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)
For those of you who use a 1 minute ES day trading algorithm for scalping (8 ticks per trade in my case), how many days of data would you use to back test it? For such a short time frame, I'm not sure it is necessary, or even pertinent, to back test beyond a few weeks. Your thoughts?
Can you help answer these questions from other members on NexusFi?
I like minimum 1 year, I've little itch to check larger sample size. Market change their behavior more than we think and its always better to check your strategy no matter the time frame on largest possible data range.
It really shouldn't be that much of trouble to stretch for at least that much.
Thanks for your reply. I have wondered about about data integrity when using futures that many days ago, especially with 1-minute charts, since trading on the ES gets pretty thin when you go past the last roll over date. I use NinjaTrader, and I have it set to merge back data, but I am not yet confident that it provides accurate data much past the last roll dates for a 1-minute chart.
Plus I am testing a strategy for choppy markets, so market conditions are important. To truly back test this strategy, I need to filter out trend days.
I don't use Ninja so I can't really comment, but here is little suggestion for future use, it is better to buy past tick data for what you trade or plan to trade so you can do back testing at your end. Lot of people do that, especially if they are coders. Plus, you get advantage of using what you know how to code or are comfortable with.
Generally, there aren't any great issues in terms of data integrity on popular platforms, because there are quite a few professional people doing back tests on very large sets of data on as small as tick level, let alone one minute. So you should be fine. If you still have doubts about integrity, you can adjust for some errors in you final results. I'm sure you wont need it though.
Thanks for the input, Bobb. I have read up to 1,000 trades. But since this is a 1-minute intra-day strategy intended for sideways markets, I am thinking that to apply it longer-term in a trending market would under-estimate the effectiveness of the strategy.
Thanks for the replies. Let me give a little more information regarding the background of my question.
This is an auto strategy which would have produced 202 trades in November, and produced $4987.50 in profits on 1 contract with a 1-contract scale in at -16 ticks. The max drawdown was -$1,312.50. December to date is $3850 on 130 trades with a max drawdown of -$1600. To me, these are good results.
But it is a 1-minute 8-tick scalping strategy for sideways markets, so it should be applied to sideways markets. November and December were sufficiently sideways to product these results, but a 6-month backtest only produces $8862.50 with a Max drawdown of $8187.50. Obviously not desirable, but there were many big, trending days during this time period for which the strategy was not designed.
So the premise of my question is this: if I am back testing a strategy which should only be profitable in sideways markets, shouldn't I back test it in sideways markets?
How does your strategy know WHEN it is a sideways market? Saying you should not run it last Sept and Oct because those months were trending is one thing, but knowing that in real time is another.
If your strategy can't tell when it is sideways, but can only exploit sideways, your 2 month backtest will not really help. It just confirms what you know - in sideways markets, strategy works great.
My only question is simply, can you tell in advance that you are going to have a sideways market? Not looking back in time and saying, "Oh, this was a trending market, I can't use the strategy for that," but looking forward in real time, and say, "Oh, this will be a trending market, I can't use the strategy now."
Or to put it another way, if you decide you will only test in periods when, as you look back at them, you can see that the market wasn't trending, you are seriously curve-fitting your test -- you are selecting the data you will test it for, because you know what the outcome was.
That will make your test essentially not worth anything. And that makes your question kind of irrelevant also, since you're only going to test during times when you know it will do well. Why test at all then?
By the way, I understand the dilemma because I have it too. I know that certain things work well during trends but not during sideways non-trends, and vice versa. There's a sort of Catch-22 there, because you can't use the method that works well in non-trending markets profitably unless you already know they will be non-trending.
But you do need a way to tell the difference, ahead of time, and then you need to incorporate whatever criteria you have developed to answer the question into the test itself, so you can test it realistically under the same type of conditions you will face with it when going live.
I hope this makes sense and is useful to you. Good luck with the strategy and your testing. And be sure you can answer this question.
Bob.
When one door closes, another opens.
-- Cervantes, Don Quixote
I know that certain things work well during trends but not during sideways non-trends, and vice versa. There's a sort of Catch-22 there, because you can't use the method that works well in non-trending markets profitably unless you already know they will be non-trending.