why have you accepted the price-predictability assumption? | Page 103 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

So I accepted a Prop Trading job in South Beach Miami 44 replies

Why you shouldn't show how much you have made in FX 35 replies

Incorporating Slippage Assumption in backtesting 4 replies

Market Predictability - Technical Signals 25 replies

Contesting the Accepted View of Bernanke 0 replies

Attachments: why have you accepted the price-predictability assumption?
Exit Attachments

why have you accepted the price-predictability assumption?

  • #2,041
  • May 14, 2014 3:18pm May 14, 2014 3:18pm
Quoting Kanzler
{quote} Friend you're still looking for the holy grail in an EA. It doesn't exist, and the edge you believe exists here....well, it doesn't. The only way you're ever going to make consistent money in the markets is by knowing with good probability where highs and lows are on your charts.
How do you know it doesn't? Are you sure it doesn't? Notice there are two different hypotheses I am testing; even if it doesn't produce an edge on its own, as long as it produces an edge over the rest of the risk-reward ratio it can still be useful in informing strategy design.

and I am not touting using EA to trade; this experiment can be carried out manually too, but it is way too time consuming; EA is mentioned here because it can help make it easier to conduct the experiment - THAT IS ALL.

I agree about the pinpointing high/low part, but that's a different thing altogether - these can always be modified as constraints to the experiment later on but for a start it'd be good to use a random sampling method to take trades.
Build good relationships with others.
  • #2,042
  • May 14, 2014 3:25pm May 14, 2014 3:25pm
Quoting shellsnail
{quote} How do you know it doesn't? Are you sure it doesn't? Notice there are two different hypotheses I am testing; even if it doesn't produce an edge on its own, as long as it produces an edge over the rest of the risk-reward ratio it can still be useful in informing strategy design. I agree about the pinpointing high/low part, but that's a different thing altogether - these can always be modified as constraints to the experiment later on but for a start it'd be good to use a random sampling method to take trades.
I can't produce a paper for you to review on it, but personally I'm certain of it. From the years I've spent trying to find these "edges", my intuition is to say that if they do exist, they are short lived and not worth the research. You'll probably blow that position off and continue working on it, which is up to you. Just my two cents.
  • #2,043
  • May 14, 2014 3:28pm May 14, 2014 3:28pm
Quoting Kanzler
{quote} I can't produce a paper for you to review on it, but personally I'm certain of it. From the years I've spent trying to find these "edges", my intuition is to say that if they do exist, they are short lived and not worth the research. You'll probably blow that position off and continue working on it, which is up to you. Just my two cents.
Interesting; I have spent around 9000~10,000 hours on trading trying to find these "edges" too and I have arrived at the opposite conclusion: my intuition tells me it exists...

But then again intuition is extremely fallible at times...
Build good relationships with others.
  • #2,044
  • May 14, 2014 3:31pm May 14, 2014 3:31pm
Quoting shellsnail
{quote} Power of the long tail. Yes, I am inclined to think that no matter what time frame there's an edge in this. It is to forget about 1:2 or 1:3 and only go for the trades that have potential of 1:7 and above at least because these events while rare are almost guaranteed to happen in a large enough sample of outcomes. Think Poisson distribution. In other words, the hypothesis of if I place 1:1RR trades in the markets randomly, then I will lose 50% of my trades implies that if I set 1:7, I will lose 7/8 of my trades is not true... the scaling...
Lol I wish I was as smart as you are in math. I can see the picture but to put it down like that is something that gets me everytime. Keep searching man, don't let anyone discourage you from finding an answer to your questions.
Being a trader is lonely, but being a great trader is lonelier still
  • #2,045
  • May 14, 2014 3:46pm May 14, 2014 3:46pm
Just to add, many relatively recent studies have shown that financial markets fluctuations are better modelled by a power law distribution (which has fatter and longer tails) than a normal distribution.

http://www.nature.com/nature/journal...ture01624.html
http://physicsoffinance.blogspot.co....n-finance.html

Quote
The first regularity is the distribution of returns over various time intervals, which has been found generically to have broad power law tails -- "fat tails" -- implying that large fluctuations up or down are much more likely than they would be if markets fluctuated in keeping with normal Gaussian statistics. Anyone who read The Black Swan knows this.
Quote
So, its the directionality of price movements which has very little predictability, whereas the magnitude of changes follows a process with much more interesting structure. It is in the record of this volatility that one sees potentially deep links to other physical processes, including earthquakes. A particularly interesting paper is this one, again by the Boston group, quantifying several ways in which market volatility obeys several quantitative laws known from earthquake science, especially the Omori Law describing how the probability of aftershocks decays following a main earthquake. This probability decays quite simply in proportion to 1/time since the main quake, meaning that aftershocks are most likely immediately afterward, and become progressively less likely with time. Episodes of high volatility appear to follow similar behaviour quite closely.
​

This seems to lend weight, if not directly confirm the hypothesis I am proposing to test.
Build good relationships with others.
  • #2,046
  • May 14, 2014 3:52pm May 14, 2014 3:52pm
Quoting shellsnail
{quote}. It would be interesting if someone can write an EA that opens trades completely randomly (random buy/sell; random time of the day), open up 20 demo accounts and do a test from 1:1 to 1:20RR and see what results we get over say a 1 year period.


I did that half year ago.Guess what, the RR proportion will have random outcomes, the higher the Reward than risk, the less samples you will have to analyze.Even the cumulative the expectancy will be higher in some cases, with Reward>3*Risk, but it is unpredictable.

You cant just say : i`ma have a 20 pip stoploss and 200 pip TP and will make a million dollar trading this way.Its nonsense.

Money management will not give you any edge...You have to look for another edge.

Besides read my previous 2 posts, i wrote those to educate people not to look like a troll even if you think it was a little bit shocking.
"There's a sucker born every minute" - P.T. Barnum
  • #2,047
  • May 14, 2014 3:54pm May 14, 2014 3:54pm
Quoting Proximus
{quote} I did that half year ago.Guess what, the RR proportion will have random outcomes, the higher the Reward than risk, the less samples you will have to analyze.Even the cumulative the expectancy will be higher in some cases, with Reward>3*Risk, but it is unpredictable. You cant just say : i`ma have a 20 pip stoploss and 200 pip TP and will make a million dollar trading this way.Its nonsense. Money management will not give you any edge...You have to look for another edge. Besides read my previous 2 posts, i wrote those to educate...
Would you mind making public the data you've documented and explaining briefly the experimental and statistical methods you've used? If it is not done in a rigorous fashion then really nothing can be concluded.
Build good relationships with others.
  • #2,048
  • May 14, 2014 4:07pm May 14, 2014 4:07pm
Anyway, if anyone is interested in exploring this question further together, do drop me a PM.
Build good relationships with others.
  • #2,049
  • May 14, 2014 4:21pm May 14, 2014 4:21pm
Quoting shellsnail
{quote} Would you mind making public the data you've documented and explaining briefly the experimental and statistical methods you've used? If it is not done in a rigorous fashion then really nothing can be concluded.
I`m afraid i dont have the data anymore, and it wasnt any fancy data i`m not a scientist, i just made an EA which exactly did the same as you mentioned, it was only a simple EA which put trades totally randomly.

1) Firsts i tried each trade to be total random, and ran through on H1 timeframe, 10 years of historical data, on all combinations of TP/SL from 10 to 1000 pip with 3 pip steps, it took me 3 continuous days to finish, and after that the conclusion was that, there was a slightly better performance that Reward>Risk scenarios performed better, but there were also cases when they performed worse than the average which was around 0 expectancy slightly negative due to spread.
2) Secondly i tried randomizing the daily trades, to trade in daily clusters, meaning that if the dice was bullish, we only put buy orders that day, and ask for another random number the next day only
3) I tried in weekly and monthly clusters

Same situation in all of them, the overall situation was better for Reward>Risk , and also the weekly randomized cluster was the best, but there was no single situation where you could say: aha 10 pip SL and 600 pip TP is the holy grail, or like that.

Even there were some cases where through 10 years, some pairs of SL/TP gave profit, but in these cases there were whole years where it didnt, so even if 600 pip TP/ 10 pip SL gave you a 1000% growth over 10 years, it may be -20% for the first 2 years.

Now i dont think that any trader here would risk being 2 years in the minus for that ,knowing that i might never come.

So the conclusion is that the profit was also somewhat random, meaning that money management is not a source of edge.Sorry!

Oh and btw i used true random numbers, not pseudo.

Quoting shellsnail
Anyway, if anyone is interested in exploring this question further together, do drop me a PM.
We could talk like that if you dont want to pollute this thread with this, but i think its better for anyone to see this and learn from this.
"There's a sucker born every minute" - P.T. Barnum
  • #2,050
  • May 14, 2014 4:25pm May 14, 2014 4:25pm
Quoting Proximus
{quote} I`m afraid i dont have the data anymore, and it wasnt any fancy data i`m not a scientist, i just made an EA which exactly did the same as you mentioned, it was only a simple EA which put trades totally randomly. 1) Firsts i tried each trade to be total random, and ran through on H1 timeframe, 10 years of historical data, on all combinations of TP/SL from 10 to 1000 pip with 3 pip steps, it took me 3 continuous days to finish, and after that the conclusion was that, there was a slightly better performance that Reward>Risk scenarios performed...
Hmm... ok thanks for the explanation!, but I will need to know more about how you randomised the trades, and also how you determined the parameters (SL/TP) for the different time-frames, and whether did you account for spread, did you take on more than one trade at a time, were there correlated positions etc. These could all affect the outcomes in many ways.. You mentioned that reward > risk scenarios performed slightly better - there needs to be a statistical test carried out to see if this is a significance difference or not cause this is the part that can reject the hypothesis.
Build good relationships with others.
  • #2,051
  • May 14, 2014 4:29pm May 14, 2014 4:29pm
Quoting shellsnail
Just to add, many relatively recent studies have shown that financial markets fluctuations are better modelled by a power law distribution (which has fatter and longer tails) than a normal distribution. http://www.nature.com/nature/journal...ture01624.html http://physicsoffinance.blogspot.co....n-finance.html {quote} {quote}​ This seems to lend weight, if not directly confirm the hypothesis I am proposing to test.

Those links are some good reading. I would have to go over it a few times to grasp some of it. But they do make some sense.
Being a trader is lonely, but being a great trader is lonelier still
  • #2,052
  • May 14, 2014 4:39pm May 14, 2014 4:39pm
Quoting shellsnail
{quote} Hmm... ok thanks for the explanation!, but I will need to know more about how you randomised the trades, and also how you determined the parameters (SL/TP) for the different time-frames, and whether did you account for spread, did you take on more than one trade at a time, were there correlated positions etc. These could all affect the outcomes in many ways.. You mentioned that reward > risk scenarios performed slightly better - there needs to be a statistical test carried out to see if this is a significance difference or not cause this...
Well its more complicated that that, i also ran a volatility test, to test in which time the volatility was high, cos its obvious that the trade cost would be the most optimal for us in high volatility with this strategy.

So say 600 pip TP and 10 pip SL in normal conditions would be no different than 10 pip TP and 600 pip SL if we look at expectancy, but if the price swings with big amplitude waves, there is a slight edge, because if the average amplitude is > 600 pip then we have an edge with 600 pip TP/ 10 pip SL.

But of course the swing size is unpredictable, and always changing so, if you use 600 pip TP /10 pip SL for 10 months and you make a ton of money, the market could enter in recession where traders pull out their money, and the volatility decreases, so you need to adjust your "swing-catcher".

Also did a daily volatility check, and yep the London Session still rocks.My analysis was filtered by this 2 criterias, the monthly/ seasonal volatility and the daily volatility sessions.And still havent found a reliable combination which would perform well through many years.

As i said, there were cases where the expectancy was positive, and not just individually but in clusters, for example from 500 pip TP to 600 pip with 10 pip SL it all performed well, but not in every year

And you dont want negative growth for an entire year, its unacceptable, even if the long term might, not certain, give good results.

The spread, is irrelevant, i used the daily average spread, but in high volatility conditions, where our strategy might work, the spread could be horrible.So thats why i tested it only on H1, lower than that its not good, higher than that = not enough data.

Quote
You mentioned that reward > risk scenarios performed slightly better - there needs to be a statistical test carried out to see if this is a significance difference or not cause this is the part that can reject the hypothesis.

Yep, so does trading from 7 am GMT to 16 pm GMT, and in high volatility days like tuesday/wednesday or excluding december/january where the volatility was low.

But still, it was not acceptable.Sorry this wont give you a reliable edge, trading like this is gambling.
"There's a sucker born every minute" - P.T. Barnum
  • #2,053
  • May 14, 2014 4:41pm May 14, 2014 4:41pm
Quoting shellsnail
{quote} Hmm... ok thanks for the explanation!, but I will need to know more about how you randomised the trades, and also how you determined the parameters (SL/TP) for the different time-frames, and whether did you account for spread, did you take on more than one trade at a time, were there correlated positions etc. These could all affect the outcomes in many ways.. You mentioned that reward > risk scenarios performed slightly better - there needs to be a statistical test carried out to see if this is a significance difference or not cause this...
I made the same kind of experiment with the same conclusion. One trade at a time, spread taken into account, repeating the backtest several times. The expectancy quickly converges to minus the spread. Worse, I got the same conclusion when always opening the trade in the direction of the slope of a MA instead of flipping a coin.
I even tried very small TP/huge SL to see if trading the noise was an option: http://www.forexfactory.com/showthre...12#post6512412
No greed. No fear. Just maths.
  • #2,054
  • May 14, 2014 4:41pm May 14, 2014 4:41pm
Quoting Proximus
{quote} Well its more complicated that that, i also ran a volatility test, to test in which time the volatility was high, cos its obvious that the trade cost would be the most optimal for us in high volatility with this strategy. So say 600 pip TP and 10 pip SL in normal conditions would be no different than 10 pip TP and 600 pip SL if we look at expectancy, but if the price swings with big amplitude waves, there is a slight edge, because if the average amplitude is > 600 pip then we have an edge with 600 pip TP/ 10 pip SL. But of course the swing...
But if you can prove to a certain degree of confidence that a high RR outperforms a low RR in a random trade scenario, you can potentially make a case to abandon all low RR trading systems.

I'm not looking for a standalone trading system from this - just possible verification of the hypothesis that low R:R does not perform well. Do we have enough evidence/data to make that conclusion?

And why am I doing this? For curiosity sake really. I have a working method and I am not actively looking to change my system, but if something interesting pops out from this research, who knows, it might make a positive difference to my current methods too.
Build good relationships with others.
  • #2,055
  • May 14, 2014 4:46pm May 14, 2014 4:46pm
Quoting PipMeUp
{quote} I made the same kind of experiment with the same conclusion. One trade at a time, spread taken into account, repeating the backtest several times. The expectancy quickly converges to minus the spread. Worse, I got the same conclusion when always opening the trade in the direction of the slope of a MA instead of flipping a coin. I even tried very small TP/huge SL to see if trading the noise was an option: http://www.forexfactory.com/showthre...12#post6512412
Yes, I am aware the expectancy might converge to minus the spread, but my question is in two parts...

1) Does any RR have an edge on its own?
2) Does high RR outperform low RR?
2b) Under what conditions might one outperform the other?
Build good relationships with others.
  • #2,056
  • May 14, 2014 4:51pm May 14, 2014 4:51pm
Quoting shellsnail
{quote} But if you can prove to a certain degree of confidence that a high RR outperforms a low RR in a random trade scenario, you can potentially make a case to abandon all low RR trading systems. I'm not looking for a standalone trading system from this - just possible verification of the hypothesis that low R:R does not perform well. Do we have enough evidence/data to make that conclusion? And why am I doing this? For curiosity sake really. I have a working method and I am not actively looking to change my system, but if something interesting...
If you gamble like that with the "swing/amplidute catcher", higher Reward>Risk gives you a slight advantage,obviously because your exploiting the volatility, however, its still gambling, and you will inevitably lose money after a time like with all gambling.

But in trading, from my experience, money management is irrelevant to the edge.Precisely configuring SL/TP configurations will not give you any edge, no matter what is the ratio between Reward and Risk.

I currently trade with a 1:2 Reward Risk method, and i`m very profitable, so i cant say that Reward<Risk is bad, nor that Reward>Risk is bad.The only possible answer is that its irrelevant.And so do i believe.
"There's a sucker born every minute" - P.T. Barnum
  • #2,057
  • May 14, 2014 4:56pm May 14, 2014 4:56pm
Quoting Proximus
{quote} If you gamble like that with the "swing/amplidute catcher", higher Reward>Risk gives you a slight advantage,obviously because your exploiting the volatility, however, its still gambling, and you will inevitably lose money after a time like with all gambling. But in trading, from my experience, money management is irrelevant to the edge.Precisely configuring SL/TP configurations will not give you any edge, no matter what is the ratio between Reward and Risk. I currently trade with a 1:2 Reward Risk method, and i`m very profitable,...
I can't see the connection between having a slight advantage => gambling.

If there is indeed a slight advantage then why not shift all systems to have high R:R to exploit this slight advantage? Or does this slight advantage disappear under certain conditions? These are questions I want to explore.
Build good relationships with others.
  • #2,058
  • May 14, 2014 5:06pm May 14, 2014 5:06pm
Quoting shellsnail
{quote} I can't see the connection between having a slight advantage => gambling. If there is indeed a slight advantage then why not shift all systems to have high R:R to exploit this slight advantage? Or does this slight advantage disappear under certain conditions? These are questions I want to explore.
I was trying to say that if you gamble with that amplitude catcher method, which is gambling btw because the average expectancy is <0.That is the definition of gambling.You have higher chance, with Reward>Risk however still <=0 , and the uncertainty is gigantic.

Even if you tune your system to a today's perfect ranging market where the amplitude of each swing has a small deviation from the mean, you can never know when a trend kicks in and the trend might be going for years, like in the case of EUR/USD from 2002 to 2004.

In an uptrend like that, if you randomly trade, you will have a negative expectancy for 2 years, cos you put equal weight to BUY/SELL orders because u put orders randomly, but the market puts more weight on BUY orders.

Its just unpredictable, unreliable and very risky.
-------------------

However if you have a real edge, and ur trading not gambling , then it doesnt matter what is your RR ratio.I have a 1:2, but i dont really care about it, it could be 2:1 aswell, doesnt matter, i put the SL and the TP where my method tells me.

So if you are a trader then RR ratio is not your concern, focus more on predicting the market, rather than adjusting your position.And also manage risk, but not because of seeking an edge, but because to protect the capital.
"There's a sucker born every minute" - P.T. Barnum
  • #2,059
  • May 14, 2014 5:13pm May 14, 2014 5:13pm
Quoting Proximus
{quote} I was trying to say that if you gamble with that amplitude catcher method, which is gambling btw because the average expectancy is <0.That is the definition of gambling.You have higher chance, with Reward>Risk however still <=0 , and the uncertainty is gigantic. Even if you tune your system to a today's perfect ranging market where the amplitude of each swing has a small deviation from the mean, you can never know when a trend kicks in and the trend might be going for years, like in the case of EUR/USD from 2002 to 2004. In an uptrend like...
Ok I think we diverge fundamentally on some beliefs regarding the market. At this point I have no idea who's right/wrong because most of these beliefs are formed from personal experience and sometimes faulty reasoning/ bias so I am open to accepting new beliefs provided they come grounded with evidence...

For me I think RR is a huge concern and predicting the market is really secondary. All I do is to draw a few horizontal lines (the prediction part), take a few entries around those lines, set a good RR, move my SL to BE when trade goes in my favour sufficiently and then let it play out. Each scenario is different and the only thing that keeps me sane and profitable is the high RR. The RR is really the edge because the shocks/moves in the market are somewhat auto-correlated like the aftershocks of an earthquake, and having a good RR allows you to take advantage of this systematic part of the market... But this remains a hypothesis - I want to see if there indeed is an edge from high RR over low RR due to this nature of the market (that I am assuming)
Build good relationships with others.
  • #2,060
  • May 14, 2014 5:23pm May 14, 2014 5:23pm
Quoting shellsnail
{quote} Ok I think we diverge fundamentally on some beliefs regarding the market. At this point I have no idea who's right/wrong because most of these beliefs are formed from personal experience and sometimes faulty reasoning/ bias so I am open to accepting new beliefs provided they come grounded with evidence...
It is ok to be skeptic, and if you believe me or not, you can always test this yourself, but eventually you will come to the same conclusion as i did.I`m just saving you time and telling it to you now.

Quote
or me I think RR is a huge concern and predicting the market is really secondary. All I do is to draw a few horizontal lines (the prediction part), take a few entries around those lines, set a good RR, move my SL to BE when trade goes in my favour sufficiently and then let it play out. Each scenario is different and the only thing that keeps me sane and profitable is the high RR.

Then you are a gambler not a trader.You try to exploit the volatility or the structure of the market, which is unpredictable.You may be succesful for now, but this wont last forever.Focusing on the direction however and succesfully predicting it, will give you the so seeked edge, and after that, the RR will be irrelevant.

Quote
. The RR is really the edge because the shocks/moves in the market are somewhat auto-correlated like the aftershocks of an earthquake, and having a good RR allows you to take advantage of this systematic part of the market... But this remains a hypothesis - I want to see if there indeed is an edge from high RR over low RR due to this nature of the market (that I am assuming)

Yes, that is a volatility /breakout strategy, which is good, but eventually in trending markets you will lose money.Lucky for you because the EUR/USD is now in somewhat ranging, but it will not be always like this.

Its either a gambling method, or you have a real edge and you dont really know that you have it, but one thing is certain it doesnt come from the RR.I can trade trending or ranging markets with any RR and still make profit,because i focus on prediction rather than position adjusting.
"There's a sucker born every minute" - P.T. Barnum
2 traders viewing now