Welcome to NexusFi: the best trading community on the planet, with over 150,000 members Sign Up Now for Free
Genuine reviews from real traders, not fake reviews from stealth vendors
Quality education from leading professional traders
We are a friendly, helpful, and positive community
We do not tolerate rude behavior, trolling, or vendors advertising in posts
We are here to help, just let us know what you need
You'll need to register in order to view the content of the threads and start contributing to our community. It's free for basic access, or support us by becoming an Elite Member -- see if you qualify for a discount below.
-- Big Mike, Site Administrator
(If you already have an account, login at the top of the page)
I have developed a trading system that did very well from 2007-2011. total of 700 trades taken. however, I feel that this does not tell me the whole story. I feel that I need at least then years back data for my trading system to see if it really works. However, because its an intra day strategy, I only got 4 years back data. Is there a way to tell if this strategy will work in the future or was this just luck.
Can you help answer these questions from other members on NexusFi?
The only way you can tell if it will work in the future is to trade it in the future. But, since that's a fairly useless answer, there are characteristics which give us clues:
1. Profit factor
2. Sharpe ratio
3. SQN
4. Percent wins
5. Expectancy
6. Max drawdown
7. Average trade, winner, looser
8. Is the equity curve smooth or jagged?
All of the above have been thoroughly discussed here on Big Mike's. There is one more I like to see which is not discussed much. If the system takes both longs and shorts, then:
9. Is long and short profitability fairly evenly balanced?
For (9) I use this formula, 1 is a perfect score:
min(long net, short net) / (long net + short net) / 0.5
Forward testing your strategy through enough weeks to trade in at least one each of a trending , ranging and congested market is the only place to start . Backtests are fraught with inconsistencies that could take more (valuable) time to unravel than to forward test in a live market on a simulator .
I and most others here have had great tests that flopped hard in the live market . Dunno exactly why but its true , so invest your time in testing it in hand to hand combat long enough to shake out the bugs .
If you develop a trading system it is always biased towards past data, and there is no guarantee that it will work in future.
To avoid curve fitting, you should only use a part of the past data to select and optimize the parameters. You can then use the remaining data for a walk forward analysis and Monte Carlo simulation. This will at least tell you whether your system is curve fitted or not. It does not tell you of course, whether the market conditions in the future changes. Is this what you did? Then it should give you some comfort.
2007-2011 is not a bad base for testing a system, because it includes both a bull and a bear market, so it should not be biased towards long or short trades.
If you have good back tested results, turn the system on live and watch the system trade. Is it following your rules, on entries and exits?
When you're comfortable take it live... with a finger on the quick close in the event your stop criteria doesn't execute as planned.
Once you gotten through that stage set it and forget it... you might also choose a broker that will allow for a max daily loss restriction for double fail safe.
My question to you is how many instruments does the system work on?
Some version of this is a frequent theme here. I think you could be saying any combination of 3 things:
1. Historical data feeds don't match real-time data feeds.
This is true, but they are certainly very highly correlated
2. The past is not predictive of the future
This may be true, but if it had no predictive value at all then no trading system would work
3. The systems you've tried execute differently over historical data vs real-time data
This is true for NT by default, but it can be largely overcome
So... I find backtesting to be highly correlated with actual results. In fact, for each system I trade, I check daily after the close that the backtesting engine gives the same performance I experienced in real-time. If there is any difference at all which is not explainable by slippage, then I treat it as a programming error and seek to correct it.