For reference, 538's day before 3/15 Primary forecast

[morningfog]

Here are the "Highest Probability" Polls-Only Projected Results from 538 as of 3/14/16 10:20 am EST:

FL: H - 65.2, B - 31.9
IL: H - 54.8, B - 42.1
MO: H- 50.7, B - 44.8
NC: H - 60.9, B - 35.9
OH: H - 57.2, B - 40


I think the actual results will be pretty far off this, with the exception of FL.

http://projects.fivethirtyeight.com/election-2016/primary-forecast/florida-democratic/#polls-only

[onecaliberal]

1. I guess they don't learn.

[LonePirate]

5. Their forecasts/predictions are entirely dependent on polls they do not conduct.

They do not offer gut predictions without polling data to back it up. The site was last updated yesterday before a slew of new polls were released today/late yesterday.

They need to perform another update based on the latest polls.

Why do people have no or little idea how that site makes its forecasts/predictions?

[oberliner]

2. Which states do you think Bernie will win?

Ohio and Illinois?

[morningfog]

3. I think he will win MO, and possibly win OH and IL.

Those three will be very close and he could sweep them.

I think the NC results will be closer, although Hillary seems to have done well in early voting.

[oberliner]

13. Would that give him more delegates on the day?

Presumably Ohio, Illinois, and Missouri will be fairly close so that even a Bernie win would only net him a few more delegates than Hillary in those states. If Hillary wins Florida by a wide-ish margin and wins NC, she would most likely end up winning more delegates than Bernie even if he does take those three states that you mention.

I am not saying this to be disparaging - I am just thinking about how the media will spin things even with Bernie having a great day.

[morningfog]

14. I think the delegate net for the day will be to Hillary, probably an additional 20-30, best case.

Could be more. If less, that would be great.

But this is Hillary's last "good day" on the calendar for some time. From here, Bernie will be chipping away. There are several big states to come later (PA, NY, NJ, CA) which would hopefully be effected by his momentum and what the narrative should become.

[oberliner]

20. Agreed

I think Bernie has some very favorable states ahead of him, and all the Deep South states that favored Hillary have already voted.

If Bernie can actually beat Hillary in NY, where she was a reasonably popular Senator, then that would be a political earthquake.

[Attorney in Texas]

4. Nate has let his personal bias taint his predictions this cycle. He's great but he's consistently

overstated Hillary's support and has missed almost as many Republicans races as he's called correctly.

Unlike other analysts, he will figure this out and get back on track for the general election but he's burned much credibility on his primary failures this cycle.

[LonePirate]

7. He has illustrated no personal bias. He is simply using polling data that is monumentally wrong.

His predictions are entirely reliant on polling data which has been significantly wrong this year. He has not introduced any personal bias into his predictions. Disagreeing with what his numbers tell him/us is not a reason to accuse him of personal bias.

[Attorney in Texas]

10. No, if you listen to his weekly podcasts, he has definite biases and his predictions skew toward his

preferred candidates. Coincidence? Maybe. Maybe not.

[LonePirate]

11. His predictions are based entirely on his mathematical modeling. His numbers have no personal bias.

Why is this such a difficult concept for people to understand?

[Buzz Clik]

6. 538 is about to lose relevance. Along with everyone else, they tanked on MI. Unlike everyone else...

... they learned nothing.

It's part of OUR learning curve -- who to trust, who to dismiss.

Tomorrow: large victory margins for Clinton in FL and NC. Tight races in the other three, likely all three going to Sanders by slim margins. If Clinton pulls out any of IL, MO, or OH, it will be big for her and very bad for Sanders.

[LonePirate]

9. What exactly should they have learned?

They have a defined methodology for gathering and processing polling data which leads to their predictions.

Are they supposed to throw away their system and simply predict a Sanders victory even though the numbers they have do not suggest that?

The problem with Nate's model is the highly inaccurate polling this year - polling which he does not conduct. If the polls were closer to being right, then so would he.

[Buzz Clik]

16. The open primaries are creating havoc, and 538 is not compensating.

Rather than change their methodology and models to account for the dynamics in open primaries this year, they are clinging to their standard approach. Illinois is a perfect example -- a closed primary in which both PPP and YouGov have adjusted their approach and are now calling IL to be very very close. 538 is clinging to the old approach and gives Clinton a 98% chance of winning with a likely margin of 56-44.

Pollster has adapted and is weighting YouGov and PPP very heavily, calling it for Sanders by 3%.

538 can either adapt or perish.

[LonePirate]

19. He has to be careful not to overcompensate and double the open primary effect intro'ed by some polls

Nate's model may or may not be one that can be modified on short notice and still produce results which he feels are sound. Does he rely on the polling people to account for open vs. closed primary differences? If they do it, do they make that fact known? If Nate has to account for open vs. closed in his modeling instead of relying on pollsters, does he risk overcompensating and thus generating results which are wrong in the other direction?

I personally think it is too late in the primary season to make meaningful modeling and polling changes. I think any poll and any prediction based on polls could be on the money or way off and neither result would surprise me. This campaign season is simply too volatile (unstable, that is) to make any sort of meaningful predictions. I think we are now seeing one party's primaries and contenders now impacting the primaries of the other party which is an entirely new variable that people (pollsters) are not considering.

If either Dem candidate wins all five states tomorrow, the race is over for the loser. If results are mixed, then the race will continue forward for at least another month.

[Buzz Clik]

21. Excellent points.

1. If 538 doinks on IL (and I am 99% certain they will with their current numbers), then they need to fold their tent and wait for next time.
   
   
2. I'm banking on PPP and YouGov being dead on, and they have IL really close.
   
   
3. "we are now seeing one party's primaries and contenders now impacting the primaries of the other party which is an entirely new variable that people (pollsters) are not considering." Agreed! Some Dems (assuming HRC is a lock) voting "Not Trump"; Independents voting for Sanders; Republicans crossing over to tamper on the Dem side. Not easy to predict but possible to pick up in polling data.
   
   
4. "If either Dem candidate wins all five states tomorrow, the race is over for the loser." I cannot see how Sanders could win 5 or even 4; HRC probably has 3 locked up with only a slim chance at 5. Sanders is in deep trouble and must get OH, IL, and MO; 5% or more win in each of these are needed to pretend he has hope.

[JimDandy]

8. Thanks. Bookmarked for later comparison. n/t

[Land of Enchantment]

12. The polls have been off

because (I think) the polling methods are obsolete. Once again look at the low turnout of 2010, 2012 and 2014. All cycles had low voter turnout. If those people who had come of age to vote did not register or vote they do not show up on the radar and are not polled today. Have you ever seen a poll of 'New Voters'? All I have seen are the usual dinosaur methods of 'likely voters' for instance. Add in mobile phones and no polling of Independents and -VIOLA- Inaccurate polls. I have ZERO faith in the polls because they do not perform accurate sampling of any given population.

[Betty Karlson]

15. Bookmarking this, just for tomorrow's laughs.

[GreatGazoo]

17. I want to invite 538 to my next poker game

I could use the money....

According to our final polls-plus forecast, Hillary Clinton had a greater than 99% chance of winning the Michigan primary.

http://projects.fivethirtyeight.com/election-2016/primary-forecast/michigan-democratic/

[basselope]

18. The most recent polls showed ALL races much closer than this.

I feel like 538 isn't even trying anymore.

[Buns_of_Fire]

22. I'm sure he's trying to work out the disparity. But good luck with that.

Problem is, as an aggregator, he's running into one of the first laws of IT: Garbage In, Garbage Out. His stuff can be no more accurate than the polls he's using.

The only statistical bias I can see being introduced is WHICH polls he's using for his final numbers. But even at that, I'm sure he's using the ones that proved most accurate in the past -- but the old polling methods just aren't cutting it anymore.