see. losers made the same mistakes again and again even the same thing been said again and again and again just like that "enough data" used all all market conditions.Who told you that??
Say that again??
A good backtest will contain enough data to test the system under bear, bull and flat market condition.
Best Back Testing Results
- Thread starter Satan's Helper
- Start date
Oh I see.
So from your brilliant statement we must conclude that winners......
1: Do not backtest anything, they just use their magical sixth sense to generate extremely accurate and highly profitable trading signals.
2: Only need 2 days of real-life trading (or less in most cases) to prove to themselves and to the world that their trading system works and is indeed profitable without a shadow of a doubt.
The way i back test is using the S&P 500 as a benchmark and try to outperform that in index points, not in USD. Backtesting and simulating in USD is not completely 'fair' in my opinion if the trading tool is some sort leveraged product (futures, options etc). In other words, benchmark and trading tool should have the same sort of risk/volatility.
Not at all, you only need to use the ROI (return on investment) in percentage, not the dollars and cents, or even the points or pips approach.
That way you can still compare apples with apples.
But don't you think that comparing two results with different kind of risks is fair? Making 3% in fixed income or 3% with trading futures in my opinion is not the same.
Also, another reason I'm comparing with a benchmark like the S&P500 is to see if it would be better to simply trade an always-long (leveraged) etf on that index.
Good back test = Trades > 10,000
No.
I have a system with 100 trades....has been working in forward testing exactly like backtests for 1.5 years now
I have a system with 1 trade in the backtest and it's working well:
Buy if close > 0
Backtested in SPY from 01/2010 to 12/2012 and has made good money since. Does this say anything whether the backtest is good or not?
Nice counterargument you pull off there
However, I am seriously saying that to develop good systems, you don't need 5000 or 10000 trades.
Imagine a weekly system on SPY. A year has 52 weeks, even if you backtest for 10 years, and you trade every week, at max you would have 520 data points. More realistically, with a weekly system, your holding period would be 2-3 weeks. So, assuming 2 weeks holding period on average, at most you would have 260 data points.
Will you say that this system's backtest results are suspicious and might not wok in future just because it has 260 trades in its backtest? In this context, I wanted to point that that having a requirement of 10k trades to have a proper backtest is an erroneous requirement.
A 100 trade backtest is also valid in many cases.