100 of us get shipwrecked on an island,

[kster]

and we decide we need to elect a leader. We choose two candidates based on their experiences with survival techniques.

All 100 of us vote on a piece of paper and deposit it into a ballot barrel (yes barrel thats all we could find, were on a damn island). When we are all done voting, we all gather around and break open the ballot barrel.

If we only count 2% of the 100 ballots would we be able to determine who won the election?

I heard that it is possible, I'm just not sure :shrug:

As luck would have it one person out of the 100 is an e-voting salesman, and he has one of them fancy ballot scanners. Doh!

He says if we scan the ballots thru his machine that it would count them a hell of alot faster, and that after he scans and counts them, all we would have to do is count 2% of the scanned paper ballots by hand to determine if the machine counted total was accurate.

If we only count 2% of the 100 ballots would we be able to determine if the machine counted total was accurate?

I heard that it is possible, I'm just not sure :shrug:

[napi21]

1. 2% one time is not a big enough sample.

You would have to go to at least 20% on a number that small. If your group was 10,000, you could do it with 5%.

[kster]

3. Thanks........nt

[Votergater]

18. Supervisor Ion Sancho wants to manually recount 3% to 4%

of all his county's elections held on any Optical Scan system (increasing the percentage for elections which are closer).

[zeemike]

2. Here, have a nice glass of Cool Aid

That will set you strait.

[kster]

4. Well what do you think?

if you count 500 ballots out of 10,000 or 20 ballots out 100 is that enough? Just asking.

[zeemike]

10. No If you want to count something , you just count it all

It is not that hard we used to do it before computers and we still can now
I was just trying to be funny with the cool aid remark

[Land Shark]

5. Interesting example, i'll have to think about it, sounds like you're right

[Cos]

47. right, but irrelevant

See my comment below - this post is based on a fundamental misunderstanding of the audit requirement, it's purpose, and how it's done. Right or wrong, it's entirely irrelevant because it doesn't address anything the Holt bill does.

[Steve A Play]

6. Which e-voting company does he represent?

I hope it's not Sequoia!

Chicago ballot chaos: New computer vote machines malfunction,
unverifiable

By Christopher Bollyn
American Free Press

Apr 4, 2006, 14:33

COOK COUNTY, Illinois -- Chicago’s use of a flawed computerized voting
system operated by a privately held foreign company reveals how
meaningless and absurd the “democratic” process in America has become.

Having observed voting systems across Europe, from Serbia, Germany and
Estonia to Holland and France, this reporter has noted that the most
honest and transparent elections are also the most simple.

The more complicated methods of voting, such as the unverifiable
computerized voting systems widely used across the United States, lack
the most essential element of democratic elections -- transparency.


In one precinct on the Near South Side, for example, the Sequoia
optical scanner failed to register anything but Republican ballots.
Although “election officials” tried to repair the machine four times,
by the end of the day it had failed to register a single Democratic
ballot in a precinct in which some 86 percent of the voters are
Democrats.

<more>

http://onlinejournal.com/artman/publish/article_657.shtml

Why not just count all of the PAPER BALLOTS? E-Voting SUCKS, as does HR 550 and it's "paper audit trail" provision. They haven't been able to get it out of comittee since 2003 when it was introduced and the "improved" version doesn't stand a snowball's chance in hell of passing with this Congress. We need honest, transparent, BALLOT counts using PAPER BALLOTS IN ALL PRECINCTS, IN EVERY STATE! "We, the PEOPLE" deserve nothing less. The government should have absolutely NO SAY in how "we, the people" choose them.

Now let's see how the pro HR 550 crowd tries to spin that into a 'pro e-voting' stance! :)

On Edit: Added link

[kster]

7. Sequoia was the name of the ship......nt

[JimDandy]

8. A tie vote would be the one instance.

kster Query # 1:

"100 of us get shipwrecked on an island,and we decide we need to elect a leader. We choose two candidates based on their experiences with survival techniques.

All 100 of us vote on a piece of paper and deposit it into a ballot barrel (yes barrel thats all we could find, were on a damn island). When we are all done voting, we all gather around and break open the ballot barrel.

If we only count 2% of the 100 ballots would we be able to determine who won the election?

I heard that it is possible, I'm just not sure."


Answer to Query #1:

Because there is at least one instance in which you couldn't make a determination of the winner, then the answer must be no (79 vote for candidate #1 and 21 vote for candidate #2, but 2% of 100 is only two votes, which could result in a 50/50 split during the draw.)

But if you're really asking if there is at least one instance where you could make that determination then the answer is..... yes. A tie vote would do it. (50 vote for candidate #1 and 50 for candidate #2, and the two votes you drew were split between them, thereby accurately reflecting the tie in the contest.)

I'm still thinking about query #2, though.

[JimDandy]

9. Caveat to above:

If your query was of the "brain teaser" type then yes it is possible, in the realm of possibilities, to have an instance of tie winners. But if your goal is to know for sure who the winner is, as we need to do in real democratic elections, then no it's not possible to know for sure who the winner is.

[OnTheOtherHand]

11. actually, that is a terrible analogy

Statistical sampling works with large numbers. No one ever said that a sample of 2 out of 100 would be representative.

And by limiting yourself to 100 voters, you make the use of the scanner obviously stupid -- which, of course, you think it is. And you may be right. But the analogy is poor regardless.

[kster]

14. napi21 suggested 20% would that be enough?

[OnTheOtherHand]

20. probably not, but it depends

Since we're not actually facing a decision about 2% random audits of 200 people on a desert island, it is sort of hard to figure out what the question is. Sampling 40 ballots would offer some protection against egregious fraud. And hand-counting 40 ballots and then running those, and the other 160, through the scanner separately could afford some protection (check with yowsa-cubed on that).

If one is concerned about precinct counts being tampered with, a sample of 40 precincts can (depending on assumptions!) offer about 7 chances in 8 of detecting tampering in 5% of all precincts, more or less regardless of the number of precincts. That is pretty useful. (Now I want to go on and make the next five points, but I have to go talk about the Republican Revolution. Sigh.)

[unblock]

12. do i need to examine 2% of the molecules in the sea to know it's water?

a very very very tiny percentage of the molecules in the ocean would give clear and overwhelming evidence that it is made up vastly of h2o.

in fact, out of the gazillions of ocean molecules, if you randomly examined just a few thousand, you would have way more than enough to know that it's mostly water.


the biggest issue with RANDOM sampling is that the sample must be truly RANDOM. skew your sample and your results lose much of their meaning.

[OnTheOtherHand]

13. also sample _size_, if we are talking about Holt

The sample size matters more than the sample percentage. Corollary: a smaller race generally requires a larger-percentage sample (kster's example is extreme) for a given degree of confidence.

The Holt legislation raises lots of important issues. The OP, not so much.

[kster]

15. We would all agree that its h2o

its out in the open for all to see.

[Febble]

16. Short answer is no

And in any case, you need to specify the level of confidence you want. Do you want to be 99% confident, or are you happy with 50:50?

In other words, do you want do know whether, if you were to repeat the sampling procedure N times, what proportion of those times would you like to give you the same answer?

There is no magic percentage. There are magic sample sizes for a given required level of confidence and a given number of voters, but the percentage of the whole that will give you the right sample size will be bigger for smaller voter numbers.

[eomer]

17. Hey, Febble, how does margin of victory fit in?

If the machine count came out 51 to 49 then being off by 1 will change the result (from a win to a tie).

If the machine count came out 99 to 1 then it would have to be off by 49 to change the result.

Surely you need a bigger sample to detect being off by 1 than you do to detect being off by 49?

BTW, haven't talked to you in a long time so, :hi:


On edit: I think Amaryllis in a different thread is right that there are two different things being discussed. You're talking about an audit on the accuracy of the count while kster is talking about confidence that the winner is really the winner. I guess I was taking kster's approach in my question above but on second thought that's not what the audit is about is it? If you're auditing an official count of 70 to 30 you might conclude that the count is probably wrong but not have enough statistical power to conclude that the official winner is probably not the real winner. When you have a chance to comment I'm sure you will make this all crystal clear.

[Febble]

24. yes, I should also have specified

what margin you wanted to detect. Kster's question just asked us to say who won.

If the population is large, the standard formula for the standard error on the margin in your sample is sqrt(pq/N) where p is the proportion you get for one candidate, q is the proportion for the other (i.e. 1-p, ignoring third parties) and N is your sample size. To get confidence limits you multiply the standard error by the appropriate z score: e.g 1.96 for 95% confidence limits; 2.58 for 99 confidence limits.

So the size of the population doesn't matter, as long as it's large.

To put this in a relevant perspective: the penultimate National exit poll sample was 13047 (according to TIA). The total number of voters was 122,293,332. That means that 0.01% of voters were sampled. And yet, because the sample itself was large, the margin of error was very small.

And, as we famously know (TIA again), the exit poll projection at that point gave a result of 51% for Kerry. From this we can compute the standard error: sqrt(.51*.49) I make the confidence 99% interval 1%, we can conclude with 99% confidence that Kerry's true vote was between 50% and 52%. (I'm rounding here, BTW - for precise calculations see TIA).

However, if we take a state poll: in Ohio there were 5.6 million voters, and an exit poll sample of about 3000. So the percentage was 0.05%, bigger than for the National Poll. But because the sample size is smaller, the 99% confidence limits are nearer 2%.

Which all goes to show it is the sample size, not the percentage that matters. The smaller the population, the larger the percentage you will need to give you a given degree of confidence that the result you get in the poll is the true result.

Similarly with an audit: if you want to be sure that not more than, say, 1% of your precincts are corrupt, you need to calculate the sample size you need to give you your required degree of confidence that the true percentage of corrupt precincts is not more than 1%. Actually, I would put it rather differently, although the principle is the same: you need to calculate the sample size you need to conclude with, say 99% confidence, that a "clean" audit (no anomalies found)means not more than a given percentage of precincts are corrupt. It's the same theorem, but a slightly different formula, and the formula is different again if you have very small populations (say a single county) - the sample sizes you will need for small populations will be smaller (you can't have a sample of 500 precincts from a county with only 50) but the percentage you need will be a lot higher.


HOWEVER - just in case you think I've come over to the light side, ALL THIS DEPENDS ON RANDOM SAMPLING, as many DUers have noted with regard to audits. All these MoEs tell you is the kind of variance you'd get in your samples if you kept taking the same size sample at random, over and over again, from the same population.

So, for audits, it is ABSOLUTELY VITAL that the samples are truly random. If they are, and you have an adequate sample size, they should give you a good answer. For state level analysis 2% sounds pretty good - how many precincts are there in your average state? My calcs say that a clean audit from 500 precincts should enable you to conclude with 99% confidence that not more than 1% of precincts statewide are corrupt (but someone else might like to check).

HOWEVER, I can't end this post without saying, that however good the methodology of a poll, random sampling cannot be assumed unless the response rate is 100%, and even then, you'd have to be sure that your interviewers had truly sampled every Nth voter, and hadn't taken the Nth one that smiled back.

RANDOM SAMPLING and ADEQUATE SAMPLE SIZE are the two key elements, and sample size needs to be computed on the basis of:

• Size of juridiction to be audited


• Confidence level required


• Minimum level of corruption to be tolerated.

[OnTheOtherHand]

43. yes, as Amaryllis points out, at least two things

One is whether the winner won; the other is perhaps a bit more the reliability of the machines than the accuracy of the count, but I am too sleepy right now to settle on semantics. If we are auditing DREs, then we want every damn machine to yield a perfect count every time -- no matter what the winning margin is. If we are auditing op-scans, then maybe we don't insist on a perfect count, but darn close and not systematically biased -- no matter what the winning margin is.

[Amaryllis]

58. Eomer, thanks for noticing! I started realizing that people were assuming

a different purpose for Holt than what was intended, and also noticed that Holt himself had no category for an audit to determine the outcome of the elecion, but just for an audit to determine if machines were accurate, and a recount to detemine outcome of election. So, another category is an audit to determine outcome. People were arguing apples and oranges and not realizing the key is PURPOSE of a particular methodology, and you have to clarify this to have a meaningful discussion about audits.

[kster]

19. The confidence level that made sure that

100 people walked away at the end of the day, knowing with full confidence that they won or lost fair and square.

[Febble]

27. You would need to

keep pulling votes out of the barrel until you had 51 votes for one candidate. Then you could stop.

Your minimum sample size would be 51 - your maximum sample size would be 99.

But the Holt audit isn't going to tell you which candidate won more than 51% of the vote. The audit is going to tell you, with a given degree of confidence, whether fewer than a given percentage of precincts were totalled incorrectly. For 100% confidence that no precinct votes were totalled incorrectly, you need a 100% audit. But for 99% confidence that fewer than 1% were totalled incorrectly, you'd need, I make it, about 500 precincts out of a large jurisdiction. For a state, that might be around 2%. For a county, the sample size could be smaller, but the percentage would be much larger. For a hundred precincts, for that degree of confidence, you'd need to audit 99, i.e. 99% of the total.

[kster]

44. So the best thing to do is

throw the vote counting machine in the ocean? Because we have to hand count the ballots either way to get all the voters (winners and losers) 100% confidence that the election was fair?


Speaking of the Holt bill. A machine audit would give the salesman confidence that his machine performed correctly, and of course the people who voted for the guy that the machine may or may not have selected would be confident, but the people who lost will have 0 confidence.

We all should have confidence (winners and the losers) in the election don't you think.

[Febble]

45. Well, I have to admit

that I just hope to God that in the UK, they don't emerge from the ocean in the first place. So yes, I would say that would be the best thing - chuck 'em in. But it does seem as though your ballots are so horrendously complex (possibly because your history of mechanised voting has allowed them to become so) that Luddism isn't a serious option.

Next best, I'd say, is good optical scanners with proper audits, checked by hand-counts.

But if you CAN'T get rid of the DREs (and in principle they should be 100% accurate, more accurate than optical scanners) then the next best option has to be good audits. And it does strike me that a 2% statewide audit of DREs is a legitimate way to monitor the accuracy of the machines, as long as there are plenty of safeguards to make sure it really is an audit (random selection, custody of ballots, etc).

But of course, confidence in the outcome is absolutely fundamental to democracy. I couldn't agree more.

[Steve A Play]

21. If the government wants to "audit" my books

will they be satisfied with looking at 2% of my receipts if I choose them randomly?

If not, why not? :shrug:

That's what they think I should accept when I want to "audit" their books to assess their 'political capital'.

[Cos]

48. They don't audit everyone, do they?

The IRS doesn't have the resources to audit every single taxpayer. How do they enforce compliance? They audit a random sample of people. That means that you never know if you're the one who will be audited, which gives you some incentive to do the right thing.

If they choose to audit you, obviously, they will want to look at all of your numbers.

It would make no sense for the IRS to try to audit 2% of the all receipts, spread out randomly among all taxpayers. But it makes a lot of sense for them to choose a random sampling of people, and audit each of those people completely.


Similarly, the minimum audit called for in the Holt bill requires that at least 2% of precincts be randomly selected. But each of those selected precincts is then audited completely. They count all of the ballots, and compare to the machine count, to find out if the machines gave correct numbers.

It would make no sense to audit an election by randomly selecting 2% of all ballots and hand counting them, because no machine count of that same set of 2% of ballots went into the official tally. So you're auditing nothing by doing that. Fortunately, that idea has only sprung up in the misled minds of people criticising the Holt bill without reading about it in detail or thinking it through first. That idea is not in the Holt bill itself, because it would be preposterous, and Holt is a smart man.

[Steve A Play]

65. Thanks for making my case for me

The IRS doesn't have the resources to audit every single taxpayer. How do they enforce compliance? They audit a random sample of people. That means that you never know if you're the one who will be audited, which gives you some incentive to do the right thing.

Your statement shows the futility of this type of "random" audit! By your logic, if the IRS finds nothing wrong with the returns that they do audit, it's considered PROOF that no one else cheated on their taxes. :crazy:

The possibility of getting caught does not stop one from cheating unless the penalty for getting caught is severe enough to act as a disincentive. I find nothing in the Holt legislation that addresses stiffening penalties for someone who is caught. While there is a remedy suggested if problems are discovered, (full recount of all precincts) there is no real disincentive to cheat because the person who rigs an election can do so in a number of ways that would make it almost impossible to discover their identity.

It would make no sense to audit an election by randomly selecting 2% of all ballots and hand counting them, because no machine count of that same set of 2% of ballots went into the official tally. So you're auditing nothing by doing that.

No one that I'm aware of is suggesting any such thing. It's a 'straw man' argument proffered by you. Your next line shows just how weak your 'argument' really is. What I've maintained all along is that nothing short of hand counting all of the ballots will adequately protect our elections. This bill is dangerous in that it helps to further legitimize the use on 'non transparent' voting technologies and does so at the Federal level.

Fortunately, that idea has only sprung up in the misled minds of people criticising the Holt bill without reading about it in detail or thinking it through first.

When you can't argue your position effectively, make shit up about your opponent and hope that people believe you. That technique only serves to cheapen any real dialog and tells intelligent people more about you than it does about your opponents. It might work on a site like FreeRepublic, but it doesn't cut it here.

I've not only read about HR 550, I've read the text of the bill and discussed it with several attorneys and numerous election reform activists since it was first introduced in the House. I tend to pay close attention to what attorneys say when it comes to matters of law. It is said that a man who acts as his own attorney has a fool for a client. I tend to agree with that statement.

[jkd]

68. Something is better than nothing

If the audit shows the recorded vote of the precinct was an accurate reflection of the VVPBs, then the process worked well. It says nothing about any other precincts.
In my state, Utah, we just bought 10,000 new Diebolds. A 2% audit would require a random accounting of 200 machines. If any of those precincts failed to correlate the recorded vote to the VVPB’s then a red flag would go up. The county where in the failure occurred would then need to be 100% hand counted. The public would not stand for any stonewalling by the BOE, if this occurred. Well, maybe they would, but I trust they wouldn’t. I’m quite confident that any organized fraud in my state would be detected, but perhaps the 2% requirement could be tweaked if that’s found to be inadequate
Such a situation might require sequestering the machine in question and subjecting it to an ITA investigation. This would in turn make all the new Diebolds suspect. The lieutenant governor and the manufacturer hold all the cards in Utah at present. They would have to adhere to these federal guidelines if the bill becomes law.
Any move toward more transparency even if it leaves a lot to be desired, would be helpful. I believe the Holt bill would be better than what we have at present.

[EFerrari]

22. If 100 of us were stranded on an island, I hope FIVE

would still find a way to vote up the Daily thread.:hippie:

[FogerRox]

64. good advice sfexpat2000

[Chi]

23. You would have to be really, really carefull...

I hear some of the ballots might be reluctant, it's my understanding they could
be hiding at the bottom, or just out-of-reach... like that last sock in the dryer.

:evilgrin:

(hehe...I kill me sometimes)

[anaxarchos]

25. "You want fries with that?".... n/t

[Chi]

26. ROFL (Chi wipes the tears from his eyes) good timing Anax.

I needed that. 8)

How the heck are ya :hi:

[anaxarchos]

53. I'm good, man...

how about t'choo?

I don't get here very often, though...

Part of the problem is that my half-life as a Democrat isn't very long...

About the third time that some senator says something about "keeping our powder dry" or "do nothing and we win", and I start thinking about writing-in my next door neighbor again (who, btw, would make a terrific president).

So it goes...

Did you see the immigration rallies on CSPAN (or in person)?

Shit!

"the people united, will never be....", etc.

Nice to hear that cadence again (especially from a few hundred thousand voices).

[Febble]

28. Absolutely

Random sampling is absolutely essential.

Reluctant ballots could be a major problem.

[Chi]

29. And ya know what...

"Reluctant ballots could be a major problem."

Or it could be used as an unprovable excuse to cover miscount.
:shrug:

[Febble]

30. Unless the randomisation protocol

is absolutely transparent, and absolutely unpredictable, and custody of the ballots between election and audit is absolutely assured.

[Chi]

31. You be sure to let me know when that happens. ..n/t

[Febble]

32. You'll have to tell me

I won't be there.

[Chi]

33. If that happens...I'll be more than happy to keep you informed.

[Febble]

35. Well, I'll keep checking into DU

to see how it's going....

[Chi]

38. The stage is yours (Chi sweeps his arm and bows slightly)

out of respect for the other people who have to read this crap.

[Febble]

41. Huh?

[eomer]

34. Are you saying...

in case of rapture, this vehicle will be unmanned?

[Chi]

36. That's too funny...

I have a bumper sticker on my Jeep that reads...
"When the Rapture comes...can I have your car" 8)

Heya Eomer (waves)

[eomer]

37. Hey, Chi, I want that sticker!

But then would I have the guts to drive my car to my workplace full of Corporatists?

[Chi]

40. I wouldn't do it....

I wont put one on my work vehicle...I'm sure it would effect my income 8O

[Febble]

39. Nah.

While you guys are busy with your Rapture, we'll probably still be trying to prevent the replacement of our perfectly good, dirt cheap, completely transparent HCPB voting system by corruptible postal ballots and DREs.

[eomer]

42. I have to agree.

I'm with Ned Ludd when it comes to voting.

[Cos]

46. This post is a fundamental misunderstanding

Nobody, not Rush Holt, not even Diebold, has suggested determining the results of an election by counting 2% of the ballots.

Nobody has even suggested an audit that counts 2% of ballots, randomly selected. Such an audit would provide zero value. What would you do with the number you got from it? Nothing.

What the Holt bill does is require that at least 2% of precincts be fully hand counted, as an audit.

When you count all the ballots in a precinct, you can compare your result with the machine's result, and you can know whether the machine made a mistake (or had a bug, or was cracked and manipulated, or whatever).

To put it in terms of your story, let's say the e-voting salesmen offers us his machine which will count all the ballots, and we're not sure we trust it. We set up 6 barrels instead of just one, and each voter puts their ballot in one of the barrels. When we're done, we use his machine to count each barrel, separately, and then add the results together. We also roll a die, and based on that die roll, select one barrel and hand count it. If the hand count gives us a different number than the machine gave for that barrel, then we declare the machine buggy (or a fraud) and hand count the other five barrels too.

It's not perfect, but it's a significant improvement over just trusting the machine, isn't it?

It has nothing to do with whether counting some percentage of the ballots can give you the whole result. That's a straw man, and a pretty mind-boggling one at that.

[paineinthearse]

49. Cos

Welcome to the forum.

[kster]

50. How about an Exit poll we

could have two people stand at two of the barrels taking an Exit poll, what % would we have to poll? As an extra security measure.

[Cos]

51. Exit polls are a different matter entirely

Exit polls add no security. They're done for different purposes, primarily:
1. So news media can predict results before the official counts are in (understanding they may get it wrong).
2. To collect lots of other valuable data about voting patterns, such as the reasons people chose particular candidates, or correlations between demographic groups and support for candidates.

Even if you polled 100% of voters at all six barrels, that wouldn't add any security, because you can't tell if people lie. What they tell the pollster isn't the ballot that will be counted - what they wrote on the ballot is what matters.

[kster]

52. So if you ask 100

people who they voted for as they left the barrel and they said A and the machine counted total went for B, it wouldn't throw up any red flags to you?

[OnTheOtherHand]

59. are you aware that you changed the subject?

What's this about? Your OP seems to have nothing to do with Holt, and these questions seem to have nothing to do with the OP. What are you trying to accomplish on this thread?

Not speaking for cos, plenty of people have "red flags" about the 2004 exit polls, but the result hasn't changed. Would you say that exit polls added security to the 2004 election?

If you want to discuss whether exit polls could be used to add security to future elections, we can discuss that. But we have to be serious about it. Whether so-and-so or someone else would experience "red flags" is subjective, personalistic, and irrelevant to election integrity.

[Febble]

60. Sure it would throw up a red flag

But it still wouldn't tell me who won.

The fire alarm goes off in the hospital where I work the whole time. Usually it's someone having a secret fag. We still have to take it seriously.

Same with exit polls - when they go off, check for a fire AND check for a faulty alarm.

Sometimes you may find both. But if the alarm is faulty, remember it may not even tell you the right place to look for the fire, even if there is one.

[kster]

67. In America 2004, TIA any many others threw up that red flag

about the election, but no one (gov/media) took it seriously, no one even checked for election fraud or a mistake with the vote counting machines. Suspicious? To me it is.

Imagine that, if you will, a fire alarm goes off in your hospital and no one gets up. Every one just sits there, like nothing happened. Wouldn't you get suspicious?

[kster]

66. I just wanted to

know if it would throw up any red flags to you, if you don't want to answer thats ok.

[Land Shark]

55. Cos, with DREs one machine typically serves 5-10 precincts

this is one of the capabilities of dREs and why they have a big leg up if early voting is instituted (very limited numbers of polling places need to be able to have lotsa lotsa ballot styles, easy on DREs)

so these precinct audits wont' check a single DRE that the local election officials don't want checked, because all they have to do is use a common arrangement for DREs and let voters have access to more than one per precinct. :argh:

[Cos]

63. Yes, so?

I don't understand the point you're trying to make here, but whatever it is, I don't think it contradicts what I said. However many precincts the DRE serves, it will produce a separate count for each precinct. If you hand count the paper, you will produce a count for the precinct as well, which you can then compare with the DRE's count. That allows you to check whether the DRE counted the precinct correctly (compared to the paper record, which the voters had a chance to see and presumably complain about).

I understand there are many problems with this system:

1. Not all voters will check the paper record, so if the DRE sometimes produces paper votes that differ from the voter's intent, it's possible that will slip by with no complaints.
   
2. When voters do complain, sometimes local elections officials won't do anything about it, or won't report it, which increases the chances of a faulty paper record slipping by.
   
3. The paper record produced by most DREs is hard to hand count and takes a lot of time.
   

However, none of this makes the original poster's analogy even remotely valid.

The key here is the die roll: If every precinct must have the same chance of being selected, then local elections officials will not know which precincts will be audited. It doesn't matter if several precincts share the same DRE. Whichever precinct is selected will be recounted, and to the extent that the paper record is valid, any problems with the count the DRE reported for that precinct will be detected.

The original poster, however, posited a ridiculous situation in which you somehow "audit" randomly selected ballots from a common pool. If you do that, you don't get a complete count for any precinct, so there's nothing to compare your result to, and you haven't audited anything at all. It has zero value.

The original post's example is a straw man that bears no relation to the Holt bill. That's what I'm trying to explain.

Do you disagree with that?

[jkd]

62. audit or exit poll?

If kster’s OP has reference to the Holt Bill, then I certainly understand your frustration. That’s what I thought. You have a couple of posts on this thread that explain the audit intentions of the bill very well. Feeble and OTOH understand exit polling as well as anybody on this forum, if that’s what kster is concerned about.
I haven’t posted anything here for months, but I often stop by to read about the hot topics. I believe we should use paper ballots, but this legislation is a good stopgap. Welcome to DU, I appreciate your impute.

[FogerRox]

54. even on a deserted Island you would be required to recommend the ERD

[kster]

56. Already done,

Thanks for all the reminders I have to admit I need them every now and again.

[autorank]

57. This post is ALL TIME for brevity, wit, and persuasiveness.

I saw the subject field in the scan table and and though WTF is kster up to?

Then this. Outstanding.

I have a question, what's the tiebreaker on the 2 ballots? There are only three outcomes (presuming none of the ballots are "spoiled" by the elements)

--2 - 0 - not representative for obvious reasons
--1 - 1 - fine but the result is just a coin flip; may or may not be representative
--0 - 2 - not representative for obvious reason

So I can't make the damn thing work. Lets recount 3 votes, or maybe all of them.

We have nothing else to do on this island.

[kster]

61. Why Thank you,

I'm just a regular voter trying to get answers.