12 September 2020 #BetweenTheLinesDotVote Analysis 6:00 And Still No Published Data - 3 Hours Late I'm pushing my memory, and while it may have happened in 2016, I don't recall Dornsife missing their 3:00 AM EST deadline. At least NOT by 3 hours! election.usc.edu
2) Here's how they state their practice: "The "Daybreak poll" is updated just after midnight every day of the week." Midnight in California, is 3:00 AM here on the East Coast. The term "just after" is loose, understandably so. But 3 hours is not an insignificant period of time.
3) Obviously, I may only speculate as to the cause. That said, I rather have to believe they're recalibrating something. An anomaly, a glitch in their calculations, the data not looking right, something. I do it all the time myself. I input the data, then I look at the chart.
4) The old idea that computers can't make mistakes is long, long past its sell-by date. Sure, I input my own data, but that's only because I'm not a programmer. Programming remains a human, error-riddled art too. So, you always have to check the output, carefully.
5) But by the time you're 3 hours late on a deadline, the least you could do would be to post a notification. I know, people often don't do things like that, but it's a shame. We all should. I'll keep checking through the day, and soon as they update I'll chart it and report.
Thread ends - for now - at #5.
6) I wrongly thought, idiot me, that I'd head back to bed for a bit more Saturday morning sleep. Alas. It was @KateScopelliti's comment yesterday that's been haunting me. She said, "6% margin for error?", after she glanced at this chart.
7) So, while Dornsife has still NOT published its daily data 3:40 minutes late now, I went and looked at the actual numbers of their margin. It looks to be above 5, but less that 5.5%. So, I decided to just round it down to 5%. And epiphany, check these numbers out!
8) Dornsife's last number for Biden was 52.69%. It's last for Trump, 41.08. Now, being a margin for error, what if it applied both ways? Could it be that Biden's number is more like 47.69%? Could it be that Trump's number is more like 46.08%? A margin for error allows each.
9) If one uses the worst case scenario approach, attacking Dornsife's integrity, one can see how Trump might, once again, be absolutely ahead of Biden, but they, like all the other outfits out there, are attempting to hide the truth, while pretending to provide the truth.
10) Linger with me on the topic of margins of error. Statistics is, by definition, NOT a precise art. It deals in probabilities, not certainties. It is, in that sense, the precise opposite of simple math. 2 + 2 = 4. That's simple math. Or, 100% vs 0%. Also simple math.
11) I always bristle when I see the normal 3, or 3.5% margin of error most polls claim. It sounds like they 97, or 96.5% sure they're right. It means NOTHING of the kind. The only thing it applies to is the numbers, the way they calculate their data.
12) What polling should do, is hold itself to account. Instead of a margin of error, it should publish a margin of confidence, if it must. My own preferred method is just to offer a prediction, and let it be right or wrong. I reject the entire margin of error concept, completely.
13) But it was Kate's observation that Dornsife is playing with a 5 - 6% margin for error. The moment she said that, I immediately felt profound regret. I watched their margin for error area in 2016, but never quite took it as far as her glancing analysis, yesterday.
14) I do cut myself this slack. Yesterday was the first time I ever charted it, and showed my chart to Kate. I should have charted it back in 2016, but didn't. Oh well. So it goes.
15) We just passed the 4 hours late mark, still no data update. Clearly, there are some big whigs in a meeting somewhere in Southern California, and they're obviously NOT releasing the update. Isn't that interesting? It sure is to me.
Thread ends - for now - at #15.
I just saw today's data come in!!! I'll graph it and return. Man do I love this stuff...
16) I'll speak to the 6 hours late below. Let's focus first on the overall chart here, and see how Dornsife represents the current state of the campaign.
17) Clearly, the Biden lead continues. We'll look at his, as well as Trump's individual charts shortly. I've added a new feature to my handwritten comments, the Insignificance Area's actual percentage. Today's, as you can see, was 5.26%. Who knew? I'll be studying this.
18) So, what's up with these 6 hours? On its face, it looks hinky. One has to imagine the computing power an outfit like Dornsife has at its ready disposal. We know their formulas were set in place long before the current campaign began. So, why be 6 hours late?
19) One might easily blame millenials. Imagine the team, they're in Southern California. And, well, the guy who's supposed to hit the "publish data" button really had a great party last night. He doesn't show up at midnight, right?
20) On a more serious note, being 6 hours late with data published daily, shows, if not lack of seriousness alone, very possibly worse. Unhappiness with the results and the need to massage the data. This is a blistering attack. I cannot prove it, and am loathe to make it.
21) Okay, I'm being too nice to myself. I confess. I kind of enjoy making it. Alright, I really enjoy it. And here's why. Back on 20 August, and since, I've been calling Dornsife out and am everything but to the point of rejecting their numbers this year. And then, 6 hours late?
22) You have to let me have my fun. Here I go. I picture this. There's a bunch of suits (but they're in their pajamas) meeting with a crew of scientists from midnight till 6:00 AM PST. The scientists have printouts of my analyses, and they're saying, we told you about this guy!
23) The yawning pajama wearing suits are saying, so what? Who cares? All we need is data showing how GIGANTIC Biden's lead is. What's the problem? Just add a few points to Biden's numbers and we're there. Okay? It's not quite as easy as that, the millenial scientists say.
24) It's 5:30 AM PST and some really big muckety muck comes in and says, show me the numbers. He glances. He says, here are the results. Writes them down. Heads back to his Belair mansion furious over having had to drive all the way over to USC.
25) You know, I've never been in those meetings. I honestly have no real idea how they run. I can only imagine. But you can tell, I'm not impressed by data that suddenly shows up, 6 hours late. You shouldn't be either. And one last point. No explanation? You have to be kidding.
26) Yesterday, I published honest guidance for the Biden campaign's consideration. I do not mess around about such things. Counsel is counsel, and I meant it. I won't do that again today, for them. But, let's look at his Dornsife numbers.
27) Lingering right there at his resistance line, one imagines that the COVID tapes from Woodward might be able to help us punch up above that line soon. If I'm a Biden campaign guy, I'm going, AWESOME! We hold out near our top, while Trump drops lower. We're cool!
28) At a deeper, more strategic level, if I believe these numbers, I'm going, this election is ours to lose. We have this. Our challenge now is simply to not mess up. America is done with Trump. MAGA is over. The Biden era is about to begin. No kidding, that's what I'd feel.
29) Now let's turn to Dornsife's Trump numbers.
30) If I'm the new guy, or one of the new guy's new guys, my heart is dropping right now. Who's this new guy? And what happened to Parscale? How did Parscale fall from grace? It's all so confusing. If I'm on Team Trump, and I believe these numbers, I'm not happy.
31) Okay, no need to overreact. It's just a tiny blip beneath what was a steep support line anyway. Support lines aren't perfect. We may well be well above the support line tomorrow. I might comfort myself that way. But then, there's this Scopelliti guy publishing this stuff.
32) Again, imagine the Dornsife people, and their LA Times masters, actually watching. Imagine them saying, WE MUST NOT LET TRUMP rise OVER his now happily broken support line! Besides, we do not enjoy showing up in our pajamas, having left our Belair pajama parties.
33) Let's turn serious again. Let's imagine that Dornsife is simply spot on. They had a glitch in their computers, who wouldn't be able to relate to that, and their data is 100% honest, right? What do we do with that?
34) First, we grant that Trump has, for the moment, lost the COVID issue. We grant that the Democrats knew they had Woodward in their corner, and they knew they'd once again take charge over the COVID conversation. More, they knew they'd land a body blow with Woodward.
35) So, we take a step back and ask ourselves, where are we really? The answer is, shockingly, exactly where we want ourselves to be. That is, from a position of unstoppable strength, we have regained the...wait for it...UNDERDOG'S ADVANTAGE.
36) We can always trot out the very truth, that we managed the COVID Crisis as well as it could possibly have been managed, and our man Trump always does so, at every possible moment. We're good there. But, better still, the entire polling apparatus is lying.
37) So, what we do is this. We imagine that we're about to fall lower, before we find our bottom, and rise again. But, we do NOT imagine that our bottom is actually below Biden's high. We imagine we're actually substantially ahead, and they're lying about every damned thing.
38) What's our stance, our posture? Underrated winners. We who come from a falsely perceived behind, to clobber our way to a massive, and once again, completely unpredicted victory. And we fight for every inch. What's an inch?
39) An inch is every number on every chart. If we break a support line, no matter how the great and the mighty may have corrupted that line, we'll care. We come out with confidence, but with 100% attention to detail. We fight for every vote, every last one of them.
40) And we pretend that Dornsife, and all the others might actually be right. We fight forward from there. It's understandable. The MAGA Movement has always been the movement of the Forgotten Americans. We remind them, they're about to be forgotten again.
Thread actually ends at #40.
Comments