe ≈ Sharpe × √N (N = number of observations)
- 3.Position size ∝ signal strength / volatility (Kelly-type sizing)
- 4.Maximum drawdown = largest peak-to-trough decline in equity curve
- 5.Strategy is 'significant' when t-stat > 3.0 (quant standard)
Key Formulas
Worked Examples
Problem
A strategy produces daily returns over 252 trading days with mean 0.08% and standard deviation 1.2%. What is the annualized Sharpe ratio? Is it statistically significant?
Solution
Daily mean = 0.08% = 0.0008. Daily std = 1.2% = 0.012.
Assume risk-free rate ≈ 0 for simplicity.
Daily Sharpe = 0.0008/0.012 = 0.0667.
Annualized Sharpe = daily Sharpe × √252 = 0.0667 × 15.87 ≈ 1.06.
Is it significant? t-stat = SR × √N = 1.06 × √252 ≈ 1.06 × 15.87 ≈ 16.8.
t-stat >> 3: very significant! But note this is for 1 year. If we only had 1 month (21 days): SR_annual ≈ 1.06, t = 1.06 × √21 ≈ 4.85. Still significant.
Answer
SR ≈ 1.06, t-stat ≈ 16.8 over 252 days: highly significant. Rule of thumb: a Sharpe of 1.0 over 1 year of daily data gives t ≈ SR·√T ≈ 1.0·15.87 ≈ 16.
Common Mistakes
- !
Not annualizing: daily Sharpe is 16× smaller than annual Sharpe (divide by √252). Always specify the period.
- !
Thinking high Sharpe in-sample guarantees future performance. Overfitting can make any strategy look great historically.
- !
Using the wrong number of observations for the t-stat. If you have monthly data (not daily), N = 12/yr, not 252.
- !
Forgetting that Sharpe doesn't capture tail risk. A strategy with perfect Sharpe can have huge drawdowns if returns aren't normal.
Practice Problems
Click "Show Answer" to revealA fund has annualized return of 15% and annualized volatility of 10%. Risk-free rate = 3%. What is the Sharpe ratio?
How many daily observations do you need to achieve t-stat = 3 with an annualized Sharpe of 0.8?