Its all about "tests"

[steve46]

And one last comment as regards bet sizing. As mentioned in one of my previous posts, there are a couple ways to bet this.

1. You put on your whole position size at once and look to get a specific pop (lets say 2 points).

2. You put on a token or "placeholder" position and look to add or "scale in" if and when you see price move to give you some breathing room. In this instance you can play for the bigger move and you can maintain a more frequent presence in the market. You assume that you will enter and exit several times before you get a "runner".

Which to choose. I use both. I watch price and try to analyze the day in terms of potential movement. I look at the previous day's action and at the action I see during the open and early in the present day.

Finally I look at the current setup. If for instance I am entering long and I see a pivot or possible test 3 points away, I may decide to put on my whole position at once, playing for the test, exiting before the pullback.

Good luck,
Steve

 

[BENG]

Thanks Steve, but the test on the regression channel depends on the period you set, and the difference can be big.

What's your setting on the 5min chart?

 

[steve46]

Well actually it does not..

You do need a specific minimum sample size (at least 100 data points)

The odds of price moving one way or another, depend on where along the distribution, your data point exists.

If for instance price tests the upper or lower channel of an LRC sized to 2 standard deviations, then the odds of continuation are only about 5%. Conversely the odds that price will retrace back into the channel are about 95%.

Most of the time, I use a channel sized to 2 standard deviations and I calc the odds on the fly.

 

[ER9]

steve, i've noticed your time template is set the same as mine (eSignal) from 6:30 to 13:00, 5minute charts.

it seems as though you draw your information, PP's, trend lines etc...from the data or price patterns only within this time period.

i have noticed days when the high or low we see with that time template may have been broken before or after those trading hours, obviously it will not show with that template setting though.

do you ever use those high's or lows as reference when drawing trend lines? or do you exclusivelly stick with what you see on your chart from 6:30-13:00?

 

[BENG]

Quote from steve46:
Well actually it does not..
You do need a specific minimum sample size (at least 100 data points)

I think it does, even a 200 period LRC will be different from the 100 period. I understand the your calculations on odds, but LRC also updates itself, so a single one big bar could change the LRC a lot. In hindsight, of course it does bounce the price nicely.

Some people think that 30 data points are the minimum requirement. How do you determine 100 data points are the minimum requirement? Any stats equation to back it up? I'm not trying to pick on you, I'm just trying to learn. I prefer to select a period based on something solid. If you tell me to pick on the period base on the cycle period, I will be fine with it.

TIA!

 

[steve46]

Somehow I am able to make it work.

I use several charts with candles showing daily, 60 min and 5 min time frames. I put in pivots and I use my own support and resistance analysis. During the summer months, I play for breakouts of the day's high or low, and as mentioned in my previous posts (and shown on the charts) tests of the previous day's high and low.

When I hear comments about "hindsight", what that tells me is that the person has trouble looking at a chart and "seeing" the possiblities for the next day. After years of doing this, I am able to open up a chart, and plan to execute a set of "opening plays" based on what I think price is likely to do. Quite often I am correct. I am pretty sure that every good discretionary trader does this or somthing similar so that they can visualize the market. It is a skill that can be learned.

My intention is not to provide a rule set for you to trade from, but to give you something to consider and build on, using our own ingenuity.

Good luck

Steve

edit:

by the way, when I say that some technique is a "skill that can be learned" what I mean to say is that I can help you with it if you are interested. Learning how to visualize the markets is not difficult. Simply post here to indicate your interest. Or not.

 

[JMowery1987]

Good stuff.

I actually was reading about this something when I first started trading and had the equidistant channel thing set on my charts when I first started, although was a bit rough on the 1 minute timeframe.

Have to explore it further if I was to do a systematic based approach to trading in the future on a bit longer timeframe.

 

[steve46]

Quote from ER9:

steve, i've noticed your time template is set the same as mine (eSignal) from 6:30 to 13:00, 5minute charts.

it seems as though you draw your information, PP's, trend lines etc...from the data or price patterns only within this time period.

i have noticed days when the high or low we see with that time template may have been broken before or after those trading hours, obviously it will not show with that template setting though.

do you ever use those high's or lows as reference when drawing trend lines? or do you exclusivelly stick with what you see on your chart from 6:30-13:00?

I switch to 24 hour settings shortly before the market opens and I continue that way until the first hour is completed. This allows me to see "tests" of the Globex session.

I prefer to use pivots taken from RTH only because others are doing the same thing.

I use highs and lows from RTH only.

Hope that helps.

Steve

 

[steve46]

As regards sample size, this is one of the basics of inferential statistics (what is a statistically significant sample size?). Rather than try to teach beginning statistics here, I will just say that I like to have 100 data points on my charts if possible. At the open it is not possible so I use a modified format (I change time frames and I use multiple time frames).

As I mentioned once before, while I am willing to point people in a productive direction, you will have to do some of the leg work yourselves.

Good luck,
Steve

 

[fader]

Quote from steve46:

If for instance price tests the upper or lower channel of an LRC sized to 2 standard deviations, then the odds of continuation are only about 5%. Conversely the odds that price will retrace back into the channel are about 95%.

hi Steve - i appreciate your insights, so i don't mean to criticize, but i don't think the above is correct.

95% refers to observations before their occurrence, i.e. before they get to 2 stdev... - this means that 95% of observations will be below 2 stdev, it doesn't mean that they will get to 2 and turn back... - for example, since 68% won't even get to 1 stdev, they can't possibly "retrace back" from 2 stdev - these odds say nothing about what happens after an observation gets to 2 stdev.. - normal distribution assumes independence of observations, not conditional probabilities.. - the odds of continuing / reversing after 2 stdev will be the same, assuming normal distribution, i.e. they will be consistent with the odds on a shorter timeframe.