to good to be true?

[davidh1819]

i do not look at the markets in terms of flipping a coin 50-50

i look at them on a 6 month chart use a few indicators and look at the candle sticks and i can decide if there is a higher probability that the marked will go in a particular direction so it could be more like 30-70

 

[Scotty2Cues]

Sure he can calculate a confidence interval. But firstly, the less samples he has, the less precise the interval, and secondly, it says nothing about what happens in that 5%. Consider the banks and their VAR, or consider how a strategy with no stop loss and small target would look from just 12 trades if you created a confidence interval.

would the standard deviation be large on that though?

What if we assume r:r is always 1:1 and after 12 trades results are

-1, -1, 1,1,1,1,1,1,1,1,1,1 the mean is 0.67r and standard deviation is 0.78r

so using formula i just looked up for confdence intervals then if we view each trade as drawn from blocks of 12 then 95% of time we have win rate per trade of

0.67r+ or - 0.44r i got 0.44 from 1.96*standarddeviation/sqroot(number of trades)

 

[Shakone]

i do not look at the markets in terms of flipping a coin 50-50

i look at them on a 6 month chart use a few indicators and look at the candle sticks and i can decide if there is a higher probability that the marked will go in a particular direction so it could be more like 30-70

Sure, I understand that. But how are you getting this 30-70 figure? Backtesting? The 12 trades? You see my point.

I still would echo what others have said that you shouldn't worry about the 15 mins aspect. If you have an edge, forget about that. BUt as Shadowninja said, trade the same way for a year before thinking it is easy. You are missing that experience.

 

[Shakone]

would the standard deviation be large on that though?

What if we assume r:r is always 1:1 and after 12 trades results are

-1, -1, 1,1,1,1,1,1,1,1,1,1 the mean is 0.67r and standard deviation is 0.78r

so using formula i just looked up for confdence intervals then if we view each trade as drawn from blocks of 12 then 95% of time we have win rate per trade of

0.67r+ or - 0.44r i got 0.44 from 1.96*standarddeviation/sqroot(number of trades)

So your calculations are showing what exactly? That with 95% confidence you might LOSE 0.4 per trade. So stats tells you that 10 out of 12, means that with 95% confidence you could lose 0.4 per trade, and in the other 5% you have no idea what you would win or lose........now, should you apply for the mortgage now because you'll have 6 figures a year?

And maybe the opening poster got 9 wins out of 12, or 8 out of 12 etc. And there are other uncertainties that are ignored in the stat calculation too, but your numbers should tell you 12 is not nearly enough.

 

[Scotty2Cues]

So your calculations are showing what exactly? That with 95% confidence you might LOSE 0.4 per trade. So stats tells you that 10 out of 12, means that with 95% confidence you could lose 0.4 per trade, and in the other 5% you have no idea what you would win or lose........now, should you apply for the mortgage now because you'll have 6 figures a year?

And maybe the opening poster got 9 wins out of 12, or 8 out of 12 etc. And there are other uncertainties that are ignored in the stat calculation too, but your numbers should tell you 12 is not nearly enough.

no +-0.44 is the interval so 95% chance of being between 0.23R and 1.11R

 

[Shakone]

Ok, I got ya. So with 95% probability you get between 0.23R and 1.11R, and the other 5%? Still could be a negative expectancy couldn't it. And it is just an estimate, so this will change with each new trade. If the next is a loser, now what is the confidence interval etc.?

Then as I mentioned, we've ignored a lot of other issues. You are assuming it is normally distributed right? Otherwise, how else did you get 1.96? Is it normally distributed? If not, then how confident are you? Also assumed things are identically independently distributed, but again, are they? I'd say no.

Consider how a no stop loss system with small targets will look over a small sample of trades and what the confidence interval looks like. These things are fine as far as they go, but you have to consider reality first.

 

[Scotty2Cues]

Ok, I got ya. So with 95% probability you get between 0.23R and 1.11R, and the other 5%? Still could be a negative expectancy couldn't it. And it is just an estimate, so this will change with each new trade. If the next is a loser, now what is the confidence interval etc.?

Then as I mentioned, we've ignored a lot of other issues. You are assuming it is normally distributed right? Otherwise, how else did you get 1.96? Is it normally distributed? If not, then how confident are you?

Again, consider how a no stop loss system with small targets will look over a small sample opf trades and what the confidence interval looks like.

the other 5% of the time it lies outside the inerval, although we assuming 1:1 so cant ever be outside 1. Doesnt need to be normally distributed. If we take any distrubution and plot means of say blocks of 12 trades, that distribution will always be normal and will have the same mean (Central limit theorem)

But with no sl how big will your losses be?


Ordered these lol

 

[Shakone]

the other 5% of the time it lies outside the inerval, although we assuming 1:1 so cant ever be outside 1. Doesnt need to be normally distributed. If we take any distrubution and plot means of say blocks of 12 trades, that distribution will always be normal and will have the same mean (Central limit theorem)

But with no sl how big will your losses be?


Ordered these lol

And now you make even more assumptions without realising perhaps, that the statistics has knowledge of what you're doing and have taken that into account. They haven't. How does it know the maximum loss is 1r? It doesn't. So at the beginning of the calculation you've assumed things are normally distributed, then later on, you've imposed a boundary which ensures it definitely ISN'T normally distributed (meaning your calculation is off-the 1.96 is specific to normal distributions, another distribution and it isn't the same,; plus I wouldn't use the central limit theorem since that is in the limit, and 12 is a long way from infinity, and there are other problems too).

This is the problem with aplying stats like this. There is a tendency to see what you want (the 95% confidence interval) and ignore the reality (that it is just a fluid estimate changing with each sample, and the other 5% counts), and to make assumptions that don't hold.

 

[advfntrader]

beginners luck, you'll blow the lot before 2011 is out. I can almost guarantee it

 

[meanreversion]

beginners luck, you'll blow the lot before 2011 is out. I can almost guarantee it

Harsh, but funny.

 

[Scotty2Cues]

And now you make even more assumptions without realising perhaps, that the statistics has knowledge of what you're doing and have taken that into account. They haven't. How does it know the maximum loss is 1r? It doesn't. So at the beginning of the calculation you've assumed things are normally distributed, then later on, you've imposed a boundary which ensures it definitely ISN'T normally distributed (meaning your calculation is off-the 1.96 is specific to normal distributions, another distribution and it isn't the same,; plus I wouldn't use the central limit theorem since that is in the limit, and 12 is a long way from infinity, and there are other problems too).

This is the problem with aplying stats like this. There is a tendency to see what you want (the 95% confidence interval) and ignore the reality (that it is just a fluid estimate changing with each sample, and the other 5% counts), and to make assumptions that don't hold.

i havent assumed normal distribution as it doesnt matter and not sure if you are missing the point about central limit. Also if we vary r:r and our results were:

-1,-1,-1,-1,-1,0,0,0,1,4,4,4

so most were losers/breakeven but we hit a couple of home runs then the mean turns out to be the same as before but now the standard deviation is 2.1 (our results are much more spread out) which is much larger than before.

our interval is now 0.67 + or - 1.2 (or -0.53 to 1.87) so we can't place as much confidence in our strategy and so need a bigger sample size (think there is a formula to calculate the sample size needed).

PUBS!

 

[Scotty2Cues]

so given 2 traders to choose from with 12 trades each and results (using 1:R so result 4 means 4 times risk)

A: -1, -1, 1,1,1,1,1,1,1,1,1,1

B: -1,-1,-1,-1,-1,0,0,0,1,4,4,4

I would choose A as his strategy/results are alot more reliable even though they both made the same profit!

definitely pubs

 

[adriandecleir1]

Now 15 minutes a day of watching a screen within the next decade can make me an abundance of money? Earning Enough each year to never work again?

Is there something I am missing?

Would it be possible to backtest your system?

 

[davidh1819]

Ok Some of you are using complex statistics and probability here i like to keep things simple higher probabaility higher risk reward ratio and go for it then stick ot it.

Also my strategy is not written in stone the nature of market dynamics will prompt me to make alterations to it in the future so i am told

 

[Shakone]

i havent assumed normal distribution as it doesnt matter and not sure if you are missing the point about central limit. Also if we vary r:r and our results were:

-1,-1,-1,-1,-1,0,0,0,1,4,4,4

so most were losers/breakeven but we hit a couple of home runs then the mean turns out to be the same as before but now the standard deviation is 2.1 (our results are much more spread out) which is much larger than before.

our interval is now 0.67 + or - 1.2 (or -0.53 to 1.87) so we can't place as much confidence in our strategy and so need a bigger sample size (think there is a formula to calculate the sample size needed).

PUBS!

When you make a confidence interval, what distribution have you assumed? Do you think every distribution has a confidence interval like that? Where do you think the number 1.96 comes from?

You HAVE assumed normal distribution, whether you realise it or not is another issue.

Happy new Year Scotty

 

[Scotty2Cues]

When you make a confidence interval, what distribution have you assumed? Do you think every distribution has a confidence interval like that? Where do you think the number 1.96 comes from?

You HAVE assumed normal distribution, whether you realise it or not is another issue.

Happy new Year Scotty

Im far from an expert and my books havent come yet :cheesy: but from what i understand if you take ANY distribution and then plot distributions of samples from that, the result will ALWAYS be a normal distribution and its that distribution where the confidence interval is calculated. The link below explains it well

http://www.usablestats.com/lessons/central_limit

 

[musslewhite]

You are probably NOT missing anything... people think that they SHOULD work more because they are doing very well, somehow human psyque makes them feel they don´t deserve to have good returns because they are not burning the midnight oil, the truth is that simple strategies work, it will be important for you to follow your system´s dicipline and not to stray away from it when you continue to have good returns, don´t fix what aint broken brother.

Congrats!