The terms used across this site, each with where to see it in real data.
Backtest
A simulation that applies fixed trading rules to past prices to see how the rules would have behaved. It is evidence about past behaviour, not a forecast. See strategy backtests
Virtual trading
A record built going forward, trading on live prices with simulated fills and no real orders. Because the rules are fixed before it starts, it covers time the backtest never saw. See the track record
CAGR (compound annual growth rate)
The yearly rate that compounds the starting amount into the ending amount: (end ÷ start)^(1 ÷ years) − 1. Annualising a few weeks of record inflates it badly, so it is not used for short records. CAGR in real backtests
Max drawdown (MDD)
The largest fall of the equity curve from a previous peak to a later low. Losses and recoveries are not symmetric (a 50% drawdown needs a 100% gain), so read it before the return. More on max drawdown
Sharpe ratio
Average return divided by its volatility (standard deviation), annualised: how much was earned per unit of wobble. Strategy backtests on this site do not subtract a risk-free rate. Sharpe in real backtests
Calmar ratio
CAGR divided by the absolute max drawdown: how many percent a year were earned per percent of drawdown endured. Calmar in real backtests
Volatility
The standard deviation of returns: how widely they scatter around the average. Daily volatility is annualised by multiplying by the square root of trading days per year.
Benchmark
What a strategy is compared with. This site uses buying and holding SPY (the US S&P 500 ETF) over the same dates. A lower return can still matter if the drawdown is clearly smaller and steadier, so both are shown.
Rebalancing
Resetting asset weights that have drifted with prices back to their targets. Results depend on the schedule (monthly, yearly) and the trading costs.
Trading costs and slippage
Fees, bid-ask spread and the gap between the intended and the filled price (slippage). The more a strategy trades, the wider the gap between backtest and real account. How costs affect results
Survivorship bias
The illusion created by testing the past only with assets that survived to today. Dropping what was delisted in between removes losses. Test with every candidate that was tradable at the time. Check it in the checklist
Look-ahead bias
A backtest using information that was not available at the moment of decision. The classic case: deciding on today's close and assuming a fill at that same close. Fills must come after the decision. Check it in the checklist
Overfitting
Trying many parameter values and keeping the one that fitted the past best memorises past luck rather than a rule. Re-check on data not used for the choice and on a forward record. Check it in the checklist
Out-of-sample period
Data not looked at when the rules or parameters were chosen. The famous method experiments show the last 30% of each period separately as out-of-sample. See the experiments
Walk-forward / forward test
Checking fixed rules on new data that arrives afterwards. The virtual trading record on this site is one; nothing is ranked before 28 days. See virtual trading rankings
Leveraged ETF decay
2x and 3x ETFs reset their leverage daily, so in choppy markets they lose value even when the index ends flat. Backtests on real market prices include this decay.
Momentum
Using the tendency of recent winners to keep rising for a while. Picking what rose more than other assets is relative momentum; checking whether it beat cash is absolute momentum. See the dual momentum experiment
Trend following
Holding only while prices trend up (for example above a moving average) and stepping aside into cash or bonds when the trend breaks. It cuts deep falls but pays for frequent trades in sideways markets. See the golden cross experiment
Educational material, not investment advice or a trading signal. Past results do not guarantee future returns.