Daily ranges / averages

[trendie]

I don't want to hijack another thread, but there was a post about daily ranges....

(Mike Farrells EJ Daytraders v Rest of World)

He mentioned the EJ had a recent range of a mere 60-ish, but had a one-day move of 260+ pips.

That got me thinking about daily averages.

If your strategy relies on a daily range, you have to treat big range days as windfalls.

For example, if the average is 100 pips. and your strategy relies on taking 70 pips of a given average, you CANNOT treat a random (possibly news-related) 300 pips day to skew your strat>?

Or can you?

If 4 of 5 days there is a 100 pip average, and there is a random 300 pip day, if you decide to factor in the windfall day and "hope" to stretch your average returns you run the risk of losing out on good, stable, returns for the sake of random outliers.

If you average daily range is 100, and every 5 days you get a 300 pipper......

You can either make a decent 70 pips a day, or skew your averages ( 4 x 100 plus 300 = 700).
700 / 5 days = 140 pips. 70% of 140 = 98 pips.

Not counting the outlier, the average would be 70 pips; with the outlier = 98 pips.

For the sake of stretching for an extra 20 plus pips, you could watch a decent trade fall back from a 70 pips win, to a null trade, or drag out over 2 days.

I think we need to apply some non-mathematical reasoning and recognise between "true" daily averages and windfall, outlier moves.

For example, If the market moves 100 pips a day for 4 days, then there is some outlier event for 300 pips, the "true" range is still 100 pips, but the market has corrected, or there has been some sentiment shift, and having made its correction, has settled back to its "normal" daily range.

Do any of you guys recognise the concept of an "outlier event", or would you just use simple averages to think that the daily range has changed?

 

[wackypete2]

I don't want to hijack another thread, but there was a post about daily ranges....

(Mike Farrells EJ Daytraders v Rest of World)

He mentioned the EJ had a recent range of a mere 60-ish, but had a one-day move of 260+ pips.

That got me thinking about daily averages.

If your strategy relies on a daily range, you have to treat big range days as windfalls.

For example, if the average is 100 pips. and your strategy relies on taking 70 pips of a given average, you CANNOT treat a random (possibly news-related) 300 pips day to skew your strat>?

Or can you?

If 4 of 5 days there is a 100 pip average, and there is a random 300 pip day, if you decide to factor in the windfall day and "hope" to stretch your average returns you run the risk of losing out on good, stable, returns for the sake of random outliers.

If you average daily range is 100, and every 5 days you get a 300 pipper......

You can either make a decent 70 pips a day, or skew your averages ( 4 x 100 plus 300 = 700).
700 / 5 days = 140 pips. 70% of 140 = 98 pips.

Not counting the outlier, the average would be 70 pips; with the outlier = 98 pips.

For the sake of stretching for an extra 20 plus pips, you could watch a decent trade fall back from a 70 pips win, to a null trade, or drag out over 2 days.

I think we need to apply some non-mathematical reasoning and recognise between "true" daily averages and windfall, outlier moves.

For example, If the market moves 100 pips a day for 4 days, then there is some outlier event for 300 pips, the "true" range is still 100 pips, but the market has corrected, or there has been some sentiment shift, and having made its correction, has settled back to its "normal" daily range.

Do any of you guys recognise the concept of an "outlier event", or would you just use simple averages to think that the daily range has changed?

The daily range on EJ HAS changed since Mike started his thread. He talks about an outlier, but even without removing that outlier the range has constricted. Just look at a chart. That's why his method is not giving the same results. Of course it's easy to see in hindsight, I'm not criticizing. The cause for change lies in how much interest there is in a currency pair due to recent macro or semi macro events. The outliers are generally caused by a skewed data point or some politician blowing hot air.

Peter

 

[Purple Brain]

The choice of what period you choose to analyse the range will make an obvious difference to your results and any attempt to use a smoothing average guarantees you consistently miss hitting your target. Daily range appears to exhibit even fewer characteristics than does the price itself of normal distribution, but the lack of any directional bias in range data would suggest narrowing down analysis of it to using oscillator type tools.

You suggest using ‘non-mathematical reasoning’ to consider the very mathematical construct of an outlier which itself can only really make sense within a discussion involving even more mathematical kurtosis and normal distribution. But the real show stopper for me it that since the middle of June this pair was a lot less volatile (75-150 pips daily) than the period from the beginning of the year (95-270). How can it make sense to decide what is an outlier and what is valid – especially as you transition from volatile to non-volatile periods?

In trying to understand what strategy would utilise the daily range the only method I could come up with is that of attempting to guess where the price is less likely to go and consider that an outlier in the context of your query. Then it occurred to me that some brokers give you that data by their pricing of box options. They probably know better than us on any given day what the probabilities are of the price hitting any given level. You’d obviously need to back engineer that data to whatever your own method involving daily range is of course.

Personally, at my current level of experience I have enough trouble with just the basic price development itself to risk getting into even deeper abstractions, but I’ll be interested to see your progress on this issue.

 

[Liquid validity]

Most pairs are back to pre 08 volatility levels now.
As for outliers, depends whether or not you're pro / anti Taleb...

 

[NVP]

good potential in this thread ..........I see Ed sakoytas whipsaw song in your signature area .........if you are a pure trend follower then you have surely answered (or chosen to ignore) these questions raised already ?

N

 

[trendie]

good potential in this thread ..........I see Ed sakoytas whipsaw song in your signature area .........if you are a pure trend follower then you have surely answered (or chosen to ignore) these questions raised already ?

N

For myself, I have answered it. I recognise I simply don't have the inside knowledge to be clever, so just let price do what it needs to, and follow where price leads me.

The thread is about those who take profits based on a function of an average of X days, and whether they ignore outlier days, in order not to skew a normal days potential profits.

 

[trendie]

The choice of what period you choose to analyse the range will make an obvious difference to your results and any attempt to use a smoothing average guarantees you consistently miss hitting your target. Daily range appears to exhibit even fewer characteristics than does the price itself of normal distribution, but the lack of any directional bias in range data would suggest narrowing down analysis of it to using oscillator type tools.

You suggest using ‘non-mathematical reasoning’ to consider the very mathematical construct of an outlier which itself can only really make sense within a discussion involving even more mathematical kurtosis and normal distribution. But the real show stopper for me it that since the middle of June this pair was a lot less volatile (75-150 pips daily) than the period from the beginning of the year (95-270). How can it make sense to decide what is an outlier and what is valid – especially as you transition from volatile to non-volatile periods?

In trying to understand what strategy would utilise the daily range the only method I could come up with is that of attempting to guess where the price is less likely to go and consider that an outlier in the context of your query. Then it occurred to me that some brokers give you that data by their pricing of box options. They probably know better than us on any given day what the probabilities are of the price hitting any given level. You’d obviously need to back engineer that data to whatever your own method involving daily range is of course.

Personally, at my current level of experience I have enough trouble with just the basic price development itself to risk getting into even deeper abstractions, but I’ll be interested to see your progress on this issue.

I concede that perhaps outlier needs to be defined.

If the average move over 10 days is 8 days of 100 pips, and 2 days of 250 and 300, the average will be the sum of the 10 divided by 10.
But, "non-mathematical reasoning" may lead you to think that the 2 outlier days are abnormal, and should not derail you from the reality that 80% of the time the daily range is 100. (I think thats what I mean by non-maths)

By trying to stretch to win profits on the 2 abnormal days, you might undermine your chances of taking plodding, pedestrian, bread-and butter daily returns.
Actually, even on the outlier days, if you have modest targets, even the outlier days could be profitable.

Just an idle curiosity about how we average and its meaning.

 

[Purple Brain]

Just an idle curiosity about how we average and its meaning.

I imagine it was idle curiosity that led to the simple moving average having exponential and linear weighted variants. A lot to be said for idle curiosity.

Anyone trading set stop and target methods will, as you say, make their bread and butter on large range outlier days just like any other. It's the less than average outliers that'll get them.