Talk:Nationwide opinion polling for the 2020 Democratic Party presidential primaries/Archive 3

1326:, I like the idea of a lead column, it is just that I'd think that we need some way to indicate who is leading in the lead column, so we either would need to have it say who is leading and then by how much, use colors and have it in the candidate's colors, or have two seperate columns, one naming the leading candidate and then one with the value. The first and third would be weird for sorting, the second requires us to determine who gets what color. I'm for the third, unless you can find a better solution. 3401:

third party sources like fivethirtyeight.com - this can be seen in the jan 15-16 poll where Kloubachars and Yangs polling numbers are coming from a third party, while Steyer, as well as 'others' and 'undecided' are left blank. I don't think we should be listing polls which are not fully documented, so I suggest to delete the two polls and wait until they release either the full information on each candidate, or until they release the full dataset, like they have in their jan 8-9 poll.

1684:, the aggregate average was changed to exclude data from the New York Times, which used one less significant figures than the other sources. (The NYT has since been removed from this section of the article.) On December 19, I noticed that The Economist changed its averages and is now doing the same, rounding to the nearest whole number. (And yes, I do click on the face of each candidate. They honestly changed their rounding method last week.) Accordingly, I 31: 173:

of Democrats and Democratic-leaning independents, I decided to remove it in line with precedent here so far (where one sample tends to be kept in the mix while the others are done away with - the same goes for favourability). That said, if you'd like to set up a section for polls of the general population regarding the Democratic primary on this page, I'd fully support the poll's inclusion there.

3063: 2998:, thank you for your attention. I see the legends have been moved below the graph but they are now missing the main candidates whose colors are integrated into the aggregation table just below. Works for me, except perhaps the legends ought to be moved a pixel or two lower as "other candidates" currently overlaps with the first letters of months on the graphs axis. 2642: 1480:, I think it's the fact that the width is in px (when no units are involved I believe it goes to px by default), so on certain screens it wouldn't show up, try setting with a % and look at what happens. Although, in some cases I'm not sure how to translate code to wikipedia, since it isn't used in the same way all the time. 815:, thanks! That's interesting. I did not know it. Like half of all views to this page are mine, and this week my main concern was whether Gabbard will get the fourth poll she needs to qualify. I saw two new polls added but she wasn't listed. That's why I commented here. Thanks again, I will leave you all to it. Cheers! 1755:

Someone in the group asks how old the fossil is, and the docent replies that it is 1,000,014 years old. Someone else asks the docent how it can be calculated with such precision, and the docent replies, "When I started working here, I was told it was one million years old, and I've worked here for 14 years!"

3911:

The Economist should be restored and FiveThirtyEight should remain. They're attributed, and it can be noted that they adjust for different variables. The polls that feed into the numbers are transparent though, and the tenths place is less than relevant given the margin of error in primary polls. I'm

2288:

That seems like an odd case to make. This is essentially just a timeline with important dates added for context. Often times news reports say that a candidate dropped out "after the last debate" or something along those lines. SYNTH would be if we added in wikivoice that said that a candidate dropped

1833:

says that as of December 20th, Yang only has 1 qualifying debate (i.e. with 5% or more). However, looking at this page, it seems that he already has the 4 that he needs. Since the DNC hasn't declared him to qualify for the debates yet, I am assuming this page is the one that's off. Or am I missing

3664:

I understand there was good reason for including Bennet and Patrick in the February table but since they withdrew relatively early and the columns are populated with 0's and 1's only, could we remove them? Also, withdrawal of minor major candidates was omitted from the polling tables earlier. So, we

1229:

In the graph legend, I have moved "Others" to the bottom of the list, after active candidates and before withdrawn ones. It makes more sense this way, otherwise readers may be confused that "Others" line represents a sum of all other candidates besides the top 3 contenders (Biden/Sanders/Warren). Do

982:

I don't think Harris dropping really had that big of an impact on Klobuchar. She was already on a very slow incline, and every candidate has had their numbers go up ever so slightly when Harris dropped out (though this was basically negated by Bloomberg buying his way to where he is now in the polls

215:

No. There's already a standard in place, but if candidates' whose polling has been worsening fail to meet that standard within a couple of weeks, they will be removed from the December table. I fully expect there'll be some streamlining in a short while, but think it's best to exercise caution until

100:

Ok. I agree Gabbard's and Booker's columns should have been removed - there seemed to be a user consensus to remove Booker's October column if he did not average close to 3% within the month's first 15 polls, but that was not implemented. I'm not sure why Gabbard's December column was added to begin

84:

Hi everyone. Myself and some of the other people originally maintaining this page have been largely absent, and I'd like to thank everybody adding new polls to the article. However, we've had some instances where editors are adding columns for low polling candidates such as Tulsi Gabbard, which only

2660:

With Williamson no longer being listed in the polling average, I'd advocate revisiting this decision - the table would only grow by two columns if candidates were listed separately. Having tested this with a preview page, it seems the text does not wrap excessively in a table listing all candidates

645:

To clarify, it is probably a qualifying poll for the January debate, since it would have been for the December debate had it been released earlier. That said, the qualification rules have not yet been released by the DNC for the January debate, so it's impossible to say anything definitive. Between

244:

Is that an agreed standard? I recall having a discussion with others (in the archives) in which a heuristic of 3% in 3 of the last 15 polls was used (to decide whether or not Booker should be included in a specific month's polling table). In the archives, I've not found any user consensus for 5% as

172:

I concur. In some primaries, independents who do not register as Democrats will not be able to vote (that is why there is value in polling Democratic-leaning independents over independents as a whole), and that goes twice for registered Republicans, which the sample included. As it was not a sample

119:

a poll showing Gabbard at 9% among Likely Voters nationally in October. Why did you remove this poll, which was reliably sourced? Your edit summary states, "Democrat-leaning independents, where polled, are better than pure democrats if they are likely primary voters, but otherwise, we tend to stick

3161:

uses doesn't seem to allow us to line-item exclude items from the legend. However, we could deactivate the legend entirely and in its place put a simple wikitable, as below. This would remove duplicate entries on the legend like the two Bloombergs and the six debates. It also has the added benefit

2154:

to me, by implying that there's an important relationship between them, and that this relationship is the thing that deserves such a high degree of weight. Beyond that, it's quite awkward to read the graph and I don't think there's precedent in reliable sources or other Knowledge articles for this

926:

Ok, now that this merger of all of 2019 has since been reverted, my moot question now has meaning again! Let me ask the question again: Considering Harris was only in 3 polls for December (And dropped out only on December 3rd), would it make sense to leave here in the "other" section? O'Rourke was

2259:

I was able to get a single function to work, but it required adding a third data point for each debate (there are currently two, to define the bottom and top of the vertical line), and removing the interpolation feature (which is used to smooth the candidate's lines). I did not like the result.

1936:

Yeah, at first I made them all the same color, and I tried to have only one legend "Debates", but that prevented me from displaying the bars vertically. Once I was forced to list each debate on a separate legend, I figured that alternating colors would be easier to distinguish. No big deal either

684:

that's not what I was talking about at all. For the first and second debates, the Las Vegas Review-Journal and Reuters were approved pollsters, but they have not been approved since then. My point is that even if polls by a certain pollster have counted for previous debates, there is no guarantee

199:

I'm seeing a lot of lower-polling candidates clogging up the table and making it look clunky. Should there be some sort of higher standard implemented, say, at least 3% in 4 polls in a given month, or alternatively at least 4% in 2 polls that month? This would remove several more minor candidates

3435:

I use RVs where available (and where there aren't LVs). I edited out an 'adult' sample of Dems and independents to add in an 'RV' sample. The new releases (see Chris Kahn's twitter) omit the results for old candidates so should not be used where fuller datasets are also available. However, where

3400:

Looking at the Ipsos/Reuters polls from jan 15-16 and 22-23, it appears as if the data from the Ipson/Reuters polls has not yet been released. Instead the poll numbers are coming from an article written by Reuters, which is excluding some candidates, whose polling numbers need to be sourced from

1754:

Especially as an aggregation, I don't think it's all that necessary to round to the tenths place. The margin of error in these polls is well beyond 1%, let alone 0.1%. Reminds me of a joke in which a docent is taking a group on a tour of a museum, and they stop at a fossilized dinosaur skeleton.

2479:

I agree that a comprehensive aggregation table is not a good idea. It's less accessible on mobile devices, and doesn't tell the reader very much. It could be an even stricter threshold than 2%: both FiveThirtyEight and the Economist headline a smaller number of candidates on the pages for their

1800:

Adding a number (like the Economist aggregate) that has less significant digits than the rest of the numbers into the set of numbers for the average calculation reduces the precision of the average. Therefore all aggregates that have no digit behind the comma need to be removed from the average

3811:

Additionally, yesterday the Economist reported a 21% aggregate polling for Bloomberg although Bloomberg's highest result in Economist's poll table was 19%. And Economist uses a smaller set of polls than all other aggregators, which causes Economist's aggregate to skew our average towards their

3420:

No, it should not be excluded. But the original numbers should have been restored. Someone edited it so as to show only the numbers according to Democratic registered voters (n=428) when previously the numbers used the combined totals of registered Democratic and Dem-leaning independent voters

2673:

Not just yet. Booker just bowed out, and while Patrick may try to take the "black candidate" mantle from him, it's hard to see how Delaney and Bennett would remain in the race for much longer. In any case, candidates who never reached 1% should not be mentioned here at all, in my opinion. —

2324:

This page is about opinion polling for the primary, so it makes sense to have a graph of opinion polls. Adding debate lines stops it being a graph about opinion polls and makes it into a graph about the relationship between the debates and opinion polls, which isn't what the page is about.

732:

Considering Harris was only in 3 polls for December (And dropped out only on December 3rd), would it make sense to leave here in the "other" section? O'Rourke was not included in November and he dropped out November 1st. Otherwise there's going to be a column of "-" for almost the entire

1268:

There was one thing that I realized when looking at the polls. The lead column is missing. I'm probably biased (due to my work on other pages with a lead column) but I think it would be better if it was shown in its separate column. To give you an example here is what it looks like on

2949:

I think it looks fine in the example further up this talk page because it is just an image. The graph on the main article page has the image graph and text as separate elements which seems to be what's causing issues with smaller displays wanting to force the text to the next line.

3404:

I won't be removing the polls until I can gather that there is a consensus to do so, but keep in mind that it is only Ipsos/Reuters who there are such issues with. Normally when the full dataset is not released from another pollster, it is released within just a couple of days.

485:

Personally I think it looks better to have the tables merged: it comes at the cost of having a few low-polling candidates visible in months when they don't need to be, but it does provide consistency and make it easier to look at trends in polling over the whole year. Thoughts?

2607:

If I'd been editing on WP in 2015/2016, I would have supported drastically reducing the number of candidates displayed in the Republican nomination polling averages for the same reasons I support it here. It's worth noting, however, that consensus (?) has shifted since 2016.

3790: 1681: 272: 268: 264: 260: 64: 59: 3550:

My question is really about whether the functionality exists to color them this way without manually calculating and editing the colors in each cell (which would be a nightmare). I know there are heatmaps in articles about cities to show their temperatures, such as

3458:

Now the number of candidates has dropped down to 11, I think it'd make sense to remove the "others" column in the poll aggregation table and just have a column for everyone, and also possibly do the same in the February 2020 polling table when it's created. Thoughts?

1801:

calculation. The Economist aggregate has another big problem that reduces its usability for the average: The Economist uses a smaller set of polls than all other aggregates and thereby has a lower precision than the other aggregates. Onetwothreeip is correct, btw.

2494:

Agreed, and also I think the 2% threshold is the best option at the moment. When there are 7-8 candidates, we can include all remaining in the race, but at the moment, there are still too many candidates to include them all in a table as is currently being done.

2866:

In my mobile desktop view, legends now overlap the graph. I remember seeing some readjustment edits in the past few days. Did someone do something to mess with the relative positioning or is it just my device that's got this issue? Please take a look. Thanks!

2468:

added all the remaining candidates to the polling aggregation table. I think that's overkill, and we should keep following the usual consensus to list only candidates whose average is at 2.0% or above. What's the general feeling about that among editors? —

128:

were TODAY, which candidate would you support?" (emphasis mine). You've twice removed this sourced data, both times with misleading edit summaries. I'm asking you to restore it until a consensus can take place on this page, for the sake of transparency and

3142:

Just a thought, but it seems unnecessary to include debate numbers in the legend, since they're all the same color anyway. I would suggest just writing Debate after the legend - I don't know if that makes any substantial changes to the way the graph works

3054:

The following table uses the colors from the table on the primaries page; it also changes Harris' color to a darker shade to make it more distinct from Buttigieg's. For the reasoning behind the other color changes, see the talk page on the primaries page.

777:

Well, the table for all of 2019 has been merged now, that might change things. Harris and Beto are included while Gabbard is not. How did she miss out (criteria for inclusion)? She's still in the race and almost made it to the debates in December as well.

2180:

After a few weeks of this experiment, there seems to be diverse opinions on the appropriateness of including vertical lines for the debates. To avoid slow-moving edit wars, I'd like to gauge consensus. Here is what participating editors have said so far:

1448:

to represent success of sports teams. Unfortunately, it didn't go quite as well as I planned. The worst problem is that the graph does not display on mobile, even if I log out. Could someone give me an idea what's causing that particular issue? Thanks!

47: 17: 1121:

no, multiple polls from the same state count, as long as they are from different pollsters. Actually, I think for the later debates that duplicate early state polls actually would count, but there weren't any repeats during the qualification period.

2738:

Can we protect the page? We have IP users removing the debates from the graph (it's happened multiple times now) even though there is consensus to keep them. Furthermore, they don't seem to want to start a discussion and just keep removing them. -

1028: 3723:

I do not think the thing about 21% is a big problem (I am fairly sure that 538 also does things like that) but not showing the polls used and especially the removed decimals means it should be removed, since that calls its accuracy into question.

85:

clutters the article and extends the width of the table. This makes it increasingly hard for readers to navigate on mobile devices. I'd just like to urge everybody to be aware of this, and move data into the Other column where applicable. Cheers.

549:

I admire the effort and the courage to try something like this, but it's too much with displaying lower tier candidates. If another candidate rises into notability, a column would have to be added for the entire article, and maybe one removed.

3421:(n=552), as is typical. None of the polling aggregators that utilize Ipsos data — such as 538, The Economist, or 270ToWin — report the numbers as they are on the page right now; they use the previous version, with independents included. — 2560: 1638: 3162:

that the current legend is somewhat incorrectly titled, as it is called Candidates but also includes debates and a caucus and is overall very bloated. Thoughts on this? I'm also not great at making these so I'm sure it could be improved.

1024: 355:

I'm guessing it still counts as a reliable source since it's technically 538? That's what we did for info that came from Twitter during the Canadian election (i.e. if it came from a journalist from a trusted publication we used them). -

1443:

As some of you know, I originally created the opinion polling graph. It has since been implemented into other polling-related articles and improved by users far better at code than I am. I decided to use the same kind of graph at

685:

that they will count moving forward. It's been a rule since the first debate that multiple polls of the same geographic area by the same pollster cannot count (with the exception of the early state criterion in later debates).

3926:

Very wise decision to remove Economics aggregator poll from the article. It was a very strange polling without any logic behind the numbers! It seems they just put some random numbers for candidates based on their own desire.

1160:

There should be a way to correlate the polls on the debate page with this page since they are linked. Why does Amy Klobuchar have 8 qualifying debate polls and Andrew Yang only has 4 yet he is out polling her on a national

663:, I think it was a poll by NBC, but they didn't remove it. They just said that you can only count one poll sponsored by NBC (one campaign was counting two NBC polls at the time I think). IIRC, it was Andrew Yang's campaign. 1688:

The Economist from the average, using the same reasoning, but that has since been undone. I am ok with keeping the Economist in the average, but just wanted to make it known that this is a change from our past practice.

1176:

Because the four-poll threshold includes polls from the first four primary/caucus states (IA, NH, NV, and SC). Five of the eight qualifying polls for Klobuchar are from Iowa, where she is polling higher than Yang.

3766:

I should clarify, the methodology used to condense the polls into an average is not shown. Unless I'm missing an explanation somewhere, then I don't see how we can verify this. This also brings up issues with 538.

2661:

even if the page width is restricted to that of the graph. I would suggest that unless higher thresholds for inclusion than the current ones are supported, we now move to inclusion of all remaining candidates. —

1739:

Also, for the NYT part of the argument at the time was that they didn't update frequently enough, whereas the Economist does. I'm not voicing an opinion on the Economist here btw, just leaving this as a reminder.

348:

So, I was wondering what you do you think of including 538? They've been doing updates every few days for about a few weeks now. The thing though is that they currently do them through Nate Silver's Twitter :

3665:

should probably either add them back to earlier months for completeness or remove Bennet's and Patrick's withdrawal from February list. Are there good reasons for the inconsistency to be maintained? Regards!

972:

I think the Harris column should stay for December. Polls taken recently show that her campaign withdrawal boosted support for Amy Klobuchar, and we would lose that if we lumped her into the "other" votes. —

1309:, I'll be honest I didn't think of that when writing the paragraph. The reason they do that there is due to the fact that the polls have different parties, whereas here each candidate is in the Dem Primary. 1529:

Looking at the December polls, I would suggest adding Booker and Gabbard to the table, because they are credited with 3% or above in several polls, thus meeting our usual inclusion threshold. Thoughts? —

1064:

I am unable to identify the DNC qualifying debate polls. This page does attempt to highlight the qualifying polls but it doesn't add up to the totals on each monthly debate on the debate article page (

1946:

The lines make the graph harder to read. I doubt we need them at all. If consensus is that the lines should exist, I agree with the above comments that one color should be used for all of the debates.

3512:

That's a pretty good idea, right not the favorability section is a bit of a mess. I personally don't have time to work it, but you certainly have my full support to propose some good colors to use.

1554:

3% has been the consensus threshold for inclusion, although only when a candidate reaches 3% repeatedly in recent polls, so we avoid one-off upticks that indeed can be ascribed to error margins. —

3591:

I agree with you, I definitely prefer the blue shading, the other colors becomes a bit too intrusive in my opinion and it also complements the blue colors of the Democratic Party. I'm all for it!

1270: 702:, Oh, sorry about that. Thanks for the correction. I kind of was focused on the Canadian election for a while and my memory got fuzzy on the Democratic Primary. Man, do U.S. campaigns last long. 1973:

Sure, as I mentioned when creating them, it's experimental. That had been a requested feature in earlier discussions. If most editors think they're not great, I'd be happy with removing them. —

2393: 1596:

The main problem with an average is that at one time they could very well have an average above 3%, and then they have less than 3% in a new poll and they drop below an average of 3%.

2160:

The graph without them clearly and effectively shows one thing, which is the change in support of primary candidates, over time, according to opinion polls. Anything added detracts.

2935: 1274: 1041:

Looks good now, thanks for @all help. I have now transcluded the early-state polling graph into the main primary page, alongside the nationwide polling graph for easy comparison. —

3797:

to remove their aggregate as well. Eventually Economist was given a grace period to see whether they would return to reporting decimal places. They did not, which is why I removed

2047:

Perhaps we should make the lines more transparent, so they look like the 'others/undecided' line. That would allow the actual lines in the graph to be more visible by comparison.

646:

the second and third debates, Reuters and one other pollster (can't remember which at the moment) were removed from the list of approved pollsters, so anything is possible.

3895:, especially given the lack of preciseness (the removed decimals) and due to the lack of clarity of which polls are used and how they end up with that aggregation/average. 3940: 3921: 3906: 3876: 3861: 3835: 3821: 3776: 3761: 3747: 3733: 3717: 3674: 3654: 3625: 3600: 3585: 3564: 3531: 3506: 3482: 3467: 3445: 3430: 3414: 3389: 3354: 3337: 3319: 3301: 3171: 3152: 3137: 3123: 3106: 3091: 3041: 3026: 3007: 2989: 2960: 2926: 2909: 2892: 2876: 2841: 2824: 2797: 2780: 2748: 2721: 2695: 2678: 2668: 2653: 2630: 2602: 2588: 2571: 2548: 2534: 2517: 2489: 2473: 2445: 2436: 2422: 2355: 2334: 2319: 2302: 2283: 2269: 2254: 2234: 2225: 2207: 2169: 2139: 2124: 2087: 2073: 2056: 2042: 2022: 2008: 1994: 1977: 1968: 1941: 1931: 1917: 1899: 1880: 1861: 1846: 1810: 1795: 1781: 1764: 1749: 1726: 1698: 1669: 1658: 1649: 1622: 1605: 1591: 1558: 1549: 1534: 1516: 1489: 1471: 1425: 1390: 1373: 1338: 1318: 1301: 1286: 1249: 1234: 1209: 1186: 1170: 1148: 1131: 1109: 1092: 1077: 1045: 1035: 989: 977: 965: 948: 933: 905: 888: 867: 850: 828: 807: 794: 772: 753: 711: 694: 672: 655: 640: 623: 604: 588: 568: 559: 540: 526: 509: 494: 460: 428: 411: 397: 380: 365: 324: 298: 284: 254: 239: 225: 209: 182: 167: 142: 110: 94: 3473:

If all other candidates meet the current threshold by then, only Patrick and Bennet won't, meaning only one column would be added to the table. I support this measure. —

3436:

alternative datasets aren't available, incomplete data is better than no data, hence the use of these releases. I've jsut edited a January sample to account for this. —

2832:, I've went ahead and asked since vandalism has occurred -- again. The last time I asked they said it wasn't frequent enough, but now it almost happens every two days. 2146:

I'm a reader of this page rather than a regular editor of it, but I'd like to suggest that debates probably shouldn't be included in the main graph. The article isn't

2579:, I can understand the case for Steyer, since he's been on the debate stage recently and is likely to make the next one as well. But, I'm not sure about the others. 3826:

I agree with this. 538 also does some weird things by adjusting numbers to eliminate a so-called "house effect." I does not seem super objective to say the least.

3016:. I'm fine with it being below the graph but I think the main candidates should be listed there as well. I might try my hand at some adjustments in a little bit. 1613:, there's also going to be less polls for a while, which does not help. Although, it's worse for the state polls, I think we haven't had any since mid-November? 3031:

I added the other candidates to the legend. I’m sure it could be presented better but switching to div col seems to have fixed the overlapping issue anyway.

2934:, I see, I'll try to fix it. I have an idea, but I'm trying to figure it out (mix the bottom legend with the colour legend) to give you an example look at 3752:

What am I missing? I see the polls listed on the same page as the averages. There are currently ten polls shown up front, with a button to "See all". --

315:

I think we should at least include anyone who is still qualifying for debates or at least nearly qualifying in the sense that they're still relevant. --

3842:

It seems like we have a pretty solid consensus. Economist has some strange numbers, as well as a lack of decimals that makes our table not as precise.

3328:, but I haven't been able to get it to always draw the chart directly to the right of the graph. if anyone knows how, it would be much appreciated! 2274:

But what's the case that including debate lines isn't synthesis? This should be a policy- or source-driven discussion rather than an aesthetic one.

200:

from the table, and would reflect candidates who have struggled to gain momentum. This would include Cory Booker, Tulsi Gabbard, and Amy Klobuchar.

2400:

Maybe, although he never had the success that, for example, Kamala Harris and Beto O'Rourke did when they peaked in the nationwide opinion polls.

2758:, and it could easily happen now. In ideal circumstances for vandalism, it might even go unnoticed. Therefore, semi-protection is a good idea. 2378:

There's still no horizontal yellow bar for when Castro withdrew, the way there is for Harris and earlier drop-outs. Seems to be an oversight?

1083:

This is because repeat qualifying polls (polls from the same pollster covering the same geographic area) cannot count twice for a candidate.

2788:, vandalism was prominent on the Canadian election pages as well, semi-protection was used throughout the campaign and beforehand as well. 2385: 2230:

Looks like no consensus either way, and some people are editing the lines in and out every few days; that's why I called for more input. —

1838: 2013:

I would prefer if lines far in the future were not included such that the graph was not squished to the left. Other than that, I like it.

2754:

Also, supporters of particular candidates are likely to adjust the numbers to help themselves, or just for the sake of vandalism — this

799:

I included everyone who's included in the main graph of polls- using the same criteria. (namely, polling average over 3% at some point)

3642: 3613: 884: 824: 790: 768: 121: 927:

not included in November and he dropped out November 1st. Otherwise there's going to be a column of "-" for almost the entire month.

3936: 500:

The way the tables were merged makes it extremely difficult to keep track of which column is which as you scroll down the table. --

3383: 2818: 2774: 2624: 2511: 2416: 1962: 1720: 1585: 1510: 1465: 1445: 1419: 1367: 134: 2325:

Emphasising that relationship instead of just providing a visualisation of the subject of the article is at least a bit synthy.

3527: 2441:

Exactly. We have consensus to mention withdrawn candidates only if they already had an individual column in the poll tables. —

2130:

Seconded, the graph looks super cluttered now. If there can't be only one entry for the debates, the lines should be removed.

1100:, it's also 1 state poll per debate I believe (e.g. if you have an Iowa poll you can't use another Iowa poll). Or am I wrong? 1027:

is non-standard and rather illegible. Could a volunteer replace it with a similar chart to what we use here? Please reply at

579:

This is the first time NPR conducted a poll during this election cycle, is it going to be qualifying for the future debates?

371:

They plan on publishing this data on their website within the next week, so it's probably better to just wait until they do.

2733: