Automated trading system developer looking for a contract.

Quote from erdewit:

After comparing the excel sheet with tick data of the DAX. the exits look plausible.
They are just fixed brackets around the entry.

But the entries only look possible if they are limit orders that are placed well in advance.
Eugene, can you tell us if this is the case?
More...

Sure, each entry executes by follow way.
When found first tick price with level price (level price mean system have +1 or -1 signal) it send market order. Tried buy by market. By market execution after lelve price waiting 5 lots volume of order price or less/over price.
Why entry simulates by market? How trades going?
price1, price2 etc... My model need some price over than prev. for entry long. it mean that prev. prices is less than level and level price is current market ask (for long). I just simulate buy from ask for long and from bid for short entries.
 

Sorry I missed your post.
What is backtest (curve fitting)? When you calculate system on some period and then analyze system. You are looking better params, rules etc... In other words you are doing equity curve better on KNOWN data and can suppose how it will work for future only. My way is Walk Forward Test.
I defined rules and do a lot of tests (for example, for each trading day) each time apply calculated system on next trading day. As result I have some paper trades. When I review trades given each time as result of apply system on UNKNOW data I can see how system will work for future.


Ok, I think I understand what your mean by Walk Forward Test. So what are the limitation of "Walk Forward Test"?

1. How long you can do Walk Forward Test? Can you do 10 years or 20 years Walk Forward Test?

2. What data set you can generate to do Walk Forward Test? Using random data generator? or live data? If life data, how can you do Walk Forward Test for 10 or 20 years?
 
Ela, your sample is only 2 months.
Do have more? It is so important to test your system over a large period. Your system must be able to survive in different markets(read:crash, low volatility, high volatility). 7-10 years is a good start for a testing period. If you don't have the tickdata just buy it. It is a good investment.
You need to put in slippage! Calculate half the expected spread as slippage per trade(or something). There are always items that you have missed in your research. Like you want to buy on the ask right at the time economic figures are coming....auw.
 
Quote from 88888888:


Sorry I missed your post.
What is backtest (curve fitting)? When you calculate system on some period and then analyze system. You are looking better params, rules etc... In other words you are doing equity curve better on KNOWN data and can suppose how it will work for future only. My way is Walk Forward Test.
I defined rules and do a lot of tests (for example, for each trading day) each time apply calculated system on next trading day. As result I have some paper trades. When I review trades given each time as result of apply system on UNKNOW data I can see how system will work for future.


Ok, I think I understand what your mean by Walk Forward Test. So what are the limitation of "Walk Forward Test"?

1. How long you can do Walk Forward Test? Can you do 10 years or 20 years Walk Forward Test?

2. What data set you can generate to do Walk Forward Test? Using random data generator? or live data? If life data, how can you do Walk Forward Test for 10 or 20 years?
More...

I agree with you. 2 month is small period. But system for NQ & ES have big trades count and I can say that system is promising. FDAX really need more testing.
For long test I need a lot of ticks and good hardware (I guess 4 good workstation allow me do research work within 1 month... that is reason why i do not need 10 year tick's history now. Market sumulation not baby game). Currently I have only 5 month ticks for ES & NQ [I'll try get more ticks from TradeStation, but my TS subscription is limited]. I'll continue my research work and if you have interesting - let me your contact info via private message.
I made deep WFT for QQQQ some time ago and have positive return (attached it for review). But this is long term system based on 15MIN. Not quick return like for futures. Futures system have differences also.
 

Attachments

Quote from elabunsky:

[I'll try get more ticks from TradeStation, but my TS subscription is limited].

I would consider CQG or buying the data direct from the exchanges if you want quality data. It is worth the cost.
More...
 
Quote from 88888888:



Ok, I think I understand what your mean by Walk Forward Test. So what are the limitation of "Walk Forward Test"?

1. How long you can do Walk Forward Test? Can you do 10 years or 20 years Walk Forward Test?

2. What data set you can generate to do Walk Forward Test? Using random data generator? or live data? If life data, how can you do Walk Forward Test for 10 or 20 years?
More...

2. Uses live data only. I can do WFT for 5-15 mins time frames. But I noticed above - my target is futures now. It allow get quick return. Sure, if somebody will be interesting for stock market research and will offer me good conditions why not?
 
You must log in or register to reply here.
Back
Top