Cuyahoga Cty - 2000 Bush Tallies Used to Fake 2004 Tallies?

Were Bush's 2000 Vote Tallies Used to Fake 2004 Tallies?

First of all, I am not a statistician. I am a programmer. I've worked on projects for major corporations where I had to reverse-engineer calculations.

For the past month I've been analyzing presidential election results from Cuyahoga County, Ohio. I started by comparing precinct results from 2000 to 2004. Some where added and some were removed, so I narrowed my focus on precincts common to both years.

I also compared the results from 11/8/2004 with the results from 11/30/2004.

The first thing I noticed was that there were 46 precincts that recorded the same number of votes for George Bush in 2004 as in 2000, either in the preliminary or final results.

This compared with only 12 precincts that recorded the same number of votes for John Kerry in 2004 as for Al Gore in 2000.

I wondered if George Bush’s 2000 numbers for each precinct were somehow used as a benchmark for altering the results of 2004. The mission from the hacker’s perspective would be to ensure that Bush would maintain roughly the same level of support in 2004 as he had in 2000. Some randomness would have to be built in. Or perhaps some precincts with week results in 2000 would be given a little boost. Basically, the hacker would use a list of precincts and build a script that would apply a specific hack to each precinct.

I was able to come up with the following formula for 11/30/2004 results:

See the precinct pairs tab http://home.att.net/~sunburst59/cuyahoga_precinct_pairs.xls


1) Determine Bush 2000 votes. Column I.
2) Determine 2004 Total Counted Votes. Column H.
3) Determine 2004 Total Independent Votes. Column U.
4) Determine the randomizing factor to be applied. Column V.
5) Determine Kerry vote. Column R:
=ROUNDDOWN((1-ROUNDDOWN(I/H,2)-V)*H,0)
6) Determine Bush vote. Column M:
=H2-(R2+U2)

(A similar formula is used to calculate the 11/8/2004 results.)

The above formula allowed me to calculate Kerry’s and Bush’s 2004 totals for over 400 precincts using Bush’s 2000 numbers and a randomizing factor. Keep in mind, I’m not talking just about the preliminary results, or the final results. This calculation applied to both results.

I was able to build some sequential number generator to determine if the formula applied to a particular precinct. If a match was found, the highest and lowest factors were saved. At first I thought they might have some correlation to the number of independent votes or the number of undervotes, but I couldn’t find a pattern. Indeed, the factors seemed to be pretty random, as they should be.

Example:

1513 CLEVELAND 1M

In 2000, George Bush received 2 votes.

11/8/2004 Results: |Bush: 23 | Kerry: 434 | Independent: 6 | Total: 463
11/30/2004 Results: |Bush: 23 | Kerry: 438 | Independent: 6 | Total: 467

That’s quite an increase in votes for Bush. Over 10 times as many in 2004 as in 2000. Let’s see how it was done.

November 8 Results:

1) Determine Bush 2000 votes. 2
2) Determine 2004 Total Counted Votes. 463
3) Determine 2004 Total Independent Votes. 6
4) Determine the randomizing factor to be applied. .0618
5) Determine Kerry vote.
=ROUNDDOWN((1-ROUNDDOWN(2/463,2)- .0618)*463,0) Or 434
6) Determine Bush vote.
=463-(434+6) Or 23

November 30 Results:

1) Determine Bush 2000 votes. 2
2) Determine 2004 Total Counted Votes. 467
3) Determine 2004 Total Independent Votes. 6
4) Determine the randomizing factor to be applied. .0618
5) Determine Kerry vote.
=ROUNDDOWN((1-ROUNDDOWN(2/467,2)- .0618)*467,0) Or 438
6) Determine Bush vote.
=467-(438+6) Or 23

Interesting how even though 4 more votes were added to the tally, the calculation still came out exactly and applied the additional votes to Kerry.

Now perhaps at this point, you’re thinking this is some sort of parlor game and that if you throw in a factor you can make anything come out the way you want it to.

But let’s put ourselves in the shoes of the programmer who was writing the script to hack the vote. What is the next precinct on the list?

1514 CLEVELAND 1N

In 2000, George Bush only received 2 votes in this precinct as well.

11/8/2004 Results: |Bush: 32 | Kerry: 470 | Independent: 0 | Total: 502
11/30/2004 Results: |Bush: 32 | Kerry: 477 | Independent: 0 | Total: 509

Gee, what if the programmer used the same randomizing factor for two precincts in a row? Would anybody notice?

November 8 Results:

1) Determine Bush 2000 votes. 2
2) Determine 2004 Total Counted Votes. 502
3) Determine 2004 Total Independent Votes. 0
4) Determine the randomizing factor to be applied. .0618
5) Determine Kerry vote.
=ROUNDDOWN((1-ROUNDDOWN(2/502,2)- .0618)*502,0) Or 470
6) Determine Bush vote.
=502-(470+0) Or 32

November 30 Results:

1) Determine Bush 2000 votes. 2
2) Determine 2004 Total Counted Votes. 509
3) Determine 2004 Total Independent Votes. 0
4) Determine the randomizing factor to be applied. .0618
5) Determine Kerry vote.
=ROUNDDOWN((1-ROUNDDOWN(2/509,2)- .0618)*509,0) Or 477
6) Determine Bush vote.
=509-(477+0) Or 32

So he did repeat the randomizing factor.

And he did for these sets of precincts:

2007 CLEVELAND 6G Factor: 0.005
2008 CLEVELAND 6H

2404 CLEVELAND 10D Factor: 0.024
2405 CLEVELAND 10E

3643 CLEVELAND HTS 3-C Factor: 0.0267
3644 CLEVELAND HTS 3-D

3825 EAST CLEVELAND 2-E Factor: 0.0263
3826 EAST CLEVELAND 2-F

3828 EAST CLEVELAND 2-H Factor: 0.0241
3831 EAST CLEVELAND 3-A

3832 EAST CLEVELAND 3-B Factor: 0.0158
3833 EAST CLEVELAND 3-C
3834 EAST CLEVELAND 3-D (3 in a row)

4432 HIGHLAND HTS 3-B Factor: 0.068
4441 HIGHLAND HTS 4-A

5121 MAPLE HTS 2-A Factor: 0.046
5122 MAPLE HTS 2-B

5164 MAPLE HTS 6-D Factor: 0.026
5171 MAPLE HTS 7-A

5801 NORTH RANDALL A Factor: 0.0363
5802 NORTH RANDALL B

Not only is there a correlation between Bush's 2000 and 2004 totals, there is a correlation between adjacently listed precincts. Some of the sets share the same physical location. Some do not.

The calculations for all these sets of precincts are included in the spreadsheet.

This may not be a smoking gun, but I believe these are the fingerprints of the hacking of the vote in Cuyahoga County.

That's just too weird to have happened naturally...isn't it???

I think we need to keep this kicked for some of the more math-inclined DU'ers, quick questions for my own information,
a)how did you calculate the randomizing factor?
b)forgive the ignorance but its been a few years since stats...what exactly does the randomizing factor do to affect the overall results

I'm impressed with all the work you put into this, the spreadsheet will now be studied further.
peace

This was calculated by generating numbers sequentially in increments of .0001. Because of the rounding, this trial and error method was the best way. The code I wrote would check each number to see if it would properly calculate the totals for the precinct for both the initial and final tallies. It would save the highest and lowest working values. So there was no set value but really a range. It was a very small range. I noticed that the ranges overlapped for adjacent precincts. The randomizing factors are numbers that are common to the ranges I found for adjacent precincts.

The randomizing factor I think was used so that Bush's 2000 numbers would not repeat as much. I also think it was used to pad his numbers in certain precincts.

It's hard to tell exactly what was being done. The rounding actually used may have been slightly different. There may have been some logic that turned this hacking on and off depending on the actual results.

By the way, I tried this same technique using Gore 2000 numbers and Kerry 2004 numbers. I was only able to get the calc to work for about 14 precinct. But no two adjacent precincts had a common randomizing factor.

That would be a big help in computing the odds of this happening by chance. With the low and high bounds it becomes the horse-shoe problem--what are the odds that this thing would land on the other one.

--MarkusQ

I've looked a bit at some of the 2000 .vs 2004 numbers. One that struck me was in Warren County.

In 2000 the number of Registered voters was 96,536.

The number of people who voted were 69,078

Bush received 69.95% of votes (48,318 votes)
Gore received 27.71% of votes (19,142 votes)

**

In 2004 the number of registered voters was 125,165

The number of people who voted were 95,512

Bush received 71.23% of votes (68,035 votes)
Kerry receieved 27.27% of votes (26,043 votes)

isn't that a bit weird?

not to mention the number of people who voted in 2000 (69,078) and the number of people who voted for Bush in 2004 (68,035).

I'm *not* a math person by any means.... just find it odd... that's all.

it certainly seems odd in many ways

-- also --

in 2000 1,618 voted for others (1,067 votes went to Nadar alone)

in 2004 145 votes went to others (144 for Michael Anthony Peroutka alone (Nadar wasn't on the ticket))

dumbass post

....in a report on various issues that I just sent to Congressman Conyers, Mr Arnebeck, Mr Bonifaz and others.

Peace.

numbers make me blurry eyed but this makes sense. Very well done. Thanks.

You should be able to find out somewhere which ones were used.

This is fascinating BTW.

regards

rig, you might very likely double them up while working from a list of all the precincts.

Thanks so much for sharing it.

I'd love to know what the likelihood of this happening just by happenstance is! and WELCOME to DU!:hi:

how the preliminary and final results for Bush and Kerry are calculated - with the only "free" variables between preliminary and final votes being the total votes and the number of "independent" votes.

I wonder what the odds are of that being random?

I believe they are capable of doing anything!

Rove was taught by Atwater and they have been doing tricks like this for years.

In Vanity Fair it talks about the Rove election fixers not needing
directions from him. They know that they have his full support to do anything necessary to win and not get caught.


Dumb Question: Were there a lot more Democrats in Ohio registered in 2004 or Republicans?

a primary. But you can assume certain types of people are likely dem voters and the other type are likely GOP voters.

Based on that the GOP was very worried about new voter registrations in OH because the new voters registered were predominantly likely dem voters.

Hint: If you live someplace that is 95% or more white you are a likely GOP voter.

Have you tried this algorithm on other counties that were counted by the same company? Cuyahoga looks like Triad (punchcard)...

See this map:

http://verifiedvoting.org/verifier/map.php?&topic_string=5std&state=Ohio

I think Cuyahoga is ES&S.

Does anyone know which counties have precinct-level results posted for 2000 and 2004?

from Richard Hayes Phillips:

His website <http://www.northnet.org/minstrel> says:

There are 88 counties in Ohio. If you can help with statistical analysis
Contact: richardhayesphillips@yahoo.com

For other county data. I'm working with Rich.

I'm at work now. I'll have to follow up when I get home tonight.

Please PM me if you can. I would welcome precinct level data from 2000 and 2004 for counties other than Cuyahoga. I'll be happy to do the same analysis.

Is posted at http://abrij.org/~bri

1, oh,"adams",28026,"Election Systems and Software, Inc. (ES&S)","Punch Card"
1, oh,"allen",108241,"Election Systems and Software, Inc. (ES&S)","Optical Scan: Precinct-Based"
1, oh,"ashland",53749,"Election Systems and Software, Inc. (ES&S)","Optical Scan: Central Count"
1, oh,"ashtabula",103120,"Triad Governmental Systems, Inc.","Punch Card"
1, oh,"athens",64380,"Election Systems and Software, Inc. (ES&S)","Punch Card"
1, oh,"auglaize",46740,"Election Systems and Software, Inc. (ES&S)","E-Voting: Other"
1, oh,"belmont",69636,"Election Systems and Software, Inc. (ES&S)","Punch Card"
1, oh,"brown",43807,"Triad Governmental Systems, Inc.","Punch Card"
1, oh,"butler",343207,"Election Systems and Software, Inc. (ES&S)","Punch Card"
1, oh,"carroll",29599,"Triad Governmental Systems, Inc.","Punch Card"
1, oh,"champaign",39544,"Triad Governmental Systems, Inc.","Punch Card"
1, oh,"clark",143351,"Triad Governmental Systems, Inc.","Punch Card"
1, oh,"clermont",185799,"Election Systems and Software, Inc. (ES&S)","Optical Scan: Central Count"
1, oh,"clinton",41756,"Triad Governmental Systems, Inc.","Punch Card"
1, oh,"columbiana",111523,"Triad Governmental Systems, Inc.","Punch Card"
1, oh,"coshocton",37132,"Election Systems and Software, Inc. (ES&S)","Optical Scan: Central Count"
1, oh,"crawford",46091,"Election Systems and Software, Inc. (ES&S)","Punch Card"
1, oh,"cuyahoga",1363888,"Election Systems and Software, Inc. (ES&S)","Punch Card"
1, oh,"darke",52960,"Triad Governmental Systems, Inc.","Punch Card"
1, oh,"defiance",39054,"Triad Governmental Systems, Inc.","Punch Card"
1, oh,"delaware",132797,"Election Systems and Software, Inc. (ES&S)","Punch Card"
1, oh,"erie",78709,"Election Systems and Software, Inc. (ES&S)","Optical Scan: Central Count"
1, oh,"fairfield",132549,"Election Systems and Software, Inc. (ES&S)","Punch Card"
1, oh,"fayette",28158,"Election Systems and Software, Inc. (ES&S)","Punch Card"
1, oh,"franklin",1088944,"Danaher Controls (Danaher Guardian)","E-Voting: Other"
1, oh,"fulton",42446,"Triad Governmental Systems, Inc.","Punch Card"
1, oh,"gallia",31398,"Triad Governmental Systems, Inc.","Punch Card"
1, oh,"geauga",93941,"Election Systems and Software, Inc. (ES&S)","Optical Scan: Central Count"
1, oh,"greene",151257,"Triad Governmental Systems, Inc.","Punch Card"
1, oh,"guernsey",41362,"Triad Governmental Systems, Inc.","Punch Card"
1, oh,"hamilton",823472,"Election Systems and Software, Inc. (ES&S)","Punch Card"
1, oh,"hancock",73133,"Election Systems and Software, Inc. (ES&S)","Optical Scan: Central Count"
1, oh,"hardin",31608,"Diebold Election Systems","Optical Scan: Central Count"
1, oh,"harrison",15967,"Triad Governmental Systems, Inc.","Punch Card"
1, oh,"henry",29318,"Triad Governmental Systems, Inc.","Punch Card"
1, oh,"highland",41963,"Triad Governmental Systems, Inc.","Punch Card"
1, oh,"hocking",28644,"Triad Governmental Systems, Inc.","Punch Card"
1, oh,"holmes",40681,"Triad Governmental Systems, Inc.","Punch Card"
1, oh,"huron",60231,"unknown","Punch Card"
1, oh,"jackson",33074,"Triad Governmental Systems, Inc.","Punch Card"
1, oh,"jefferson",71888,"Election Systems and Software, Inc. (ES&S)","Punch Card"
1, oh,"knox",56930,"MicroVote General Corporation","E-Voting: Other"
1, oh,"lake",228878,"Sequoia Voting Systems, Inc.","E-Voting: Other"
1, oh,"lawrence",62550,"Election Systems and Software, Inc. (ES&S)","Punch Card"
1, oh,"licking",150634,"Election Systems and Software, Inc. (ES&S)","Punch Card"
1, oh,"logan",46411,"Triad Governmental Systems, Inc.","Punch Card"
1, oh,"lorain",291164,"Triad Governmental Systems, Inc.","Punch Card"
1, oh,"lucas",454216,"Diebold Election Systems","Optical Scan: Central Count"
1, oh,"madison",40624,"Triad Governmental Systems, Inc.","Punch Card"
1, oh,"mahoning",251660,"Election Systems and Software, Inc. (ES&S)","E-Voting: Touchscreen"
1, oh,"marion",66396,"Triad Governmental Systems, Inc.","Punch Card"
1, oh,"medina",161641,"Fidlar Doubleday","Punch Card"
1, oh,"meigs",23242,"Triad Governmental Systems, Inc.","Punch Card"
1, oh,"mercer",40933,"Triad Governmental Systems, Inc.","Punch Card"
1, oh,"miami",100230,"Election Systems and Software, Inc. (ES&S)","Optical Scan: Central Count"
1, oh,"monroe",14927,"Triad Governmental Systems, Inc.","Punch Card"
1, oh,"montgomery",552187,"Triad Governmental Systems, Inc.","Punch Card"
1, oh,"morgan",14843,"Triad Governmental Systems, Inc.","Punch Card"
1, oh,"morrow",33568,"Fidlar Doubleday","Punch Card"
1, oh,"muskingum",85423,"Triad Governmental Systems, Inc.","Punch Card"
1, oh,"noble",14054,"Triad Governmental Systems, Inc.","Punch Card"
1, oh,"ottawa",41192,"Election Systems and Software, Inc. (ES&S)","Optical Scan: Central Count"
1, oh,"paulding",19665,"Triad Governmental Systems, Inc.","Punch Card"
1, oh,"perry",35074,"Triad Governmental Systems, Inc.","Punch Card"
1, oh,"pickaway",51723,"MicroVote General Corporation","E-Voting: Other"
1, oh,"pike",28194,"Triad Governmental Systems, Inc.","Punch Card"
1, oh,"portage",154870,"Triad Governmental Systems, Inc.","Punch Card"
1, oh,"preble",42417,"Triad Governmental Systems, Inc.","Punch Card"
1, oh,"putnam",34754,"Triad Governmental Systems, Inc.","Punch Card"
1, oh,"richland",128267,"Election Systems and Software, Inc. (ES&S)","Punch Card"
1, oh,"ross",74424,"MicroVote General Corporation","E-Voting: Other"
1, oh,"sandusky",61753,"Election Systems and Software, Inc. (ES&S)","Optical Scan: Central Count"
1, oh,"scioto",77453,"Election Systems and Software, Inc. (ES&S)","Punch Card"
1, oh,"seneca",57734,"Triad Governmental Systems, Inc.","Punch Card"
1, oh,"shelby",48566,"Election Systems and Software, Inc. (ES&S)","Punch Card"
1, oh,"stark",377519,"Election Systems and Software, Inc. (ES&S)","Punch Card"
1, oh,"summit",546773,"Voting Technologies International","Punch Card"
1, oh,"trumbull",221785,"Election Systems and Software, Inc. (ES&S)","Punch Card"
1, oh,"tuscarawas",91706,"Election Systems and Software, Inc. (ES&S)","Punch Card"
1, oh,"union",43750,"Triad Governmental Systems, Inc.","Punch Card"
1, oh,"van wert",29277,"Triad Governmental Systems, Inc.","Punch Card"
1, oh,"vinton",13231,"Election Systems and Software, Inc. (ES&S)","Punch Card"
1, oh,"warren",181743,"Triad Governmental Systems, Inc.","Punch Card"
1, oh,"washington",62505,"Election Systems and Software, Inc. (ES&S)","Optical Scan: Central Count"
1, oh,"wayne",113121,"Election Systems and Software, Inc. (ES&S)","Punch Card"
1, oh,"williams",38802,"Election Systems and Software, Inc. (ES&S)","Punch Card"
1, oh,"wood",123020,"Triad Governmental Systems, Inc.","Punch Card"
1, oh,"wyandot",22826,"Election Systems and Software, Inc. (ES&S)","Punch Card"

2000 and 2004? Sunburst mentioned that some had been deleted and some added, but that he was comparing over 400 precincts that had stayed the same.

On edit: Spelling + Clarification

:kick:

the State of Ohio site has 2000 and 2004 numbers...

If you look on the Data tab of the spreadsheet you can see them all.

Let's say that the 2004 Bush-Kerry popular vote difference could be set to be 3% more than 2000 Bush-Gore. Or is it a lot less likely something like what you're suggesting could be accomplished on a national scale?

precincts that have the same randomization factor applied to them.

The fact that 46 out of 1400 precincts have the 'same' result is not slam dunk evidence -- the doublets and triplets are very interesting.

Of the 12 precincts that have the same vote for Kerry in '04 as for Gore in '00 -- are there any doublets or triplets?

:kick:

to exactly illustrate the phenomenon you are looking at:

This was noted in both Conyers letter to Blackwell (12/8) and in Moss v. Bush --
In Miami County (Concord Southwest precinct) voter turnout was a highly suspect and improbable 98.55%. In Concord South precinct, there was a highly improbable 94.27% voter turnout. Miami County election results indicated that 18,615 votes came in after 100% of the precincts had reported. It is statistically suspicious that the extra votes came in at essentially the same percentage for candidates Bush and Kerry both before and after the 18,615 votes were counted Senator Kerry had received 10,724 votes (33.92%) of the vote after 100% of the precincts had reported. After the additional 18,615 votes were added, his percentage remained 33.92%.

Also - this site may have some of the data that will be useful for you:
<http://www.jqjacobs.net/bush/xls/ohio.html>

:kick:

used? I doubt there is a tabulator for every precinct.

For the non-math inclined on this board, you are showing that the number of B* votes increased from 2000 to 2004 by a random percentage (in the very broadest sense - I know you give more detail).

But it's precisely the randomness of the random factor that's problematic. The random aspect means that the formula can be applied to virtually any set of numbers to produce a desired result.

The formula also presumes the final number of total votes are known - how would someone know that, to a reasonable degree of certainty, ahead of time? And applying it on a precinct by precinct basis, in real world terms, would be rather time-consuming, even if you have a program running to do it - you'd have to gather the knowns beforehand (2000 B* vote numbers, 2004 final vote numbers) and apply them accurately. You also run the risk of actually lowering the real number of 2004 B* votes by using the 2000 number and just applying a random variable to increase them.

Yes, the fact that there are doubles and triples in a row raises eyebrows - I'm sorry but I can't find the random factor on your spreadsheet, even in column V. Can you clarify where it is? I would want to know how the votes in these "duplicate" precincts are actually gathered and input, as that may have an impact.

Have you tried the formula in non-Dem precincts? How would it work in heavily Repug precincts - have you come across any negative random factor numbers?

Just playing devil's advocate, for discussion..... never give up!

both initial and final tallies in every precinct.

I think 59sunburst is on to something.

The same "adjustment factor" SOMETIMES works, but most of the time it doesn't. See his spreadsheet.

I think he's close to something, but not dead on.

--MarkusQ

http://home.att.net/~sunburst59/cuyahoga_precinct_pairs_with_factors.xls

Thanks for all your comments. I've updated the file with a new worksheet called "all factors" that has all calculated "randomizing factors".

1) It includes all precincts common to 2000 and 2004.

2) It includes low and high ranges of factors that can be used to calculate:

a) Initial Results only (cols. AC and AD)
b) Final Results only (cols. AE and AF)
c) Initial and Final Results (cols. AG and AH)

If you look at the new spreadsheet, you'll see the factors seem to be in a pretty random order. That's why the doubles and triples stand out. But also keep in mind that there are over 400 precincts that come out perfect for both tallies. They also seem to come in clusters.

No doubt, some of these can be attributed to chance. I reversed this calc and used Gore's 2000 numbers to try to predict. I was only able to get 14 precincts to come out perfect for both tallies. There were no adjacent precincts with overlapping ranges.

I welcome anyone to explain how over 400 precincts lining up to both tallies is in the realm of reasonable probability.

My gut tells me that a human hand was at work here.

I think one of the pitfalls is to assume that all precincts were treated the same way. The hacking that occurred may have been conditional. Some precincts may have been left alone. Perhaps those were the ones chosen for the recounts. It's difficult to know just what was done. But just because you can't explain how the burglar broke into the bank, doesn't mean you should ignore the mud prints near the safe.

Later this week I'll post a template that will allow you to plug in your own values and perform the calcs I did.

As soon as I can I will look at other counties.

Gotta get some rest.

Peace.

:thumbsup:

of the ranges overlapping. Thanks for the data.

--MarkusQ

I wonder if any of those 400 were picked for the 3 percent handcount criteria for full handcount. This looks like a muddy footprint to me. Has TiA seen this?

this thread.

http://www.democraticunderground.com/discuss/duboard.php?az=show_topic&forum=203&topic_id=196518#196533

Non-Votes and Cross-Voting in Cuyahoga County.

It has a very different analysis of Cuyahoga.

I have to say this is an amazing discovery! You are able to make extremely accurate predictions based on nothing more than 2000 totals and a random variable. But what if you had a lot more information to work with in forming your projections? You may not need the random variable.

Assume that you have the following information available for all elections in each precinct for the past 15-20 years:

-- # registered voters and their party affiliation
-- percent of registered voters of each party (D,R,I) that actually voted in each of the past elections
-- final results for each election

You could plot these points on a graph and project the future election outcomes within a highly reasonable MOE. If you had this data for all the precincts in every county in the state, you would know how many votes your software has to tip for your candidate to win the state.

If you also knew the recount rules you could predict which precincts are more or less likely to be involved in a hand recount. You could engineer your software to avoid hacking precincts that have a likelihood of being recounted. You could then distribute the votes that need to be hacked thinly over all the safe precincts.

I would think your projections could be accurate within one standard deviation. When you calculate this, you could multiply the standard deviation expected by a randomized value. This way you could minimize the number of votes flipped in each precinct, flip a different total in each precinct, and make it look random.

Can you think of a way to test this given information that is publicly available?

Thanks! Great work!

:thumbsup:

How do I "kick" something?

Does simply posting cause something to be kicked?

I am not a mathematician (I am a graduate of English law) but on another thread someone mentioned that some of * percentages were very close to the golden mean. So I did a little experimenting with dividing one Fibonacci (sp?) number by the next and got 34/55 = 0.6181818 etc. Now where have we seen 18181 before? I might hypothesise (and let me be the first to say I may be spouting crap) that there is some difficulty using truly random numbers. Perhaps MS Access doesn't have an easy-to-use RN generator, I don't know. So the hacker is chopping up bits of easily programmable, "random looking" number sequences.

Just my 2c worth.

Guess they got sloppy, huh.

I have been thinking about the doubled-up multipliers and have a theory that could explain why. Essentially, the hacker was not being negligent in leaving fingerprints. They were writing an algorithm to run in MS Access to change the tabulated results, each time by a random multiplier in the right range to give the order of "results" they wanted. They might have been able to write a much more sophisticated algorithm that would give random-looking multipliers leaving no fingerprint pattern. However, this would use far more computing power. It would probably crash Access. Repeated blue screen would very likely cause elections officials to give up and instead tally the results in a new file which would not have the hidden algorithm or do it the old-fashioned way. In either case, the results returned are the ACTUAL results (I have assumed the elections officials were, for the most part, not in on the conspiracy).

What started me thinking this way was a post about Fibonacci-type numbers in the results. I googled fibonacci + "random number generator" and (as suspected) this is one way of producing "pseudo-random" numbers (that is, numbers that look random to all intents and purposes but are not random). Imagine you have a very long list of numbers that you use, each in turn: if the amount of random numbers you need to use does not exhaust your list, your numbers look random; but if they do exhaust your list THEY REPEAT AND GIVE YOU AWAY. (This is my understanding of the material that came up when I searched). The bigger the numbers in the F sequence that are used, the more "random" numbers are available, the more computing power is required. Doubling up would halve (at least) the time taken to run, the memory used and potentially make the program run-able when it would otherwise keep crashing.

Even if the F sequence is not the source of the multipliers, it is still the case that the better the algorithm, the more computing power is used. So the hacker has a choice: either crash MS Access so the point of the hack is defeated, or leave fingerprints of non-randomness behind. In the latter case, an obsessive number cruncher (sorry!) might pick it up, but they could easily be silenced; the former is the far bigger and more immediate danger to the project of GETTING ONE'S MAN IN.

As noted above, I like to think I know something about maths and computers but I am not an expert. I would be most interested to know what any experts around think of this theory.

were on to a similar theory in Florida -- only using Gore plus 5% -- But then they dissappeared off this site. I have tried to pm and email -- no luck. Haven't heard from them since around Thanksgiving.

:kick:

http://home.att.net/~sunburst59/franklin_precinct_pairs_with_factors.xls

The approach was the same as with Cuyahoga.

There is not as much variance between the unofficial and official counts as this was a touch-screen voting county.

Even though Franklin County has 756 precincts common to 2000 and 2004 (as opposed to 1419 for Cuyahoga), I was able to find 9 pairs (11 for Cuyahoga).

The most interesting pair is Columbus 5A and Columbus 5B.

In each of these two precincts, Bush received exactly the same number of votes in 2004 as in 2000.

Columbus 5A: Bush 2000 and 2004, 28.
Columbus 5B: Bush 2000 and 2004, 23.


Bush received the same number of votes in 8 other precincts. Kerry received the same number of votes as Gore in 1 precinct.

This file includes the macros for calculating the ranges.

If you know anything about VBA you can use this as a template for analyzing other counties.