Menu
Sister Sites

Options

Bookmark Thread

First Page First Unread Last Page Last Post

Print Thread

Similar Threads

Right time frame to trade for 3-days-average trade length 0 replies

Why EA test is forward test better than back test?? 1 reply

How to define the length of an array with an extern variable? 5 replies

Question about number of streaks of a certain length in a trade sample 1 reply

Draw a certain-length line 3 replies

Length of time to test a new system

  • Post #1
  • Quote
  • First Post: Dec 20, 2010 11:42am Dec 20, 2010 11:42am
I've started testing a "new" system,i'm just wondering how long i should test it for or how many set ups i should trade before going live with it.
Mark Douglas in his book Trading in the zone ,reckons that 20 trades of your set up should give you an reasonable accurate reading,i'm thinking this is not enough,any thoughts ?
Time probably isn't relevent ,as it differs greatly the amount of trades you get with diffrent systems and t/fs
Regards........

Arrow
20 is way too little to gain any confidence with your new system. I would say 100 trades at least to gain some descriptive statistics data on how your system performs.
Measure twice, cut once
Error is proportional to the square root of N

Think about it. Thank god Douglas decided to pretend to be a psychologist rather than a statistician !
Quoting Razzle
Error is proportional to the square root of N

Think about it. Thank god Douglas decided to pretend to be a psychologist rather than a statistician !
Fair enough Razzle,saying that,Trading in the zone helped me alot with the physological side of trading.
Quoting arrowloopboy
I've started testing a "new" system,i'm just wondering how long i should test it for or how many set ups i should trade before going live with it.
Mark Douglas in his book Trading in the zone ,reckons that 20 trades of your set up should give you an reasonable accurate reading,i'm thinking this is not enough,any thoughts ?
Time probably isn't relevent ,as it differs greatly the amount of trades you get with diffrent systems and t/fs
Regards........

Arrow
Unless you are already a highly experienced a consistently profitable trader I would suggest that your aim should be to establish the capability of your method rather than just test a new system.

The testing of a new "system" tends to exclude so many of the variables that are critical to sustainable trading, not least of which the establishment of the knowldge, skills and experience of the trader and the proper evaluation of all the assumptions that reside in the new plan/system. Testing a system, merely on outcomes for say, 20 trades, (over what period of time and in what diversity of market conditions?) often wont tell you much, or at worst may build false and untested assumptions into your trading plans that may end up loosing you money.. The "I tested the system for 3 months but then it stopped working as soon as I started trading live" scenario is often evidence of this.
Quoting arrowloopboy
Mark Douglas in his book Trading in the zone ,reckons that 20 trades of your set up should give you an reasonable accurate reading,i'm thinking this is not enough,any thoughts ?

Time probably isn't relevent ,as it differs greatly the amount of trades you get with diffrent systems and t/fs
Regards........

Arrow
hey arrow..... some of the metaquotes bunch and along with a couple of their russian friends suggest far more than 20.... they point to the hundreds...... but the type system would play a role in that decision also....

my thoughts would be somewhere between thousands to get an accurate idea and an eternity to improve it.....

it took me a full year and probably more than 20,000 live trades to nail down one..... nfa rule changes had a little bit to due with that delay....

my eternity to improve it started 10/01/10 ..... it has placed 2499 trades since..... improvements come to mind every day......h
to trade and code, keep both simple... no call to impress....h
I would backtest it until I think I have most of the bugs worked out, and forward test it on a demo account for at least two weeks, on perhaps several charts at once (especially with a basket system), to make sure there are no surprises. This way you will be sure to get a lot of trades in and feel fairly confident about what you're doing.

Of course this doesn't guarantee you'll be successful but, you can always start on a micro or mini account, to minimize your losses.
Quoting arrowloopboy
I've started testing a "new" system,i'm just wondering how long i should test it for or how many set ups i should trade before going live with it.
Mark Douglas in his book Trading in the zone ,reckons that 20 trades of your set up should give you an reasonable accurate reading,i'm thinking this is not enough,any thoughts ?
Time probably isn't relevent ,as it differs greatly the amount of trades you get with diffrent systems and t/fs
Regards........

Arrow
What 20??? That's supposed to be way too small of a sample.

In terms of history, we are talking min. 10 Years, more is better. If your system doesn't generate enough trades per 10 Years, test on more history. If still, test it on multiple instruments and aggregate the results.

Typically a good sample would not be less than 200 trades, but 500+ is recommended.

It's no problem to test a system on multiple instruments and aggregate the results if you don't have enough history or if your system is generating few trades. A good system is supposed to work on multiple instruments with slightly different results.


Regards,

Nader
  • Post #9
  • Quote
  • Last Post: Jan 7, 2011 4:16pm Jan 7, 2011 4:16pm
I try to have a minimum of 10 years with backtesting with plenty of occurrences to make it worthwhile. Then I test on a limited scale of the last 18 months to make sure the system hasn't been a loser for 18 months straight. This will give you both the larger and smaller picture of what you should expect from your trading system.
0 traders viewing now