Pavel Atanasov

@PavelDAtanasov

Decision scientist working on prediction/forecasting, calibrated trust in algorithms as cofounder . & Good Judgment Project alum. Dad.

Brooklyn, USA
Vrijeme pridruživanja: rujan 2014.

Tweetovi

Blokirali ste korisnika/cu @PavelDAtanasov

Jeste li sigurni da želite vidjeti te tweetove? Time nećete deblokirati korisnika/cu @PavelDAtanasov

  1. Prikvačeni tweet
    9. ruj 2019.

    How can we help humans make accurate predictions on complex problems such as predicting clinical trial outcomes? Excited to announce new funding to (w. ), together with at JHU.

    Prikaži ovu nit
    Poništi
  2. prije 4 sata

    The latest / Register / Selzer poll has been, once again, eerily predictive of the results so far.

    Poništi
  3. proslijedio/la je Tweet
    prije 12 sati

    Estonia e-Residency add in London. Bold ad for a country. Nice,

    Poništi
  4. proslijedio/la je Tweet

    We're taking a different approach to our 2020 modeling BTW -- When back-tested for the 2016 presidential election, a forecasting model based on an MRP-injected polling aggregate achieves a better prediction (lower Brier, RMSE) for each state than an aggregator alone.

    Prikaži ovu nit
    Poništi
  5. proslijedio/la je Tweet
    2. velj

    *11 Battleground States NBC/WSJ Poll: Biden 49% (+5) Trump 44% Sanders 49% (+3) Trump 46% Trump 49% (+1) Warren 48% Trump 47% (+1) Buttigieg 46% *AZ, CO, FL, ME, MI, MN, NV, NH, NC, PA and WI

    Poništi
  6. 1. velj

    Passed legislation, judicial nominations, and cabinet appointments will likely be similar in the parallel worlds where Sanders vs. Biden is president. Given this, few Democrats, even very left-leaning ones, will find Sanders to have been 20% better president than Biden in 2024.

    Prikaži ovu nit
    Poništi
  7. 1. velj

    Given Sanders' disadvantage in v. Trump, he has to be about 20% better president than Biden in your subjective judgment, in order to make rational sense for you to vote for Sanders in the primary. On the subjective utility scale: Trump 0 Biden 1 Sanders 1.2+

    Prikaži ovu nit
    Poništi
  8. 1. velj

    So let's say, conservatively, that they have the following chances of winning general election if nominated: Biden 60% Sanders 50% The extra 10% for Biden is a conservative translation of poll differentials int probability of winning.

    Prikaži ovu nit
    Poništi
  9. 1. velj

    Consider this: Biden is +5.4 points v. Trump nationally. Sanders is +3.7 points v. Trump. Also, Biden outperforms in key states like PA.

    Prikaži ovu nit
    Poništi
  10. 1. velj

    Persuasive arguments that Democrats will be similar in terms of actual enacted policies. So electability should loom large in the eyes of Dem primary voters, policy preferences much less so.

    Prikaži ovu nit
    Poništi
  11. proslijedio/la je Tweet

    1/There are two purposes of meritocracy: A) allocation (send people to what they do best) B) incentives (make people try hard). I think meritocracy helps with both of these, but a lot less than we tend to think.

    Prikaži ovu nit
    Poništi
  12. proslijedio/la je Tweet
    31. sij

    Incredible de facto social science experiment. Collective honking resets red light countdown to more time. Will it reduce honking? I think so

    Poništi
  13. proslijedio/la je Tweet
    30. sij

    Pledged Delegates Estimate (1990 needed to win the nomination) Biden 1720 Sanders 1286 Warren 439 Buttigieg 340

    Prikaži ovu nit
    Poništi
  14. proslijedio/la je Tweet
    29. sij

    Causal identification of 's impact on prediction markets: - Nate Silver updates model to reflect previous error: - Prediction markets incorporate new forecast. They had previously mispriced the model error. h/t

    Poništi
  15. proslijedio/la je Tweet
    30. sij

    Read this timely piece in the by about the importance of data for making school funding fairer.

    Poništi
  16. proslijedio/la je Tweet

    Watch the tax rate drop from 70% to 23% for America's 400 richest households

    Poništi
  17. proslijedio/la je Tweet

    Hello everyone, Please keep this plot in mind before naively assuming that Trump's approval rating translates directly to his vote margin. Shown here: even on the generic POTUS ballot, that's not true. kthxbye

    Poništi
  18. proslijedio/la je Tweet
    29. sij

    Got analytics chops?! Let's see 'em. Registration dates in the people analytics + non-profit world: 1) student case competition with UNICEF data (Jan 31), and 2) dataviz competition with Doctors Without Borders (Feb 2). For in Philly Apr 2/3:

    Prikaži ovu nit
    Poništi
  19. proslijedio/la je Tweet

    “The reason the world seemed like a better place during your childhood is because you were a child.” - Jon Stewart

    Poništi
  20. proslijedio/la je Tweet
    24. sij

    This week at Davos, ⁦⁩ launched a call for billionaires to give while they live. 2,150 billionaires have $10t collective wealth, yet most give less than 1% a year. The most generous give 4% or more, and done well, that has huge impact

    Prikaži ovu nit
    Poništi
  21. 24. sij

    Harvesting the wisdom of crowds for election predictions using the Bayesian Truth Serum.

    Poništi

Čini se da učitavanje traje već neko vrijeme.

Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.

    Možda bi vam se svidjelo i ovo:

    ·