That's a point I think we can agree on, in all its various manifestations. Honestly, I would be happy for OP if I could, but until he has forward tested this live there's very little reason for him to engage with any sort of excitement.
NQ Fully Automated Trading System
- Thread starter Rickshaw Man
- Start date
That's a point I think we can agree on, in all its various manifestations. Honestly, I would be happy for OP if I could, but until he has forward tested this live there's very little reason for him to engage with any sort of excitement.
Only observation without seeing more is the results do look too good to be true.
Either 10 month sample size is too short. Or some error in the back test assumptions.
But maybe the OP has found the holy grail trading money printer!
Either 10 month sample size is too short. Or some error in the back test assumptions.
But maybe the OP has found the holy grail trading money printer!
Day trading is a bit different to long term trading you dont need decades of data, he has a 1500 trade sample size, which i think is 10 months worth for his system.
A 5000 trade sample would be better, although ideally he should be testing his system under different market conditions, eg bull/bear/low volatility/high volatility.
Actually it depends on the profitability of the system; basically it's a function of the square root of the number of data points (trades) and the average return/risk of those trades. The more profitable the system, the less data we need to be confident. So it's only true that more trades automatically means more confidence if the profitability of the system also goes up proportionally as root(T) reduces. Eyeballing that system, I'd say it has a sharpe ratio of 1, which means it's about as good as a system with one month holding period and a SR of 1/root(20) = 0.22. You'd need decades of data to be statistically confident about a SR 0.22 system with monthly holding period. Hence my comment.
(This assumes there is zero overfitting in the backtest, which with just 10 months worth of data seems a little unlikely).
GAT
The sharpe ratio looks around 5.0 to me
. The table in the opening post says the monthly standard deviation is very low vs the monthly return. Another pointer to it being fitted to a small sample size or single market type.The sharpe ratio looks around 5.0 to me
I was eyeballing on daily returns, but you're right the monthly SR is 5.5 (assuming zero risk free). Which is the equivalent of about SR 1.23; which by coincide is roughly what my own backtested system has and I have ~50 years of data and ~114 different markets.
Anyway the point still stands; running a system - even a day trading system - tested on less than a year of data for a single market is absolutely nuts.
GAT
I've always felt the number of trades over the amount of time was critical to system success. I will keep monitoring it. No big hurry. I will say, it underperforms in a range-bound environment.
Hi Rickshaw Man,
The system looks good. Good luck with forward trading it!
Can you say more about why you are using such a high slippage figure?
Last edited: