Since the daily position polls and bullish-bearish polls have been offered, I have jotted down the daily numbers... couple comments about these polls:
1) The data measures sentiment, so the question is: how "extreme" is "extreme"?
2) These polls are not money-on-the-table indicators, they are votes which may or may not reflect the actual money-on-the-table status of the participants.
First the daily Bull-Bear poll since late 2003. The blue curve is a 5 day SMA of the (Bulls)/(Bulls + Bears) percentage
Click to view attachment
Next is the "Position Poll" results since its inception in the spring of 2004. Partial and Fully long- Partial and Fully short position votes are grouped as bullish or bearish positions respectively. A 5 day SMA of each group is derived, then the blue curve is determined by: Bull 5 day MA - Bear 5 Day MA
Click to view attachment
FWIW
Randy N.
Randy,
You never cease to amaze me.
That was simply amazing.
You never cease to amaze me.
That was simply amazing.
You may see more with a shorter ma ( 2 or 3 days) or with a ema
The sentiment of those self-motivated to participate is reflected in the TT polls perhaps indicates the intensity of that particular small group of individuals. Maybe that's useful in some respect in context with other concurrent polls, but in and of itself I have to take this into account:
source: http://www.ropercenter.uconn.edu/pom/polling101.html#scient
source: http://www.ropercenter.uconn.edu/pom/polling101.html#scient
I think the prolem with this study is how did the traders vote? Did he/she voted by their feelings or by their work?.....It is like if I give you $1000 bonus now, how do you feel? Also, the result can be influences by some traders in this board......
Paul/Speilchekr,
Take this data for FWIW, it is only presented for amusement and disregard it for any reason you see fit. In one respect it is "more credible" than LowRisk.com in that you can only vote once in Mark's poll, where at least I think, the LowRisk poll allows unlimited votes by individuals.
As far as sample size, of course the TT polls do not provide sufficient data points for statisitical significance, that is obvious. However, I recall Jason Goepfert saying the AAII polls have 100 votes in a busy week, and the more common vote count for AAII surveys is the neighborhood of 50 to 60 responses (maybe less, memory ain't what it used to be) in a week. Jason mined this information from talking with one of the AAII staff.
Soooooooo.... despite low sample spaces, the AAII surveys have provided some usefullness over the years... at least I constantly hear the survey referenced and over the past couple of decades does have a credible track record for sentiment analysts.
Sample space debates.... a topic you can get all (and more) than you want on John Bollinger's Markets List forum.
Randy N.
Take this data for FWIW, it is only presented for amusement and disregard it for any reason you see fit. In one respect it is "more credible" than LowRisk.com in that you can only vote once in Mark's poll, where at least I think, the LowRisk poll allows unlimited votes by individuals.
As far as sample size, of course the TT polls do not provide sufficient data points for statisitical significance, that is obvious. However, I recall Jason Goepfert saying the AAII polls have 100 votes in a busy week, and the more common vote count for AAII surveys is the neighborhood of 50 to 60 responses (maybe less, memory ain't what it used to be) in a week. Jason mined this information from talking with one of the AAII staff.
Soooooooo.... despite low sample spaces, the AAII surveys have provided some usefullness over the years... at least I constantly hear the survey referenced and over the past couple of decades does have a credible track record for sentiment analysts.
Sample space debates.... a topic you can get all (and more) than you want on John Bollinger's Markets List forum.
Randy N.
QUOTE (mortiz @ Sep 9 2006, 07:13 AM) 
Paul/Speilchekr,
Take this data for FWIW, it is only presented for amusement and disregard it for any reason you see fit. In one respect it is "more credible" than LowRisk.com in that you can only vote once in Mark's poll, where at least I think, the LowRisk poll allows unlimited votes by individuals.
As far as sample size, of course the TT polls do not provide sufficient data points for statisitical significance, that is obvious. However, I recall Jason Goepfert saying the AAII polls have 100 votes in a busy week, and the more common vote count for AAII surveys is the neighborhood of 50 to 60 responses (maybe less, memory ain't what it used to be) in a week. Jason mined this information from talking with one of the AAII staff.
Soooooooo.... despite low sample spaces, the AAII surveys have provided some usefullness over the years... at least I constantly hear the survey referenced and over the past couple of decades does have a credible track record for sentiment analysts.
Sample space debates.... a topic you can get all (and more) than you want on John Bollinger's Markets List forum.
Randy N.
Take this data for FWIW, it is only presented for amusement and disregard it for any reason you see fit. In one respect it is "more credible" than LowRisk.com in that you can only vote once in Mark's poll, where at least I think, the LowRisk poll allows unlimited votes by individuals.
As far as sample size, of course the TT polls do not provide sufficient data points for statisitical significance, that is obvious. However, I recall Jason Goepfert saying the AAII polls have 100 votes in a busy week, and the more common vote count for AAII surveys is the neighborhood of 50 to 60 responses (maybe less, memory ain't what it used to be) in a week. Jason mined this information from talking with one of the AAII staff.
Soooooooo.... despite low sample spaces, the AAII surveys have provided some usefullness over the years... at least I constantly hear the survey referenced and over the past couple of decades does have a credible track record for sentiment analysts.
Sample space debates.... a topic you can get all (and more) than you want on John Bollinger's Markets List forum.
Randy N.
Took another look at the data.....If you do a 5 period ma on you 5 day ma.....You will get the SPX.....
Interesting data and charts, Randy. Thanks for taking the time and effort to share that with us.
mss
mortiz,
Thanks for the munchies; good stuff.
Best,
Tom
Thanks for the munchies; good stuff.
Best,
Tom
Randy,
VERY nice work. One thing that you'll find interesting is looking at the Fully long or Fully short positions, especially with a moving average.
In the years that I've been doing the FF/WSS surveys, I've found that when the sample group is relatively stable, the statistical caveats are pretty meaningless. I look at the same group of voters week after week. They don't change and sans a couple vacations, the number of responses is rather stable. As such (and as Carl Swenlin will attest), I can glean a lot more out of the data than a standard statistical approach would be able to or would predict.
The FF board is similar, but there are some potential problems. One is the "hot hand 'guru'" problem, where one or more of our number who are particularly respected and or have been very right in public for a while sways a number of votes, perhaps making our poll "lean" more than it might otherwise. That said, it's also pretty possible that many in the segment of the market that FF represents are ALSO looking at the same things that make that "hot handed 'guru'" hold a given position. After all, there's very little that's new under the sun, and if someone is using something available to all in multiple TA packages or on multiple sites, then it doesn't really matter that greenie (e.g) has found the holy grail of trading and is sharing it on TT--some other geniuses are doing the same thing elsewhere. IOW, it's pretty likely that TT sentiment, at least when smoothed, fairly consistently measures a certain segment of sentiment.
I also think it has value to look at participation. The mere act of voting, say, "100% Long" says something about sentiment. When a lot of people are voting that way, it says even more. Also, when voting goes way down, it says something as well. What? Well, I have some theories, but I'm not 100% sure I'm reading it right.
Mark
VERY nice work. One thing that you'll find interesting is looking at the Fully long or Fully short positions, especially with a moving average.
In the years that I've been doing the FF/WSS surveys, I've found that when the sample group is relatively stable, the statistical caveats are pretty meaningless. I look at the same group of voters week after week. They don't change and sans a couple vacations, the number of responses is rather stable. As such (and as Carl Swenlin will attest), I can glean a lot more out of the data than a standard statistical approach would be able to or would predict.
The FF board is similar, but there are some potential problems. One is the "hot hand 'guru'" problem, where one or more of our number who are particularly respected and or have been very right in public for a while sways a number of votes, perhaps making our poll "lean" more than it might otherwise. That said, it's also pretty possible that many in the segment of the market that FF represents are ALSO looking at the same things that make that "hot handed 'guru'" hold a given position. After all, there's very little that's new under the sun, and if someone is using something available to all in multiple TA packages or on multiple sites, then it doesn't really matter that greenie (e.g) has found the holy grail of trading and is sharing it on TT--some other geniuses are doing the same thing elsewhere. IOW, it's pretty likely that TT sentiment, at least when smoothed, fairly consistently measures a certain segment of sentiment.
I also think it has value to look at participation. The mere act of voting, say, "100% Long" says something about sentiment. When a lot of people are voting that way, it says even more. Also, when voting goes way down, it says something as well. What? Well, I have some theories, but I'm not 100% sure I'm reading it right.
Mark
My concern is the small sample.
Imagine if you only had sentiment data from 1998-2000. Your picture would be very biased by one direction. Up.
Your date is from the great 2 years of the 4 year cycle. The bounce after the crash.
2 years from now we may know much more! Data over the 4 year cycle is needed.
Thanks for sharing.
Scott.
Imagine if you only had sentiment data from 1998-2000. Your picture would be very biased by one direction. Up.
Your date is from the great 2 years of the 4 year cycle. The bounce after the crash.
2 years from now we may know much more! Data over the 4 year cycle is needed.
Thanks for sharing.
Scott.
Actually I'm only half right.
I'd like more data over the whole cycle, but the best 2 years are ahead, not behind us!
Sorry.
Scott
I'd like more data over the whole cycle, but the best 2 years are ahead, not behind us!
Sorry.
Scott
Randy's original links here:
Since the daily position polls and bullish-bearish polls have been offered, I have jotted down the daily numbers... couple comments about these polls:
1) The data measures sentiment, so the question is: how "extreme" is "extreme"?
2) These polls are not money-on-the-table indicators, they are votes which may or may not reflect the actual money-on-the-table status of the participants.
First the daily Bull-Bear poll since late 2003. The blue curve is a 5 day SMA of the (Bulls)/(Bulls + Bears) percentage
Next is the "Position Poll" results since its inception in the spring of 2004. Partial and Fully long- Partial and Fully short position votes are grouped as bullish or bearish positions respectively. A 5 day SMA of each group is derived, then the blue curve is determined by: Bull 5 day MA - Bear 5 Day MA
FWIW
Randy N.
Since the daily position polls and bullish-bearish polls have been offered, I have jotted down the daily numbers... couple comments about these polls:
1) The data measures sentiment, so the question is: how "extreme" is "extreme"?
2) These polls are not money-on-the-table indicators, they are votes which may or may not reflect the actual money-on-the-table status of the participants.
First the daily Bull-Bear poll since late 2003. The blue curve is a 5 day SMA of the (Bulls)/(Bulls + Bears) percentage
Next is the "Position Poll" results since its inception in the spring of 2004. Partial and Fully long- Partial and Fully short position votes are grouped as bullish or bearish positions respectively. A 5 day SMA of each group is derived, then the blue curve is determined by: Bull 5 day MA - Bear 5 Day MA
FWIW
Randy N.