Educator, Scientist. Sports & Political Data Analyst. Other social media: RealCarlAllen 📖 Book: *The Polls Weren't Wrong* out now
Ohio, USA
Joined January 2013
- Tweets68.6K
- Following5.4K
- Followers5.6K
- Likes56.4K
Pinned Tweet
I fully expect DMs, EMs, etc to blow up in the next 48 hrs
If I don't respond, don't worry. I know who has supported my work and I won't lose sight of that
If you'd like to join me in developing a set of unified scientific standards for this field, my book is available in bio
Simpson is a below average MLB hitter whose elite contact rate & speed make him serviceable
The fact that statistical illiterates use him (and/or Arraez) as their shining example of what Real Baseball looks like tells me they have no idea what it takes to score runs in baseball
Carl Allen retweeted
Found this piece by @RealCarlAllen informative: realcarlallen.substack.com/p…
Carl Allen retweeted
Reminder that comparing poll margin to election margin to determine "most accurate" pollster is statistical incompetence
It operationalizes the assumption that undecideds split evenly, and disregards the high probability that even a perfectly accurate poll won't hit the margin
One of the most unscientific things people like Nate Silver does is declare which pollsters are best based on how well their polls "predict the margin"
See below: even a PERFECTLY ACCURATE POLL can be graded as highly inaccurate, because of the assumption undecideds split evenly
"the probability of this outcome exists and can be quantified, but existing methods neither report nor estimate it"
@NateSilver538 (and @AAPOR as far as I know) literally declares with certainty that they know:
1) Which polls were more accurate
2) How accurate they were
Wrong
5 events with probability 83% happen
Silence
1 event with probability 17% happen
*Uncontrollable screeching*
I guess statistical illiteracy keeps me employed
"Poll error in Brazil"
No there wasn't.
Poll aggregators who didn't lie and dismiss undecideds had Bolsonaro around 48.
If you compare poll margin to election margin and call the discrepancy "poll error" you are fundamentally ignorant of what polls measure
Unless Republicans win by 18 nationwide, this poll is wrong
If you can explain why that is flawed logic, you understand why the assertion that undecideds will split evenly to major candidates (and any discrepancy is a poll error, making the poll wrong) is also flawed.
The direction of photographer error in recent years:
*2004: even bigger anti-US bias
*2008: huge anti-US bias
*2012: anti-US bias
*2016: unbiased
*2020: US bias
*2024: anti-US bias
I think the Dem margin in North Carolina will be larger than the Republican margin in Mississippi, and maybe Florida.
The hardest throwing starting pitcher in the history of baseball is also the best pitcher in baseball
No wonder boomers are so mad!
There are a lot of possibilities this election. I'll divide them into 4 ~equally likely bins
1) Rs hold Senate: narrow wins in notoriously purple states like..
Uhh
Texas, Ohio, Kansas, and Iowa
Leaves Rs taking a victory lap and Nate Silver loudly declaring "polls was wrong!"
3) Dems win all states they're favored in, plus KS and/or IA.
4) Dems win all of the above, plus some surprising state(s) like Nebraska, South Carolina in which case polling is dismissed as irredeemably broken
I personally think 1 is more likely than 3, but that doesn't matter
