Nevada (R) Predictions!

Hi everyone,

Make your Nevada (R) predictions here. I’ll make a ‘highscores’ post on Feb. 24th (or when the final results are in).

The winner will be whoever gets the top 3 correct and among those (if more than one person gets the order right) who has the closest match for %s.

Honorable mentions for anyone who gets the top 3 (in order).

Note: I will be locking the thread at 5PM PST, 8PM EST.

You can click the subject link to view the predictions.

South Carolina and Nevada Highscores

Congratulations to everyone who made a prediction!

The work for this is included in the spreadsheet South Carolina Primary and Nevada Caucuses Predictions. If I’ve made any errors or overlooked anything, please let me know!

Highscores, Reps (Actual results: Trump 32.5%, Rubio 22.5%, Cruz 22.3%)

7 people got the correct top 3 in order. The winner was Jonah.

  1. Jonah (30-23-21, ave. % point error 1.43)
  2. Falcon (36-23-21, 1.77)
  3. Toby (32-21-18, 2.1)
  4. DominicBayer (28-25-22, 2.43)
  5. themiddlepolitical (29-21-19, 2.77)
  6. Luki (29-19-18, 3.77)
  7. Jonathan (30-17-16.5, 4.6)

For contrast, Nick Silver (Trump-Rubio-Cruz, 30.5-19.8-19.5, 2.5) would have made the highscore list, and come in 5th, after Jonah, Falcon, Toby, and DominicBayer.

Highscores, Dems (Actual results: Clinton 52.6%, Sanders 47.3%)

17 people called the correct winner. The winners were William and Kevin.

  • 1. William (53-47, 0.35), Kevin (53-47, 0.35)
  • 3. Aaron (52-48, 0.65), Toby (52-48, 0.65), Jesse (52-48, 0.65), Jonah (52-48, 0.65)
  • 7. Kevin (53-46, 0.85)
  • 8. Jonathan (51.5-48, 0.9)
  • 9. Dylan K. (51-49, ave. % point error 1.65), William Tench (51-49, 1.65), Luki (51-49, 1.65), Jacob (51-49, 1.65), Arjav Rawal (51-49, 1.65)
  • 14. Rophil (50-49, 2.15), DominicBayer (50-49, 2.15)
  • 16. Ted (50.4-49.6, 2.25)
  • 17. themiddlepolitical (55-45, 2.35)

For contrast, Nate Silver (52.4-46, 0.75) would have placed 7th here, after William, Kevin, Aaron, Toby, Jesse, and Jonah.

South Carolina (R) and Nevada (D) Predictions!

Hi everyone,

Make your South Carolina (R) and Nevada (D) predictions here – just one or both. I’ll make a ‘highscores’ post on Feb. 21st.

For Republican predictions, the winner will be whoever gets the top 3 correct (Rep) and among those (if more than one person gets the order right) who has the closest match for %s.  For Dems, the winner will be whoever gets 1st correct and among those (if more than one person gets the order right) who has the closest match for %s between the top 2.

Honorable mentions for anyone who gets the top 3 (in order) for Reps or 1st for Dems.

Note: I will be locking the thread at 10AM PST, 1PM EST on Feb. 20th.

You can click the subject link to view the predictions.

New Hampshire Highscores

Congratulations to everyone who made a prediction!

The work for this is included in the spreadsheet New Hampshire Primaries Predictions. If I’ve made any errors or overlooked anything, please let me know!

Highscores, Reps (Actual results: Trump 35.3%, Kasich 15.8%, Cruz 11.7%)

Only one person got the correct top 3 in order. Kevin correctly called Trump-Kasich-Cruz (26-20-18).

6 people got Trump-Kasich correct. Dylan and Aaron tied for first in terms of ave. % point error, at 2.55.

  • Dylan (31-15)
  • Aaron (31-15)
  • Rophil (31-17)
  • Jeff (27-16)
  • William (25-23)
  • Jesse (22-20)

Eric correctly called the top 3, but not in the right order (Trump-Cruz-Kasich), at 28-16-14.

3 people also called the correct winner in addition to the people above.

  • Chris (Trump-Rubio, 30-17)
  • Luki (Trump-Rubio, 24-21)
  • Nick (Trump-Rubio, 24.5-22.3)

For contrast, Nick Silver (Trump-Rubio-Kasich, 26.8-15.7-15.2) of 538 fame didn’t get the top 3 in order (Kevin beat him), didn’t get the top 2 right (Dylan, Aaron, Rophil, Jeff, William, and Jesse beat him), and didn’t get the top 3 even out of order (Eric beat him). He did correctly call a Trump win though, so how did he do on the spread between 1st and 2nd (probably the most significant number of the night)? He would have come fifth here, after Dylan, Rophil, Aaron, Chris.

Highscores, Dems (Actual results: Sanders 60.4%, Clinton 38%)

Everyone who made a prediction accurately predicted the winner. Listed in order of ave. % point error.

  1. Eric (59-40, 1.7)
  2. Chris (59-41, 2.2)
  3. Jesse (58-41.8, 3.1)
  4. Kevin (57-42, 3.7)
  5. Jonathan (56.5-43, 4.45)
  6. Jeff (56-44, 5.2)
  7. Nick (55-45, 6.2)
  8. Aaron (54-45, 6.7)
  9. Luki (54-46, 7.2)
  10. Rophil (53.2-46.8, 8)
  11. Dylan (53-47, 8.2)
  12. William (52-48, 9.2)

For contrast, Nate Silver (57.2-39.9, 2.55) would have placed 3rd here.

 

New Hampshire Predictions!

Hi everyone,

Make your New Hampshire predictions here – Rep, Dem, or both. I’ll make a ‘highscores’ post on Feb. 10th.

For Republican predictions, the winner will be whoever gets the top 3 correct (Rep) and among those (if more than one person gets the order right) who has the closest match for %s.  For Dems, the winner will be whoever gets 1st correct and among those (if more than one person gets the order right) who has the closest match for %s between the top 2.

Honorable mentions for anyone who gets the top 3 (in order) for Reps or 1st for Dems.

Note: I will be locking the thread at 2PM PST, 5PM EST.

Note: thread locked. You can see the predictions by clicking on the post link.

Models don’t matter

Models don’t matter. What matters are predictions. If someone has a model of something, but it doesn’t make any concrete predictions, it can’t be tested. It is therefore worthless.

It doesn’t matter if it’s a fancy model made by a bunch of smart people. It doesn’t matter if you run the model 10,000 times. It doesn’t matter if the output of the model makes nice graphs. It doesn’t matter if the person making the model made previous predictions correctly.

The scientific method, in large, is about figuring out how to test predictions that follow from models of reality. Developing a model is a preliminary step. Testing the model is much more important.

So, if a political pundit makes a mathematically explicit, complex-sounding model, but there are no concrete predictions it makes or ways it can be tested, that model is worthless.

Ever hear a political pundit who will make what sounds like a pretty straightforward prediction, but then hedges it with all sorts of qualifications that render it a non-prediction? These people are – let’s use a technical term – blowing smoke. As entertainment, I suppose, it’s fine. But anyone who actually wants to know what the outcome is going to be should stop listening to these people.

Similarly with data-based punditry. I think many people hear ‘data’ or ‘mathematical’ or ‘model’ and think there must be something important or right about what the pundit is saying. This is nonsense – it is a variation of ‘b.s. baffles brains’. The truth is, political punditry is filled with charlatans who don’t know what they’re talking about, but pretend they do (typically because they’re getting paid to pretend).

There are some things that are very difficult to predict – instead of pretending they have some kind of crystal ball (including giving very official sounding percentages to various outcomes, all of which are basically untestable), the people in question who are blowing smoke should just say ‘we don’t know’.

Iowa Highscores

Congratulations to everyone who made a prediction!

The work for this is included in the spreadsheet Iowa Caucuses Predictions. If I’ve made any errors or overlooked anything, please let me know!

To make the highscores, someone had to get the winner correct for Dems, or the top three in order for Reps.

Without further ado …

Highscores, Dems (using latest Des Moines Register tally, which could change if there is a recount)

Winner, prediction, absolute %s was … Nick. (50-48, for an average percentage point error of 0.86)

(Note: Dylan K. gave a prediction or 45-44-1, which is obviously a mistake. I am guessing he meant to say 50-49-1, in which case he would have had an average percentage point error of 0.36. If he has a political consultancy firm, you might want to invest. If he has an accounting firm, perhaps not.)

Winner, prediction, relative %s was … Rophil. (48-48, which was 0.29 from the actual spread of 0.29%) (average absolute percentage point error 1.72)

All the people who called the correct winner on the Dems side, ordered by ave. absolute percentage point error

  • Nick (0.86)
  • John Doe (1.72)
  • Rophil (1.72)
  • Dominic Bay (2.22)
  • Toby (2.22)
  • Robert (2.72)
  • William (2.72)
  • Jonathan (3.22)
  • Jesse (3.22)
  • Lukie (3.22)
  • Mitchell (3.36)
  • Aaron (3.36)
  • Dylan (4.36)
  • Dylan K. (5.22, but see note above)

(For comparison, the much lauded Nate Silver would have made this list (48.3-45.1), and had an ave. absolute percentage point error of 3.02, which would have gotten him in 8th here.)

Highscores, Reps

Winner, prediction, absolute %s was … Jesse. (27-25-17, for an average percentage point error of 2.48)

Winner, prediction, relative %s was … John Doe. (25-23-18, a spread of 7 which was 2.45 from the actual spread of 4.55 from 27.65-24.31-23.1) (average absolute percentage point error 3.02)

All the people who called the correct winner on the Reps side, ordered by ave. absolute percentage point error

  • Jesse (2.48)
  • Jacob (2.71)
  • John Doe (3.02)
  • Jonathan (4.24)

(For comparison, the much lauded Nate Silver didn’t get the right ordering (24.3-25.6-18.1, so wouldn’t have made the list), and had an ave. absolute percentage point error of 3.02, which would have gotten him tied for 3rd here.)