No, it's more complex than that. There are cases of that in some datasets that need to be investigated but this is something different.
Basically they're saying they can make a relatively simple algorithm that can predict out of how many people registered actually voted with extremely high precision and you shouldn't be able to do that so easily with such a simple algorithm. Usually what you expect would be far more random and difficult to predict without a large margin of error.
What they show on the graph though does appear to have a small margin of error but perhaps not enough to show at their level of precision (less than a thousandth inaccurate).
The population line compared to the registered voter lines gives some baseline for the amount of deviation you might expect though I don't see them having done the same for that and they should have (though it looks like their model won't match at a glance).
Their algorithm to predict is is I believe for to a plotter or 2D graphics plotter that lets you draw a bezier curve with a few points to bend around. It's like that. Or like a tween (look up tweens). They found a simple one that predicts it more than 99.99% accurately when you apply that modification to the number of registered voters.
To truly verify if that's plausible or not to happen naturally isn't something I have the time for personally at the moment but at a glance it's definitely curious. The problem with a lot of this stuff is that it isn't really easy to explain it to the layman in a way they can verify, it's complex. I'm not that far off a layman myself.
You can however visually see that registered voters versus voters correspond very strongly compared to population versus registered voters so that alone should tell you there's something different happening there, whether natural or not the process is not the same, the correlation is far strong (IE if registered voters goes up 10% then so do votes but normally you might expect +-1% instead so 9% to 11%). Instead you're seeing give or take so many percent much lower than with other related variables.
This does need to be verified though. You can make an algorithm that will find the best fit for a 1st order, 3rd order, 4th order, 6th order, 12th order, etc polynomial and at some point for any dataset you'll end up with a match of some order with a one in a thousand degree of accuracy. This doesn't mean anything unless you actually do the calculations of the match for the best fit 6th order or at least some comparisons. However just six points also based on the ups and downs of the difference of the difference is a a little tight. I can't rule it in or out at a glance and I'd have to figure out the maths to determine it.
I think the warning sign is that the original graphs show around six ups and downs as well which basically pretty much makes them already sixth order polynomials so extract another curve from them and it might be the same. Though they're also adding that it's very consistent across states which still makes it curious. Given the amount of noise on the registration graph my argument against it based on shallow reasoning isn't enough to dismiss it. Basically I'm saying the difference still appears to have a big drop in variance and no immediate explanation comes to mind.
Also if it were to fit them all you would surely get sometimes 5 orders, sometimes 6, etc. You see this where in terms of for example the total population (which is a smooth curve) some graphs are 4hz, some are 5hz, some are 6hz, etc. The same if you smooth the registered curve. I wouldn't personally expect one exact "key" to unlock all doors. I could be wrong but if so it's a remarkable mathematical effect.
What really makes it hard to believe is that I have some fairly extensive experience and knowledge of how they can cheat but the scale of this if it is true is enormous.
No, this is saying that once he figured out the percentage for each state, he combined that with the 2010 Census data and it produced exactly the “actual” vote. This is a time bomb.
Is this saying that exactly 100% of registered voters voted?
No, it's more complex than that. There are cases of that in some datasets that need to be investigated but this is something different.
Basically they're saying they can make a relatively simple algorithm that can predict out of how many people registered actually voted with extremely high precision and you shouldn't be able to do that so easily with such a simple algorithm. Usually what you expect would be far more random and difficult to predict without a large margin of error.
What they show on the graph though does appear to have a small margin of error but perhaps not enough to show at their level of precision (less than a thousandth inaccurate).
The population line compared to the registered voter lines gives some baseline for the amount of deviation you might expect though I don't see them having done the same for that and they should have (though it looks like their model won't match at a glance).
Their algorithm to predict is is I believe for to a plotter or 2D graphics plotter that lets you draw a bezier curve with a few points to bend around. It's like that. Or like a tween (look up tweens). They found a simple one that predicts it more than 99.99% accurately when you apply that modification to the number of registered voters.
To truly verify if that's plausible or not to happen naturally isn't something I have the time for personally at the moment but at a glance it's definitely curious. The problem with a lot of this stuff is that it isn't really easy to explain it to the layman in a way they can verify, it's complex. I'm not that far off a layman myself.
You can however visually see that registered voters versus voters correspond very strongly compared to population versus registered voters so that alone should tell you there's something different happening there, whether natural or not the process is not the same, the correlation is far strong (IE if registered voters goes up 10% then so do votes but normally you might expect +-1% instead so 9% to 11%). Instead you're seeing give or take so many percent much lower than with other related variables.
This does need to be verified though. You can make an algorithm that will find the best fit for a 1st order, 3rd order, 4th order, 6th order, 12th order, etc polynomial and at some point for any dataset you'll end up with a match of some order with a one in a thousand degree of accuracy. This doesn't mean anything unless you actually do the calculations of the match for the best fit 6th order or at least some comparisons. However just six points also based on the ups and downs of the difference of the difference is a a little tight. I can't rule it in or out at a glance and I'd have to figure out the maths to determine it.
I think the warning sign is that the original graphs show around six ups and downs as well which basically pretty much makes them already sixth order polynomials so extract another curve from them and it might be the same. Though they're also adding that it's very consistent across states which still makes it curious. Given the amount of noise on the registration graph my argument against it based on shallow reasoning isn't enough to dismiss it. Basically I'm saying the difference still appears to have a big drop in variance and no immediate explanation comes to mind.
Also if it were to fit them all you would surely get sometimes 5 orders, sometimes 6, etc. You see this where in terms of for example the total population (which is a smooth curve) some graphs are 4hz, some are 5hz, some are 6hz, etc. The same if you smooth the registered curve. I wouldn't personally expect one exact "key" to unlock all doors. I could be wrong but if so it's a remarkable mathematical effect.
What really makes it hard to believe is that I have some fairly extensive experience and knowledge of how they can cheat but the scale of this if it is true is enormous.
No, this is saying that once he figured out the percentage for each state, he combined that with the 2010 Census data and it produced exactly the “actual” vote. This is a time bomb.