Polling Behavior: A Look At The AP Poll And Michigan Hoops

Submitted by LSAClassOf2000 on January 28th, 2013 at 11:06 AM


Last week, fellow MGoBlogger Mmmm Hmmm posted a diary which discussed the behavior of the polls with relation to teams in the Big Ten, and reading it opened the door to an intriguing possibility. As there are sites now which in fact capture the voting patterns of the polls down to the individual voter, such as Pollspeak.com, it is possible to create a visual representation of how the polls react to a team on a weekly basis. Naturally, the team I chose as a test case for this was Michigan.

The AP poll is the most readily available when it comes to individual voter data, so that’s the one I chose for this exercise. I used Pollspeak.com to generate the team reports for each week from the preseason until the most recent data available and then downloaded the information for each week (it is available in CSV format from the site) into Excel. For our purposes, we shall leave individual names off the ballots and focus on how the rankings themselves behave.

I created two tables – one for the raw voter data, and another which counts by ranking how many voters placed us at a given position on their ballot. It was not until this point that I realized how convenient the COUNTIF function on Excel can be, incidentally. The resulting graphs for each week come out of the second table and allow for week-over-week comparison of the behavior of Michigan’s ranking per the opinion of the AP voters.


For the weekly data, the horizontal axis is the particular ranking, and the vertical axis is how many votes were cast at that ranking for Michigan.

 photo APBallotsWeek1_zps0042b329.jpg

 photo APBallotsWeek2_zps0474bd0f.jpg  photo APBallotsWeek3_zpsf5af7e58.jpg  photo APBallotsWeek4_zpsdebad89c.jpg  photo APBallotsWeek5_zps3dc483bb.jpg  photo APBallotsWeek6_zpsa8efe2d3.jpg  photo APBallotsWeek7_zpsf26ebef0.jpg  photo APBallotsWeek8_zpsc67d9777.jpg  photo APBallotsWeek9_zpse77735fd.jpg  photo APBallotsWeek10_zps5c875337.jpg  photo APBallotsWeek11_zps8e918cb4.jpg  photo APBallotsWeek12_zpsfcac7073.jpg  photo APBallotsWeek13_zps06f3329a.jpg  photo APBallotsWeekAvgRating13_zpse1f072d7.jpg


On each graph, as the week’s progress, if you follow the mode (the highest bar) and look in either direction, it is clear that consensus was beginning to build regarding Michigan’s ranking at different points. From Weeks 3 through 6, for example, we attained more third place votes than anything else, and with each week, more voters began to agree with the third place ranking. From weeks 7 through 10, you can see a similar trend for our second place ranking. At Week 11, you see the poll which was submitted the day after our loss to Ohio State, and with that game, the mode fell to fourth place and the general agreement among voters vanished. Week 12 shows the poll as of January 21st, coming off the win against Minnesota, and you will note that many forgave us – in a way – for the Ohio State loss in this poll as the mode is once again building around a second place ranking.

Now that the general format is laid out for tracking these (at least on this strange-looking Excel Workbook that I have created), it will be possible (and convenient) in the future to do comparisons between teams as well in addition to tracking individual teams, at least within the confines of the AP data.




January 28th, 2013 at 11:40 AM ^

I can't believe he would have any reason to do it, and I didn't know until downloading these reports that he had a vote in the AP balloting, but it is Paul Klee  of the Colorado Springs Gazette. It should be noted that sometimes voters don't always turn in their ballot in a timely fashion, or indeed often, or perhaps in Klee's case, ever.


January 28th, 2013 at 11:42 PM ^

It's a 65 member panel, correct?  And there were 65 first place votes.  So Paul Klee must be voting, and apparently irresponsibly.  Seems like he should be held accountable for something so obviously egregious. 

Is there anywhere that you can find a list of how everyone voted?


January 29th, 2013 at 12:22 AM ^

It might be better if all the graphs had the same scale on the y-axis. That would make comparisons between each week easier to see at a quick glance.

Edit: Should have said that this is already great work as it is. You have given a very good graphical representation of the voting patterns.  I really like it.  I just had a small suggestion that might be useful.


January 29th, 2013 at 7:14 AM ^

I'm not suggesting you do it... as what you've already done is great and it will be interesting to follow as we go thru the season. BUT, I would love to see this same approach applied to say other team/conference specific trends. In other words could something like this show bias (whether deserved or not) for/against certain teams or conferences? Would be interesting to see....


January 29th, 2013 at 10:35 AM ^

I tested the format below with a few teams this morning - Michigan, Indiana and Duke - to see if you could find this very thing. Comparative results are below (thumbnails - these are huge charts at full size):


 photo OverallPatternAPVotesMichigan_zps32aeaaa0.jpg  photo OverallPatternAPVotesIndiana_zps6c4dadb8.jpg  photo OverallPatternAPVotesDuke_zpsee3ec882.jpg

The skew for Duke and Indiana was decidedly towards #1 or #2, although Indiana spikes at #5 during a few weeks as well. I imagine it would be interesting to take the Top 10 in the final poll and do a season-long study of the AP voting using this format though.