This is a summary of a thread on this board that now is over 1,000 posts -- far too many for people to 'catch up on', especially considering some of it is rather complicated mathematical analysis.
We are at a crossroads. We NEED to bring more people on board, especially those with math/data analysis backgrounds and others well suited to figuring out the best path forward. (Our eventual goal, hopefully, to produce another consolidated report, perhaps a website, and break this story in the media).
Where we are - it appears we have strong evidence of vote flipping in the Primary. Or at least, anomalies have been discovered that are incredibly difficult to explain away, though a simple algorithm would explain them.
Much of the research focuses on South Carolina, but we have extremely suspicious data on most other states as well, though we need to be careful since some states are primaries and others are caucuses, which need to be kept separate, even if both end up being fraudulent.
The basic summary:
1) Romney is always the only benefactor.
2) There is evidence of vote flipping going back to the 2008 primary.
3) The algorithm(s) being used are rather crude, often basic 1:1 flipping.
4) Votes are often, but not always, siphoned from a single candidate. This candidate is often Ron Paul, but has also been Gingrich, Santorum, and even Huckabee in 2008.
5) Romney benefits as precincts increase in size, and this increase is algorithmically 'clean' with little or no 'white noise' common to non-altered candidates. For example, we might see a steady 10% rise in Romney's votes from precincts sized 50% to 80%, at which point it increases to a steady 15% (far after any differences in size should matter).
6) Demographics are not at play, though this is the 'debunk' most often brought up by people new to the thread.
*****This is not a case of the 'urban'/'rural' divde, which is the 'go to' explanation some new to the thread jump to.
a)We see the same algorithmic rise in Romney votes even within precincts of a single city, all urban. Furthermore, in 'unaffected' counties, all precincts within a city are flatlines for all candidates. More on this below.
b) examination of 'untouched' (no anomaly) counties shows no such divide.
c) we see the algorthymic rise when evaluating sets of precincts that are practically the same size; that is, if we look at say, Clark County, NV, we see the same rise between precincts of 60 then 65 then 70 voters, as we do in a state where precincts differ by 500 or more. There is no reason to suspect such a causal demographic difference between precincts of the same size that differ only in a handful of votes.
*****EXPLANATION OF X AXIS OF MOST CHARTS
Most, but not all, charts on the following pages show 'cumulative votes sorted from smallest vote precinct to largest'. In what those working on this consider to be examples of non-affected Counties/states, you'll see all candidates quickly flatline at an early percentage point tallied. In affected counties, you'll see a distinct 'slope' for 2 (in a few cases, more) candidates, in which Romney is always the upwards slope. These slopes show less 'noise' than even the flatlines, and often are practically 1:1 siphons from the 'victim' candidate.
***** Liberty1786 has now detected same anomalies in 2008 Democratic primaries!!! 2004 appears clean. More charts added below.
This research began as a result of a study posted to the dailypaul (by a user now here, named The Man) and has been furthered by several here, most notably user Liberty1789.
The original 1000+ post thread can be found here:http://www.ronpaulforums.com/showthr...South-Carolina
The original study which began this thread can be found here:https://docs.google.com/viewer?a=v&p...U2ZGU1OWZkZjhk
I will now cut and paste the most relevant posts from the 1000+ thread. If I miss any, those familiar with the thread should feel free to post.
PLEASE NOTE THAT DIFFERENT CHARTS MAY APPLY TO DIFFERENT STATES, and that some states are primary and some are caucus. I will do my best to post relevant charts and analysis in order. Please don't post till I post a 'go ahead', I have quite a bit I need to repost.
If you want detailed info on any of this, the original thread has massive discussion. Feel free to post here as well, but this is mostly meant to be a summary thread for those new, because we NEED more people involved pronto.
EDITED TO ADD: Response to person wanting a better explanation of graphs:
Originally Posted by affa
I suggest starting with the report linked to in post 1. That's the essay that started it all. It's focused on specific counties in SC where, when charted, the data looked particularly manipulated. From there, we've (mostly Liberty at this point) have been trying to move the research forward by analyzing the data in various ways.
From there, advancements in the theory have been made. For example, originally, it was determined that only precincts of a certain size (in SC's case, 277+) were affected. That lead us to believe it might be based on vote size. However, further research showed that in fact, it seems to only occur when the % of votes in that precinct represent a certain amount of votes in the county.
Almost all charts shown represent cumulative vote across precinct size. That is, small to large. Now, areas where no data manipulation is detected follow what might be expected -- by the time you're at a reasonable amount of votes included in the cumulative total, the results 'flatline'. We see this often, and it matches what one would expect. There are plenty of non-anomaly counties where all candidates flatline early.
However, in many counties, we're seeing distinct anomalies. Specifically, these anomalies are sloped linear increases across entire spectrums of precincts (say, from 20% through 90%) in which one candidate (always Romney) increases votes at an algorithmic rate, almost always at the expense of a sole victim. The slopes are clean, with little to no noise that can be found in the results of candidates unaffected. That is, even the flatlines show noise. These anomalies do not.
The slope is generally, but not always constant. We feel demographics do not play a role because this is not happening in jumps -- that is, we're not seeing differences between small, medium, and large, but rather, a steady, consistent rise between any two affected data points -- for example, from precincts sized 20% through 30%, or 80% through 90%.
Just as odd, we find this same algorithmicly smooth (we suspect) rise regardless of whether we're talking about precincts that differ by 5 votes or 500 votes (see post 27 for the best example/explanation of this). That is, we see the same algorythmic rise in Clark County, NV where deciles are measured by handfuls of votes, verses NH when examining precincts that differ by thousands.
The linear rise is clean and algorithmic in all cases, sapping votes based on the % of total votes that the precinct represents. The percent rises, cleanly, from 0% in precincts that show no anomaly (due to small size) to upwards of 9% in the largest counties. Again, this is a smooth rise across size... there aren't jumps at certain sizes representative of say, cities.
In fact, a recent poster did a precinct evalation of a single city:
So we see this odd anomaly even with an entirely urban environment, all based on relative size.
The benefactor is always Romney.
It usually only affects one victim, though there are some more complex cases where we see it involve one victim from one decile through another decile, then switch to a different victim (but still a clean algorithmic rise) at a later decile... usually just enough to give Romney the win. In other cases, where he's getting trounced by Gingrich, we see just enough to give him a solid second, rather than, say, 3rd (this is often where he siphons from Gingrich).
I hope this is a basic enough summary to get you to better understand the charts. I am not a statistician, but I'm pretty sure this is a solid explanation of what we are seeing. It is our contention that these rises are so clean and algorithmic, based on definable factors, that it's a clear cut case of a rather rudimentary vote flipping technique. The technique tends to hide gains in ways that seem easy to explain (oh, Romney does better in cities!) but upon closer look the stats win out and prove these demographic debunk attempts flawed and incomplete.
If we're correct, then in fact, Romney is being rejected state after state. Ron Paul voters actually do vote. And other candidates, such as Santorum and Gingrich, are also doing better than reported.
EDITED TO ADD: Summary of findings from Liberty1789
Originally Posted by Liberty1789
I would like to rephrase/reiterate the following.
(affa, you might want to consider adding it to the front post)
The data shows so far:
1. In some counties, Romney's cumulative shares of vote increases linearly from a given point only when the precincts are sorted by vote tally.
2. The linear segment surge binarily kicks in at any decile from the 2nd and sometimes binarily stops before the 10th.
3. The volatity of the linear surge segment is significantly lowered ("white noise reduction").
4. Romney's surge binarily affects other candidates, simultaneously or in sequence in a county.
5. The affected opponent(s)' cumulative lines exhibit the same straightness and reduction of volatility.
6. Romney's surge is correlated to cumulative % votes cast, not precinct population.
7. Counties representing a low % of the total ballot are always unaffected.
8. Romney's surge binarily affects some counties and not others.
9. In counties without Romney's surge, the cumulative score of the candidates tends to follow the mathematical law of hypergeometric distribution. That relationship is totally broken when Romney surges, affecting identically at least one other candidate.
The hypothesis of an algorithmic vote flipper switching a % of the final score(s) to precincts proportionately to their shares of total county vote tally, from 1, 2 or 3 candidates to Romney, would explain ALL the data properties from above in one stroke.
And this is what needs to be debunked.
ONE EXAMPLE OF MANY:
Originally Posted by Liberty1789
Beautiful 1-for-1 vote flip in 2008.
What's the best explanation for this anomaly? Demographics or victim name update in a piece of software?
The 'go to' debunk attempt by people unfamiliar with the data seems to be that Romney simply does better in urban areas. This does NOT explain the anomalies being detected. User Aden has been mapping out all precincts within city limits for cities in South Carolina. That is, all precincts within the following graphs are not only 'urban', but within the same city.
We are finding the same anomalies here for counties that are 'affected', while counties that show no signs of the statistical anomalies show flatlines for all candidates within city limits. Examples:
All precincts within city limits in a county where anomalies were detected:
Originally Posted by Aden
All precincts within city limits in a county where NO anomalies were detected:
Originally Posted by Aden
This is NOT something so simple as 'Romney does better in urban areas'. There are hundreds of posts evaluating this claim, and it's been shown to be a false premise in several different ways (the above graphs being perhaps the best).
Affected counties show a systematic, no-noise algorithmic increase in Romney's votes from a specific hinge point to a specific hinge point, with the votes coming from different opposing candidates in different counties (sometimes neighboring). This is not based on the rural/urban divide, for we see this same precise incline even when looking at precincts within a specific city (in affected counties)... counties that show no evidence of anomalies have the expected, flat results for all candidates.
Has this always happened?
Well, here is a historical look at a couple counties in NH. Note that, no, this has not always happened. In fact, we get flatlines going back to 1996... except for Romney in 2012.
Originally Posted by Liberty1789
Last edited by affa; Yesterday at 12:13 PM.