Poll Rates

  • 2
  • Idea
  • Updated 3 months ago
This is an experiment, I want to know which days have the highest polls rate, I will keep an Excel from Today to next Thursdays to see if I can come to a conclusion.

It is possible that I repeat the experience for several weeks to validate my results.

I will only calculate votes that were cast during the first day of the poll otherwise, It is gonna ruin the results.

If anyone has a hypothesis on the results let me know.
Photo of cinephile

cinephile

  • 1368 Posts
  • 1762 Reply Likes

Posted 4 months ago

  • 2
Photo of cinephile

cinephile

  • 1368 Posts
  • 1762 Reply Likes
Friday: 274 votes


Photo of cinephile

cinephile

  • 1368 Posts
  • 1762 Reply Likes
Friday doesn't seem like a really good day, maybe because Face-Off isn't the most popular type of polls.

Only the poll of Ed Jones went above 100 votes

Poll: Face-Off: Riddlers of "Batman" that Never Were: 118 votes
Poll: Face-Off: 'Winter's Bone' vs. "Ozark": 83 votes
Poll: Face-Off: 'East of Eden' vs. 'Rebel Without a Cause' vs. 'Giant': 73 votes
(Edited)
Photo of Dan Dassow

Dan Dassow, Champion

  • 16885 Posts
  • 19180 Reply Likes
cinephile,

When the number of polls was under 4,000, I captured the number of votes per poll on a daily basis.

Out of curiosity, how are you capture the poll rates?
Photo of Ed Jones(XLIX)

Ed Jones(XLIX)

  • 23311 Posts
  • 27775 Reply Likes
How is he gonna really compare when there is an inconsistency in the particular days when only three polls vs. four polls go live?
Or how about when the Thursday polls go live at 7PM Pacific and Peter does the Friday Polls at 12 Midnight Friday/Saturday (29 hour spread) Or The Monday polls that are only 20 or 21 hours new?
Or is he using a specific 24hour period and what 24 hours are they?
(Edited)
Photo of cinephile

cinephile

  • 1368 Posts
  • 1762 Reply Likes
That is a good point Ed, I didn't about that my current method is to compare the number of votes for each poll between 2 days, I consider that a days is from 0:00am to 23:59pm. For now, I won't change my method of calculating because it would made the data even more inaccurate.

Second, Dan Dassow 

I'm collecting links to polls and then comparing the votes from each day. Obviously, It doesn't work if the Poll is inactive.
Photo of cinephile

cinephile

  • 1368 Posts
  • 1762 Reply Likes
I think that It would be more accurate to release the average number of votes for new polls per day. Because as Ed Jones points out there are only 3 polls released on Fridays versus 4 on the other days.

So far this is what the new Graph looks like, this represents the average number of votes by new polls for each day of the week.

Note: I will be counting Saturday's votes until 24h pm Eastern Time, but I can say that If the trend continues Saturday's poll rates will reach aproximately 139 average.

Photo of Ed Jones(XLIX)

Ed Jones(XLIX)

  • 23311 Posts
  • 27775 Reply Likes
cinephile please!
Drawing any conclusions on a small data sample is unwise.
Also Fridays sometimes has 4 polls. Also other days of the week may have 3 polls and not four.
True data will require a larger sample.
Say a minimum of one year.
:):)
Photo of cinephile

cinephile

  • 1368 Posts
  • 1762 Reply Likes
I could analyze larger-scale data, but for that, I would need a webscraper.

I send an e-mail yesterday to IMDb to know if I need or can have their written consent for the use of crawlers on the website for non-commercial use.

On a legal aspect, web scraping is really on the edge of what it is possible to do, and I want to be legally safe.

For now, that is the best that I can safely do, and the fact that there are 4 polls a day 3 polls a day, 2 polls a day, doesn't really matter since, as I said I change the data type from a straight number of votes.
Photo of Ed Jones(XLIX)

Ed Jones(XLIX)

  • 23311 Posts
  • 27775 Reply Likes
Because of the fact that there are 4 polls a day, 3 polls a day or 2 polls a day, does really matter since that condition exists, that is why a longer sample is needed (1 year) before you can extrapolate a definitive day that is dominant.
And I have observed the same thing  ElMo. Weekends are bad days for polls and weekdays are good.
(Edited)
Photo of cinephile

cinephile

  • 1368 Posts
  • 1762 Reply Likes
In my intro, I said:

"It is possible that I repeat the experience for several weeks to validate my results."

First, It can take a year, it doesn't really matter, I have been on IMDb for 1 Year and 7 Month so another year is not a problem, but as I said it is possible that I get permission to analyze data with a robot if that is the case, it will make possible to collect data from the older polls then I wouldn't be necessary to work on this for 1 year.

Second, It is possible that  Saturday and Sunday aren't the best days of the week for polls. Honesty, I don't know what would explain because during workdays people are way busier, but still, I will analyze the weekends and the Friday because that is only a hypothesis, I can't make any conclusion without solid data.
Photo of 15yearsIMDber aka ElMo

15yearsIMDber aka ElMo

  • 4470 Posts
  • 5347 Reply Likes
I can tell you one thing, it's not Saturday and it's not Sunday (Sunday is even worse than Saturday) It has been true ever since I started making polls so you can save yourself the weekend, the busiest day is a working day. 
Photo of cinephile

cinephile

  • 1368 Posts
  • 1762 Reply Likes


1-The spreadsheet is done for 365 days as Ed Jones suggested

2-I'm not displaying the spreadsheet and the graph for the average for now, because I got some comments that said that it was too early to post that.

3- Don't pay attention on the results for today, Sunday, Mar. 29th 2020 as the day is not over yet.

4- Tomorrow, I won't post the final results for Sunday, Mar. 29th 2020 because I will be distancing my post on this thread. I want this poll to be easy to understand, and it would become confusing if I continue to post there every day.

So the next results will be released on Friday, Apr. 3rd 2020.



The results for Saturday, Mar. 28th 2020 have been updated on Saturday, Mar. 29th at 23:59pm
Photo of cinephile

cinephile

  • 1368 Posts
  • 1762 Reply Likes


==WARNING==
 The results are mainly biased by some big polls

The new poll with the most vote in a single day is BonafideBoss's Best Psychopathy Film with 195 votes released on April 2, 2020.




  1. The experience will be prolonged for an undetermined amount of time since the results are too close to each other.
  2.  I will post the entire spreadsheet on google sheet in a few days


Photo of cinephile

cinephile

  • 1368 Posts
  • 1762 Reply Likes
This was published at 0:10 Eastern Time.
Photo of Ed Jones(XLIX)

Ed Jones(XLIX)

  • 23311 Posts
  • 27775 Reply Likes
Remember I said...........1 Year!
Then you'll have a decent sampling.
Photo of cinephile

cinephile

  • 1368 Posts
  • 1762 Reply Likes


In other words, I don't see in the future, but the spreadsheet (except for the colors) is done for 1 year.



(Edited)
Photo of ACT_1

ACT_1

  • 4681 Posts
  • 6308 Reply Likes
Photo of cinephile

Sorry 'Off Topic' here

Your new dancing photo ...
Have you seen this video (others by same dancer are on YT)
https://youtu.be/cn-NsWRtaSY?t=104
.

(Edited)
Photo of cinephile

cinephile

  • 1368 Posts
  • 1762 Reply Likes
cool video
Photo of ACT_1

ACT_1

  • 4681 Posts
  • 6308 Reply Likes
Source large
You could start a YouTube dance list ? ?

https://www.youtube.com/watch?v=ROXQmAgLk7k

Maybe not - not IMDb
.
(Edited)
Photo of cinephile

cinephile

  • 1368 Posts
  • 1762 Reply Likes


==WARNING==
 The results are mainly biased by some big polls

The new poll with the most vote in a single day is thenolanfan's Best Oscar-Winning Best Leading Actor of the 2010s with 207 votes released on April 5, 2020




1. Sorry if the quality of the images is not as good, I'm trying to resolve this issue right now.
2. I'm still trying to post to make the entire spreadsheet without revealing personal informations.
3. Do not pay attention to the Friday results, I have already put the numb of polls in the formula so this is why the average is so low. Otherwise, the average would be 101
4. Sorry for posting 10 min later than last week, I have been testing some stuff on Excel.

Photo of Ed Jones(XLIX)

Ed Jones(XLIX)

  • 23311 Posts
  • 27775 Reply Likes
==WARNING==
 The results are mainly biased by some big polls
I remember saying that you'll need at least a years worth of data for this to have any meaning?
Photo of cinephile

cinephile

  • 1368 Posts
  • 1762 Reply Likes
Better Images:


The titles have been shortened to fit the case, the shorter titles are at "11", the longer ones are at "8".  No favoritism intended only to fit the cells
Photo of cinephile

cinephile

  • 1368 Posts
  • 1762 Reply Likes
Ed, I'm a bit annoyed by your attitude right now. I don't know what you are trying to get. 

But I repeated more than once that the spreadsheet is done for 1 year, but I will still be posting weekly in this post. To make it a habitude for me, and for people who are interested.

If you don't want weekly results, you can come back in 1 year.
Photo of cinephile

cinephile

  • 1368 Posts
  • 1762 Reply Likes
I brought back the histogram showing results by day, I just noticed that the histogram that I posted 11 minutes ago was only listing numbers
Photo of cinephile

cinephile

  • 1368 Posts
  • 1762 Reply Likes
1) Next results Thursday, I didn't post last week's results because there was a good movie on the TV and I was too captivated, LOL.


2) Disappointing results. Presently, results seem to stabilize meaning that, the day (except for the weekend and Friday) doesn't seem to affect a poll's results.
Photo of cinephile

cinephile

  • 1368 Posts
  • 1762 Reply Likes


1) Sorry,  I missed:
April 15, 2020
April 18, 2020
April 20, 2020
April 22, 2020

I have been doing this for over one month, and I feel really exhausted. For this date, I haven't added the number of polls because I don't want to screw the average. However, I have included a list of the polls published.

2) The Good News!!!!
The Spreadsheet is available, I have activated the comments so if you want you can give me feedback there.


https://docs.google.com/spreadsheets/d/1WFzPP2rQVbJqlBqyG2Y2R4EcyPI_GAV8t2-yi0BGyX8/edit?usp=sharing
Photo of cinephile

cinephile

  • 1368 Posts
  • 1762 Reply Likes

 Poll Rates are up to date, I'm gonna try to update it daily on the Online Spreadsheet.

1) Data for Thursday aren't available for the same reason that I didn't post the graph yesterday (I was too busy)


Analysis:

Saturday and Sunday have the same average
Monday and Thursday have the same average




Good News:

I will be allowing more time for this. To provide a more detailed analysis.
(Edited)
Photo of cinephile

cinephile

  • 1368 Posts
  • 1762 Reply Likes
I have read about Statistical Data Analysis during the last week, and come to the conclusion that my Analysis could be more way more detailed than two simple phrases. So, I will be trying to optimize that in the next weeks.
Photo of Dan Dassow

Dan Dassow, Champion

  • 16885 Posts
  • 19180 Reply Likes
cinephile,

Simply comparing the averages (means) of two data sets is insufficient. The variability of the two data sets (standard deviation) may be large enough that the differences may not be statistically significant.

For additional information:
https://en.wikipedia.org/wiki/Paired_difference_test
https://en.wikipedia.org/wiki/Pairwise_comparison
Photo of cinephile

cinephile

  • 1368 Posts
  • 1762 Reply Likes
So what are you suggesting??
Photo of Dan Dassow

Dan Dassow, Champion

  • 16885 Posts
  • 19180 Reply Likes
1. Gather as many samples as practical to reduce the Standard Error.
SE  sigma  sqrtn

textSE = standard error of the sample
sigma = sample standard deviation
n = number of samples

How do you calculate standard error?
Estimate. Since the population standard deviation is seldom known, the standard error of the mean is usually estimated as the sample standard deviation divided by the square root of the sample size (assuming statistical independence of the values in the sample). n is the size (number of observations) of the sample.

2. Once you have a sufficient number of samples, you can use the T-test to compare sample means.
https://blog.minitab.com/blog/adventures-in-statistics-2/understanding-t-tests-1-sample-2-sample-and...

How To Perform T-Tests In Microsoft Excel
https://www.youtube.com/watch?v=q0ckcKsSPXU

You can use the T-Test compare each day of the week against another to determine whether any day has a significantly different average than the another. There are 21 pair-wise comparisons.