What is decent return?

[wantochangemylife]

I have been testing a sytem for predicting ftse closing up or down using snapshot data.

I have back tested this theory on every day for 3 years worth of trading.

I started with £500 and placed bets at £100 per trade.

Year 1 - proft £3600
Year 2 - profit £1100
Year 3 profit £1700

It works BUT it involves me making a trade everyday and whilst it clearly works is this a viable return for the effort? C

THanks

 

[Shakone]

"it clearly works" - well I'm not sure about that. It is a bit hard to say whether it works. Have you FORWARD-tested it?

Also, if you're risking £100 per trade on a £500 account, there is a decent chance you will lose it all if you get some bad luck in the beginning.

As for is it viable for the effort...well it seems like no effort at all. Shoudl just take you less than a minute a day to make that bet.

 

[wantochangemylife]

You are correct - only done 3 years worth of backtesting. As I am a bit of a novice could you explain how/what I need to do to forward test it. Thank you in advance for your help.

 

[Shanghai]

How much are you risking per trade? I know your trades are £100 but what is the maximum amount you can lose per trade if the trade does not work?

Its also important to know what the maximum drawdown was during your backtesting.

 

[Shakone]

Forward testing: Make your predictions for tomorrow, and the next day, and the next for at least a month (preferably more because you are only getting one trade per day), and see what the return would be. You don't need to risk money doing it, just make your prediction and see how it would have worked out.

Sometimes a backtested system which seemed successful fails completely when you go into forward testing.

Shanghai, I was under the impression that all he can lose is the 100, that it is a bet either close up or close down, and how much up or down is not relevant.

 

[Xeno]

Forward testing just means carry on with the exact same system and do it on data going forward. If you've tested up to now, that'll be real time data.

The first problem that occurs to me about your strategy is this.

In year 2, you have a win percentage of about 52-53% At that win rate there's a good chance you'll have five losers in a row in the first quarter of the year alone. So, you could be pretty likely to go bust at some point.

 

[Shanghai]

Shanghai, I was under the impression that all he can lose is the 100, that it is a bet either close up or close down, and how much up or down is not relevant.

Could be. If that is the case then that's much too much risk for a £500 account.

 

[wantochangemylife]

Thanks for the replies everyone.

I trade £100 per trade - so I either double my money or loss the £100 - the maximum loss I had over the 3 years testing was £600 (6 bets in a row). I now understanding what forward testing means, thanks. I start for real this afternoon - stuck £500 in the account and will trade £50 bets so this should be ok (unless I get 10 wrong lol). Keep you all posted.

 

[Shakone]

Good luck with it. Don't underestimate the getting 10 wrong part. Getting 10 wrong in a row is unlikely, but 45 right and 55 wrong, or even 10 right and 20 wrong are not so unlikely.

 

[Windowsill]

Thanks for the replies everyone.

I trade £100 per trade - so I either double my money or loss the £100 - the maximum loss I had over the 3 years testing was £600 (6 bets in a row). I now understanding what forward testing means, thanks. I start for real this afternoon - stuck £500 in the account and will trade £50 bets so this should be ok (unless I get 10 wrong lol). Keep you all posted.

The return is very poor, 6 days going the other way would wipe out the profit.

 

[meanreversion]

Forward testing is not as has been described here!

Imagine you have 10 years of data.. optimise the system using data for year 1+2 ("in sample") then see how that optimised system would have fared in year 3 (out of sample). Then repeat the process using year 2+3, and see how the optimised system would have fared in year 4, and so on.

You can choose whichever output you like to optimise (e.g. CAGR/max drawdown). You can then see if your system is robust... for example, if the in-sample return is 50 pct a year, but the out-of-sample is +5 pct, it might suggest curve fitting or a non-robust system.

Forward testing is fairly advanced and I don't know many people who do it. It's pretty easy to do on Amibroker, and it gives you an added layer of comfort (or not!) for your mechanical system.

 

[Shakone]

meanreversion, that still sounds like back-testing, albeit with some sensible tests. Forward testing is all that really matter though. And you can't say that you don't know many people who do it, since everyone who trades for real, IS forward testing

 

[LiamH]

Am I the only one who thinks this strategy doesn't make any sense?

We're talking about 20% risk per trade yet the profits in relative terms are tiny over a year. The op also states that he either doubles his stake or loses the £100 so the risk:reward is 1:1? Surely this then means that pct winners\losers is not far from 50\50.

If this is the case, risking 20% per trade then there is as close to a cast iron guarantee as you can get that the account will be blown within 100 trades.

 

[wantochangemylife]

Thanks all for the posts certainly food for thought.

Over 3 years of back testing I only had 6 losses once. The lowest the cash balance went to was -£300.

The first year profit was good but I noted this took place in a falling market , the other 2 years in a risining. I think I need to fine tune it as it really is simple - perhaps to simple?

Whilst here anyone think the Ftse will end lower this year?

Once again thanks for your input

 

[wistundra]

Wonder how long it would take someone at a brokerage to pick up on an account that was consistantly making 25/40% a month on margin and start duplicating the trades for the house account.
Anyone care to comment ?

 

[Hoggums]

Test it on the period 2000-2004.

 

[meanreversion]

meanreversion, that still sounds like back-testing, albeit with some sensible tests. Forward testing is all that really matter though. And you can't say that you don't know many people who do it, since everyone who trades for real, IS forward testing

Take a look at this link, it describes the walk-forward test, or what I believe is forward testing.

http://www.amibroker.com/guide/h_walkforward.html

Maybe this is just semantics, I don't know, but to me it makes a lot more sense that THIS process is "forward testing" rather than trading for real = forward testing. If forward testing is simply trading, why give it such a grandiose title? This website could be renamed "Forward Test 2 Win" --- seems unlikely.

I suspect most people hear the term "forward test" and just ASSUME it means putting real money on the line; I think this is erroneous.

 

[wantochangemylife]

Test it on the period 2000-2004.

Tested years 2003/2004/2005

 

[Shanghai]

Take a look at this link, it describes the walk-forward test, or what I believe is forward testing.

http://www.amibroker.com/guide/h_walkforward.html

Maybe this is just semantics, I don't know, but to me it makes a lot more sense that THIS process is "forward testing" rather than trading for real = forward testing. If forward testing is simply trading, why give it such a grandiose title? This website could be renamed "Forward Test 2 Win" --- seems unlikely.

I suspect most people hear the term "forward test" and just ASSUME it means putting real money on the line; I think this is erroneous.

My understanding is the same as yours. I don't consider live trading as forward testing.

 

[Xeno]

Take a look at this link, it describes the walk-forward test, or what I believe is forward testing.

http://www.amibroker.com/guide/h_walkforward.html

Maybe this is just semantics, I don't know, but to me it makes a lot more sense that THIS process is "forward testing" rather than trading for real = forward testing. If forward testing is simply trading, why give it such a grandiose title? This website could be renamed "Forward Test 2 Win" --- seems unlikely.

I suspect most people hear the term "forward test" and just ASSUME it means putting real money on the line; I think this is erroneous.

I agree it's probably semantics, and I wouldn't want people to get the wrong impression about trading/testing, but I do think trading can be forward testing (note, they're not the same act - they are two different objectives which can be accomplished with the same starting actions)

My take is this. Forward testing is all testing on out of sample data. Out of sample data implies that your method must have already been tested (and probably optimised) on in sample data. Simple as that. Makes no difference whether the data is in the past or the future. Therefore trading can be forward testing if you do test. Whether you also risk money is irrelevant.