Just because you feel threated because they know more than you doesn't necessarily mean they're wrong. Think of it like trying to find consistencies in the inconsistent.
Dilemma in modeling idiosyncratic high-frequency noise
- Thread starter bluelou
- Start date
Just because you feel threated because they know more than you doesn't necessarily mean they're wrong. Think of it like trying to find consistencies in the inconsistent.
Time based charts do not typically have volume displayed, the user chooses to display volume on a chart which is inefficient. Ineffecient because it has to be interpreted and that process causes confusion and is a distraction.
When you use Constant Volume/Share Bar Charts there is no need for time to be displayed, most people wear watches if they need to be somewhere. Displaying time on a chart again is not only as ineffecient displaying volume on a time based chart but silly as well.
Time is not the reason for the noise. Time is simply a factor in the way a chart is displayed that causes the noise. The noise is created in the inconsistency in the way the bars are created on any chart. Think about calling the "noise" something different. The inconsistency in the way the bars are created on any chart is causing that chart to be unstable. Unstable because you are putting a single chart together with nonuniform bars. Uniformity is the key to creating a noise free chart.
If you use the time it takes to create a volume bar in your strategy and you are wildly successfull . . . more power to you. I simply choose to keep my charts as clutter free as possible and to eliminate the need to interpret anything.
To each their own successes will be built based on the variability of their experiences.
Very interesting ProfLogic and would be quite interested in reading your material, please let us know where/when it would be available!
You state noise is created in time based charts due to instability. I can buy this as if you break down a 5m bar into 1m you'll likely see something completely different. But my question is how do you know that volume/share bars do not create a different form of 'noise'?
Because there isn't any inherent nonuniformity of the bars.
Minute Bar Charts contain a varying number of contracts or shares traded per bar.
Tick Bar Charts contain a varying number of contracts or shares traded per bar.
Range Bar Charts contain a varying number of contracts or shares traded per bar.
Equivolume Bar Charts contain a varying number of contracts or shares traded per bar.
Constant Volume/Share Bar Charts contain a nonvarying number of contracts or shares traded per bar.
OK, so uniformity allows for proper comparison, your comparing apples to apples with constant volume bars, is my thinking correct?
However does this create any new bias or noise. Or are your replacing one error with as yet (unknown to me) error?
Has creation of constant-volume bars resulted in stationary series or dependence, non-randomness, etc.? Empirically I assume you have found these type of bars to contain a greater signal and hence profitability?
Apologies for the threadjack but wanted to gain an understanding of the 'why'?
We are talking about noise . . .
Uniformity does allow for consistent comparison of price movement. No consistent comparison can be made with inconsistent charts.
I've studying the CVSB Charts since I came up with them and can not seem to find any inherent idiosyncrasies. That is not to say they don't exist but if you knew the time I've spent looking, you would have to imagine they don't exist. I'm really thorough.
And yes, since there is less distraction associated with these charts there is a higher profitability using them regardless of the individuals strategy. I've talked to many many traders that have switched to them in association with their own trading or investing environments and have told me they like them a lot better.
ProfLogic,
Are you also saying that tick-based bars aren't "consistent"? Do you believe tick-based bars are inferior to volume bars? If so, can you support that claim? Or, is it just that you're trading spot FX and there isn't sufficient time & sales data available for you to use tick bars?
P.S. knocks420, no problem w/the threadjack - I was done w/the topic on ET anyway.
-bluelou
Are you also saying that tick-based bars aren't "consistent"? Do you believe tick-based bars are inferior to volume bars? If so, can you support that claim? Or, is it just that you're trading spot FX and there isn't sufficient time & sales data available for you to use tick bars?
P.S. knocks420, no problem w/the threadjack - I was done w/the topic on ET anyway.
-bluelou
Wrong website. As a previous poster suggested, NuclearPhynance or Wilmott are more appropriate for this sort of thing.
The muppets on this site (my sincere apologies to Jim Henson's splendid creations), wouldn't know the difference between 'stationary' and 'stationery'
but you'll hear lots about ABC being 'perfect' (BTW - one must be extremely skeptical when you hear/read that word with relation to the markets) and XYZ being 'imperfect' - without any cold hard (read: objective) testing/data. But of course, would one expect anything different?
[Edit] Looks like you just posted on NP. [/Edit]
The muppets on this site (my sincere apologies to Jim Henson's splendid creations), wouldn't know the difference between 'stationary' and 'stationery'
but you'll hear lots about ABC being 'perfect' (BTW - one must be extremely skeptical when you hear/read that word with relation to the markets) and XYZ being 'imperfect' - without any cold hard (read: objective) testing/data. But of course, would one expect anything different?[Edit] Looks like you just posted on NP. [/Edit]
bluelou: either I'm wrong about the embedded structure in the noise or there is an embedded structure and I'm the only one who can see it.
Nuclear Phynance: there are some dynamics in the time series that your model is not addressing, and therefore is showing up in the residual.
Perhaps NP is nudging you towards the former, suggesting that you create a different mechanism to isolate the residual, and abandon efforts to understand the "noise" that is identified by your present mechanism. Or perhaps NP is wrong and you are right and there IS an embedded structure and you are the only one who can see it. If so, let us hope it is both tradeable and profitable.
Nuclear Phynance: there are some dynamics in the time series that your model is not addressing, and therefore is showing up in the residual.
Perhaps NP is nudging you towards the former, suggesting that you create a different mechanism to isolate the residual, and abandon efforts to understand the "noise" that is identified by your present mechanism. Or perhaps NP is wrong and you are right and there IS an embedded structure and you are the only one who can see it. If so, let us hope it is both tradeable and profitable.