How many trades until a system is proven, not just lucky? | Page 2 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Please code this proven system (alien's system v2) 10 replies

where is the "simple proven system" thread gone? 3 replies

Looking for a proven mechanical system 4 replies

Am I good or just lucky? 31 replies

SWING Trades using the FXTradepro "LUCKY 7" Progression Sequence 4 replies

How many trades until a system is proven, not just lucky?

  • #21
  • Jan 8, 2014 12:04am Jan 8, 2014 12:04am
Quoting grin
{quote} Certainly, I must be missing something.
You forgot to say a "bad candlestick pattern" system will scale. Of course it will.

You can't scale a news-related 1-5min chart "system" to a daily chart. You can certainly try.
Congratulations ScalaFX for coming out of the closet!
  • #22
  • Jan 8, 2014 12:09am Jan 8, 2014 12:09am
Quoting nubcake
You can't scale a news-related 1-5min chart "system" to a daily chart. You can certainly try.
That was my very first thought after reading his post lol
The single most important factor is having control of your own life.
  • #23
  • Edited 12:50am Jan 8, 2014 12:40am | Edited 12:50am
Quoting nubcake
{quote} You forgot to say a "bad candlestick pattern" system will scale. Of course it will. You can't scale a news-related 1-5min chart "system" to a daily chart. You can certainly try.

I do not understand the first line of you response. Sarcasm?

News systems focus on trading at specific times, because it is when moves are expected. They focus on lower TFs due to increased speed.

The majority of news driven systems are probably break out systems.

Breakouts occur on all timeframes, you just do not have as well defined time span. Direction to activate at specific times is an input, not the system itself.
  • #24
  • Edited 1:32am Jan 8, 2014 1:00am | Edited 1:32am
Quoting yinyang
How many trades until a system is proven, not just lucky? It should have been ... How many trades until a TRADER is proven, not just lucky?
Yes I think this applies in the case of any discretionary trader / system. If there is really a system that is consistently followed (mechanically), and the system is built on a sound foundation, offering a real advantage, then the probability stats in the 2nd post apply with 90% confidence. If this is a shoot from the hip system, then I don't think the stats mean anything.

If this is a real (mechanically tested and implemented) system that is not the result of luck, spurious backtesting or gut feel, then the stats posted by cwn723 in post #2 are of interest via the calculations in this post. The assumptions:

Quote
A system I am using has been profitable 3 months in a row. But I have only traded it 50 times. When do you know okay this system really does work?

Quote
Thanks cwn.I will definitely read that thread. I thought I had 50 but im 33 wins out of 47, Average win $134 average loss $146.

The analysis:

Quote
Ex. If 70% of those 50 trades were winners, the PDF interval with confidence of 0.9 is .59-.80. That means that you would be 90% confident that the system will win somewhere between 59%-80%.

By my calculation at 59% wins, the actual system expectancy could decrease from the current $50.6 per trade to about $19.2 with 90% confidence. I tend to be a bit pessimistic about these numbers so with 95% confidence the win% could range from 56% - 82% giving an expectancy at the low end of $10.8 per trade. If you need 99% confidence (I do) with this system the low end drops to 52% giving and expectancy with 99% confidence of -$0.40, which might not give you the warm fuzzies that you had hoped for.

On the other hand, if you can increase your trade count to 100 with the same 70% win%, your low end percentage with 99% confidence is at 58%. Not wonderful but still profitable.

To those posting that 5000 or 10000 trades are necessary, given the assumptions in this post, 1000 trades provides a very high degree of confidence, and somewhere between 1000 and 2000 trades the probabilities hit the asymptote, meaning you have a near certain statistical confidence that your win% will be very close to the backtest. So increasing the trade count above 2000 doesn't give any more confidence from a statistical standpoint than 2000 trades (because maximum statistical confidence has already been reached). But as a caveat, I've seen systems based on overfitting meet the 1000-2000 trade level and fail out of sample so proper backtesting is a must or precondition.

The more trades in the test, the harder it is to overfit. But the mere act of optimization introduces a natural bias towards overfitting in system development, even if proper system selection procedures are followed. If proper system selection / backtesting procedures are not followed, then it is very easy to overfit and all these confidence statistics mean exactly nothing.
  • #25
  • Jan 8, 2014 1:35am Jan 8, 2014 1:35am
Quoting vox dei
{quote} "atleast 1 million trade samples" Is that even possible? Typo I suppose. If not, how?
Of course it is only possible in backtest. But I did such a BT to try and explain those fools why and how they can only fail:
http://www.forexfactory.com/showthre...12#post6512412

Note: After this post the trade explorer stopped being updated after the account blow up. Exactly like expected and explained. Yet 6500 new posts have appeared in the thread which is now at 1,547,022 views... Looks like people want to be sold dreams.
No greed. No fear. Just maths.
  • #26
  • Jan 8, 2014 2:09am Jan 8, 2014 2:09am
Quoting Mikkk01
{quote} lol Where did you get this shit? Surely there're systems that work on both but lots are specifically for LTF or HTF. And there may be many different reasons for that. If you're so ignorant that didn't know - I'm sorry you're too much of a newish newbie to give advice here. Read babypips at least! Maybe after that you'll get that your understanding of markets is very poor yet... Don't mislead people. I just laugh at guys like you but some will actually buy the crap you put here - then we'll see muppets [like maggie] who were taught by other...
Sorry i forgot that you are the resident troll in every thread, too sad ....

As i said many times, if you dont believe me test it, every statement can be tested for its truthfullness.What i stated is tested stuff, the fractalness of the market is a very nice property, you should have studied it more before blaming me.
"There's a sucker born every minute" - P.T. Barnum
  • #27
  • Jan 8, 2014 2:15am Jan 8, 2014 2:15am
Quoting Proximus
{quote} I meant on backtesting, on forward test of course its impossible, but on backtest its possible.Just get a big database of historical data, and trade on lower timeframes to have enough samples. If your system works on D1 it should work on M5 too, not identically showing the same results, but since the market is fractal-like, it should be somewhat similar.
Quoting PipMeUp
{quote} Of course it is only possible in backtest. But I did such a BT to try and explain those fools why and how they can only fail: http://www.forexfactory.com/showthre...12#post6512412 Note: After this post the trade explorer stopped being updated after the account blow up. Exactly like expected and explained. Yet 6500 new posts have appeared in the thread which is now at 1,547,022 views... Looks like people want to be sold dreams.
Thank you both for the heads up. My question was strictly about backtest, not forward. I asked because from my experience strategies with the potential of being profitable usually include a number of variables that cap the sample size to thousands. Proximus cleared my doubt in the thread about EMH. At the time of the question I didn't consider that Proximus was referring to testing something as simple as candlesticks using multiple pairs and TFs.

PipMeUp: Worse yet, people sell dreams to themselves, knowingly or unknowingly. As traders we cannot afford such dishonesty with ourselves. Either we remain objective or we won't last long.
"To hold, you must first open your hand. Let go." - Lao Tzu
  • #28
  • Jan 8, 2014 2:16am Jan 8, 2014 2:16am
Quoting Proximus
{quote} Sorry i forgot that you are the resident troll in every thread, too sad .... As i said many times, if you dont believe me test it, every statement can be tested for its truthfullness.What i stated is tested stuff, the fractalness of the market is a very nice property, you should have studied it more before blaming me.
Hey, muppet!

I wonder who you didn't yet call a troll? Seems like everybody is trolling your poor ass
The single most important factor is having control of your own life.
  • #29
  • Jan 8, 2014 2:19am Jan 8, 2014 2:19am
Quoting vox dei
{quote} {quote} Thank you both for the heads up. My question was strictly about backtest, not forward. I asked because from my experience strategies with the potential of being profitable usually include a number of variables that cap the sample size to thousands. Proximus cleared my doubt in the thread about EMH. At the time of the question I didn't consider that Proximus was referring to testing something as simple as candlesticks using multiple pairs and TFs. PipMeUp: Worse yet, people sell dreams to themselves, knowingly or unknowingly. As traders...
So you believe proxy on the point that all strategies work pretty much the same way on all TFs?
The single most important factor is having control of your own life.
  • #30
  • Jan 8, 2014 2:27am Jan 8, 2014 2:27am
Quoting Mikkk01
{quote} So you believe proxy on the point that all strategies work pretty much the same way on all TFs?
?? Can you quote me anywhere saying that? No. So, there's your answer.
"To hold, you must first open your hand. Let go." - Lao Tzu
  • #31
  • Jan 8, 2014 2:36am Jan 8, 2014 2:36am
Quoting vox dei
{quote} ?? Can you quote me anywhere saying that? No. So, there's your answer.
Good!

I asked as if he doesn't know even such a basic thing why would you listen to other stuff he says? He has dim perception of markets overall...
The single most important factor is having control of your own life.
  • #32
  • Jan 8, 2014 3:41am Jan 8, 2014 3:41am
Currently traded system is never proven but you can still trade it profitably indeterminate period of time.

"..real data is only available about the past, ... There's no data on the future. The only way you can look through the fog into the future is if you have really solid theories."

http://www.inc.com/articles/201108/c...ml?cid=sf01001
  • #33
  • Jan 8, 2014 6:25am Jan 8, 2014 6:25am
this is not a good example. the trader on that post was annoyed by all kinds of trolls and naysayers and intentionally wipe out that demo so that all naysayers would leave.

of coz even if he wouldn't do that, he might still have already busted, but that TE's failure is not due to his trading failure.

btw, the method presented by Favourite in that thread is probably only profitable mechanical system i've seen in FF.

Quoting PipMeUp
{quote}http://www.forexfactory.com/showthre...12#post6512412 Note: After this post the trade explorer stopped being updated after the account blow up. Exactly like expected and explained. Yet 6500 new posts have appeared in the thread which is now at 1,547,022 views... Looks like people want to be sold dreams.
  • #34
  • Jan 8, 2014 10:45am Jan 8, 2014 10:45am
Quoting charvo
btw, the method presented by Favourite in that thread is probably only profitable mechanical system i've seen in FF.
I know no way to prove a trading system is profitable. But there are systems which can be proven wrong i.e. long term losers.
Some times BT will disprove a system quickly. Some times the BT won't be able to break the system. We only get a statistical confidence.
For this very system I couldn't prove it wrong but I could statically show (see my posts in this thread) that:
1- the exclusive use of a very small TP and a huge SL in itself doesn't give any edge even in the only direction of the trend.
2- (from 1-) dnabrasil's postitulate (not Favorite's) that the entry timing is unimportant is wrong. The R:R is then injustifiable. Since it is the root of the system, the idea of the system (R:R<<1) is wrong.
3- the initial proposed MM implies that either you lose or you bet the profit again therefore there is no profit taken at any point of time.
BTW I let you check that DNA's rules aren't even respected in the set of trades of his explorer.
No greed. No fear. Just maths.
  • #35
  • Jan 8, 2014 10:51am Jan 8, 2014 10:51am
agreed, i had doubts in his way too, i just said we can't be sure his way loses.

Quoting PipMeUp
{quote}BTW I let you check that DNA's rules aren't even respected in the set of trades of his explorer.
  • #36
  • Jan 8, 2014 11:12am Jan 8, 2014 11:12am
I must be destined to remain in darkness about the reasons systems are non-scalable to other TFs.
  • #37
  • Jan 8, 2014 11:17am Jan 8, 2014 11:17am
too many reasons

1. u didn't define 'system'
2. 2 pips spread, it plays different to a system 50tp 50sl and another system 500tp 500sl.
3. why 1 minute is same as 1 day?
there's something won't happen during a minute, will it? a hour is different from a year, no?
or many funds clear their position by December, but do they do that by the end of every hour?
there are so many difference, so many. it is in fact safer to say one system usually fits only on 1 TF, if it is profitable.

Quoting grin
I must be destined to remain in darkness about the reasons systems are non-scalable to other TFs.
  • #38
  • Jan 8, 2014 1:16pm Jan 8, 2014 1:16pm
Knowledge and Experience are not counted by years

"In Age and Great Invention (NBER Working Paper No. 11359), Jones considers data on Nobel Prize winners in Physics, Chemistry, Medicine, and Economics over the past 100 years, and on outstanding technological innovations over the same period. For comparative purposes, Jones also considers the ages of track and field record-setters and ball players who have received Most Valuable Player awards.

42 percent of innovations came about when their creators were in their 30's... 7 percent were produced by innovators at or before the age of 26 (Einstein's age when he performed his prize winning work)." (NBER Working Paper No. 11359)

Being in any field of endeavor for so many years doesn't equate (nor doesn't proportionately equate) to any measure of success nor any measure of certainty/truth. Sadly to say, it goes the same as with any trading system. The number of back-test doesn't proportionately equate to any measure of certainty. Once you know certain truth about trading, you only need to know the criteria of the trading system. You will then be able to deduce its weakness and strength, success or failure, viable or impracticable, without ever back-testing it. This arise from a simple truth; though there are infinite possibilities, the chart in front of your eyes is only a 2 dimensional flat plane with finite borders.
  • #39
  • Jan 9, 2014 7:33am Jan 9, 2014 7:33am
Quoting Mikkk01
{quote} So you believe proxy on the point that all strategies work pretty much the same way on all TFs?
I didnt said identically the same, but closely the same, i trade on m1 and if my strategy works on m1 it could work in any other timeframe.

Quoting Mikkk01
{quote} Good! I asked as if he doesn't know even such a basic thing why would you listen to other stuff he says? He has dim perception of markets overall...
Oh like you have a perception, the only perception you have is to insult people and call them "muppets" if they dont agree with your ideas, you troll on every thread, got banned from a few i guess, that was for a reason just to know.You are incapable for an intellectual debate, you can only insult people.Which is i think,very childish for a grown up person.

Let me know what useful thing you contributed to this thread, the first ever post you posted on this thread was right to insult me, nothing intellectual after that.

Please do me a favor and look into a mirror more often, because its obvious who is the troll here...
"There's a sucker born every minute" - P.T. Barnum
  • #40
  • Jan 9, 2014 11:47am Jan 9, 2014 11:47am
Quoting PipMeUp
{quote} I know no way to prove a trading system is profitable. But there are systems which can be proven wrong i.e. long term losers. Some times BT will disprove a system quickly. Some times the BT won't be able to break the system. We only get a statistical confidence. For this very system I couldn't prove it wrong but I could statically show (see my posts in this thread) that: 1- the exclusive use of a very small TP and a huge SL in itself doesn't give any edge even in the only direction of the trend. 2- (from 1-) dnabrasil's postitulate...
Possibly the only way to infer a trading system is truly profitable is when you are making bank and your broker changes the rules of the game in order to make that system untradeable.
"To hold, you must first open your hand. Let go." - Lao Tzu
0 traders viewing now