Tuesday, April 26, 2016

Northeast Tuesday

No live updates tonight, almost all the races are basically settled, and I have a gymnastics class to attend!

Predictions (no suprises):

  • Delaware
    • TRUMP
    • Likely Clinton
  • Maryland
    • TRUMP
    • Clinton
  • Connecticut
    • TRUMP
    • Clinton
  • Rhode Island
    • TRUMP
    • Toss-up
  • Pennsylvania
    • TRUMP
    • Clinton

If, when I get home around 11 (pacific time), any of the races are still in doubt I'll build a model and start updating for my fellow night owls. Otherwise, see you next time.

Tuesday, April 19, 2016

New York Tuesday

YUUUUUUGE nights for both Clinton and TRUMP. See you next tuesday!

TRUMP has a yuuuuuuuuuuuuuuuuuuge night

Call for Clinton, heading for blowouts.

Call for TRUMP obv

It feels like it's been forever since we had a primary! I fear this one will be over before it started (Clinton and TRUMP) but I modeled it anyway. Clinton and TRUMP are clearly expected to win tonight, so just eeking by clearly won't be enough. In order to beat expectations they need to have yuuuuuge nights. So in addition to the usual outputs, I added:

  1. Does Clinton win by >15%?
  2. Does TRUMP get >50% of the vote?

That should give us something to sweat, even if the actual winner is determined very early. Here's where we're starting:

As always, read more on methodology here, I use the 538 model as Bayesian priors when available, otherwise I use polling and wide wide priors (i.e. high uncertainty).

Tuesday, April 5, 2016

Wisconsin Tuesday

Call for Sanders too, that was a short night. Good night!


Call it for Cruz and (almost) Sanders


This might be an anticlimactic evening. MSNBC has already called it for Cruz and Sanders, my model isn't there yet but it's headed there.

For some reason Wisconsin gets it's own special Tuesday! Both the Democrats and Republicans held primaries in Wisconsin and only Wisconsin today. Polls close at 6pm.

Results should start rolling in soon; here's where we're starting out:

As always, read more on methodology here, I use the 538 model as Bayesian priors when available, otherwise I use polling and wide wide priors (i.e. high uncertainty).

Tuesday, March 22, 2016

Regular Tuesday

Alright Utah.

It's after midnight your time, how do you only have 12% of the vote in??? I'm not calling it (but the AP etc. have) on Utah, but I'm calling it a night on myself. Goodnight, thanks for following along.

Idaho - Call for Sanders
Of course, it seems like they releaed 100% of the votes all at once, so is call for Sanders even the right phrase?

Utah - Cruz now 100% to get all the delegates

Utah - Call for Cruz

It's interesting that the networks hav called for Sanders but not Cruz. Clearly I see it the opposite way.


Utah - First Results!


Waiting for Utah, This Sanders guy seems to think we have a rigged economy.

Arizona - Call for Clinton

Arizona - Call for TRUMP

Close to calling for Clinton too.


Added an output to test if a candidate gets >50% in Utah. Currently based on the previously discussed squishy prior odds, but should prove interesting as the night wears on. (if someone get 50% or more they get ALL the delegates, otherwise UT is proportional)

Pre-game Odds.

Caveat: Like I said, the priors are pretty squishy on account of fairly limited polling in some of these races, so these odds are more of a place to start before the real resuls roll in, than something you should be using for betting.

That said, if I had to pick an upset I'd go with Clinton in Utah, since it seems like there are a ton of expectations for Sanders on the basis of a single poll, but there's such a dearth of data it's hard to say.





A regular tuesday, but With Elections!

Maybe we could call this Cowboy Tuesday, or Mormon Tuesday?

We have

  • Idaho (Democrats only)
  • Utah
  • Arizona
Models are set and ready to go, all we need now are results. Polling data is sparse on some of these races, so the priors are a little uncertain, but all that will come to an end soon. Plus modeling the race beforehand is something lots of people do, it's the mid-race modeling you can only get here!

As always, read more on methodology here, I use the 538 model as Bayesian priors when available, otherwise I use polling and wide wide priors (i.e. high uncertainty).

All of these races have a strong favorite, so we could be in for an early night. But if there's a lesson to be learned from the election cycle so far, it's that things rarely go as expected.

Tuesday, March 15, 2016

Super Tuesday 2.0

Missouri - Call for Clinton

Sorry for the wrong call. Basically Clinton was winning St. Louis big, we took that into account, but then in the last 10% or so of votes in St. Louis she won huge, way bigger than she had been, and overtook Sanders.

Now we can say goodnight.

Missouri - Wow. St. Louis County out of nowhere broke for Clinton even more than it had been. Not only is the race way back under doubt but it looks as if Clinton is the significant favorite.

I'm gonna be up for a while, mostly to see how long it takes the networks to call MO and IL. But I'm calling it on posting for the evening. My first Super Tuesday thread was my most viewed post of all time, then my second Super Tuesday post became my most viewed post of all time, and now this post is my most viewed post of all time. These live model updates are intense, but it seems like people are enjoying them and that makes me happy. Thanks for sweating along with me.

Politico with a whole bunch new votes, including in some areas that were still outstanding, and .......... DRUM ROLL

Call Missouri for Sanders and TRUMP.

Dum dee dum

Missouri gonna keep us here a while I think

Call Illinois.

Illinois holding at 99.8% for Clinton. (7:59pm)

Cmon Illinois.

Illinois at 99.9% Clinton (8:46pm)

Missouri is just being stubborn at this point. But Illinois mayybe we're headed toward closing time?

After the furious action earlier tonight things continue to be fairly stagnant.

There's what's clearly an error in DuPage county reporting Says 0.3% of precincts have reported. Were that true there would be over 4M votes in a county with 1M people. Downsizing that county's importance gives a "winding down" view of Illinois.

Sanders hanging around

Illinois still a sweat?

Illinois winding down

Call Illinois for TRUMP

Illinois and Missouri. Allllmost ready for that call.

Missouri - are we heading toward a night of Clinton and TRUMP with a Kasich footnote?

Illinois - closing time on both sides?

Call Ohio for Kasich. Ran behind the networks on that one, where's my dataaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa

Not over for the Rs in Ohio. Quite...

Ohio new batch of votes came in, call for Clinton

Ohio - There are calls I'd like to make but we don't have quite enough of a sample of returns from different counties to say we're all the way there.


North Carolina - Call for TRUMP


Reminder that we do make calls, unless otherwise noted they're based on 100 simulations (gotta be quick!), so they really mean >99.5%


Missouri and Illinois Pregame


Florida - Call for TRUMP

What is the network's problem? Florida looks like settled sauce. I'm not sure what they're waiting for.

Pregame for Ohio and North Carolina

NC D - Clinton is expected to win easily, if that changes I'll start modeling.

Already looks settled for Ds, not sure what the networks are waiting on.

Here's the Pre-game for the Rs

With all the big states voting today, and all the delegates being awarded (many of them winner take all on the Republican side) I think this day truly deserves the name Super Tuesday 2.0.

As usual I'll be updating as we go, it's been an incredibly busy week but I have models for all 5 races for both D and R. I may not use them in some of the less suspenseful states (e.g. D for North Carolina).

Unlike last Tuesday, it's going to feel like basically all the polls close at once; I'll be updating as quickly a I can / as quickly as it's material.

We have results now
Florida (likely Clinton, TRUMP)

Polls close at 430 PDT
North Carolina (likely Clinton, TRUMP)
Ohio (D is uncertain, R is likely Kasich)

Polls close at 5 PDT

Both Toss-ups