@Anonymous wrote:@Anonymous There are variations in implementation among the bureaus and different scoring metrics are weighted differently or possibly not weighted at all, and that’s one of the things that we have to investigate and find out.
We do know it affects fico version 8; I don’t know if we can say it definitively affects all three bureaus. As you said, we all know the number of accounts with a balance metric is not weighted very heavily, if at all at Experian, but the metric exists and is weighted more heavily at the other two bureaus.
If your total revolving balance hasn’t changed much, then you most likely would not have seen point changes associated with the metric, unless you just happened to be next to a threshold.
Next the point changes are not believed to be of a great magnitude, so they will be at the bottom of the list of reason codes and frequently missed because only 2, 3, or 4 are typically displayed.
Again reason codes are not strong evidence; they are proof. But as previously stated, that does not mean all options are implemented at all bureaus. They are customized otherwise there was arguably be no need for three bureaus.
I can’t say what caused SJ’s changes definitively and I don’t think Cassie is attempting to either, but I’ll let her speak for herself; however, I can say raw dollar amounts do cause small changes, and the majority of those thresholds are believed to be found under $50,000, which would be the optimal bracket of 1% to 10% roughly on a half million dollar profile.
Consequently while moving aggregate utilization in that interval on a profile with that size TCL, you’re going to get noise, from the thresholds for amounts owed. Yes, change less than 5 points has been considered noise traditionally and not worthy of investigation to determine cause as maybe not even being possible. But with the advances in our knowledge, we are now beginning to learn a lot of things and apparently this is where some of that noise may be coming from.
As for the numbered paragraphs, number 1 is definitely true, IMHO, though as I previously qualified, it may not be active at Experian or it may be weighted less.
As for number two I would say known thresholds because it is possible another could be found and I think the working belief would be that yes, there should be no utilization changes from crossing 6, 7 or 8. And as stated, these are small changes, so 5 point changes would afford even further evidence, actually.
For number three, I would say for those with very high total credit limits, yes, utilization changes could be confused with raw dollar changes at low aggregate utilizations, i. e., where the 0 - 9% interval falls around $0 - $25,000 - $50,000, thereby placing it right in the middle of probable raw dollar thresholds.
Another major concern respectfully is when you rely upon a frontend that calculates utilization one way versus the algorithm calculating utilization by a different method.
Consequently, we don’t have the true utilization numbers, so there’s no way to make any conclusion for paragraph number four by anyone!
Well you don’t have to think the reason codes are absolute, they are; the only two pieces of information you get directly from the algorithm are your negative reason codes and your score change. There is nothing else; they are absolute.
Of course the wording is not precise, they have a team of lawyers there for obfuscation to keep people from suing. It’s up to us to discover what each reason code means; that’s part of the puzzle. If it were that easy we wouldn’t need any of these forums or any experiments or research. Again it’s purposefully obtuse that’s why we’ve been studying it for years to figure out what each code correlates to, but some are as plain as day.
No you need to study the reason codes, there’s different reason codes for different versions I believe and if you actually examine the chart, it indicates which ones are for industry options.
JMHO as always.
Well, it seems that we agree that it is up to us to figure out what the precise meanings of the reasoning codes are and how much weight they carry on a given Fico version.
But based on that I would call them "relative" and "evidence" and not "absolute" and "proof" 
.
I certainly agree that reasonings code are very valuable, but I do prefer actual data points. @SouthJamaica should have enough data to determine whether the score changes could be due to raw dollar amounts, and if yes, what the thresholds are.
@Anonymous wrote:
@Anonymous wrote:@Anonymous There are variations in implementation among the bureaus and different scoring metrics are weighted differently or possibly not weighted at all, and that’s one of the things that we have to investigate and find out.
We do know it affects fico version 8; I don’t know if we can say it definitively affects all three bureaus. As you said, we all know the number of accounts with a balance metric is not weighted very heavily, if at all at Experian, but the metric exists and is weighted more heavily at the other two bureaus.
If your total revolving balance hasn’t changed much, then you most likely would not have seen point changes associated with the metric, unless you just happened to be next to a threshold.
Next the point changes are not believed to be of a great magnitude, so they will be at the bottom of the list of reason codes and frequently missed because only 2, 3, or 4 are typically displayed.
Again reason codes are not strong evidence; they are proof. But as previously stated, that does not mean all options are implemented at all bureaus. They are customized otherwise there was arguably be no need for three bureaus.
I can’t say what caused SJ’s changes definitively and I don’t think Cassie is attempting to either, but I’ll let her speak for herself; however, I can say raw dollar amounts do cause small changes, and the majority of those thresholds are believed to be found under $50,000, which would be the optimal bracket of 1% to 10% roughly on a half million dollar profile.
Consequently while moving aggregate utilization in that interval on a profile with that size TCL, you’re going to get noise, from the thresholds for amounts owed. Yes, change less than 5 points has been considered noise traditionally and not worthy of investigation to determine cause as maybe not even being possible. But with the advances in our knowledge, we are now beginning to learn a lot of things and apparently this is where some of that noise may be coming from.
As for the numbered paragraphs, number 1 is definitely true, IMHO, though as I previously qualified, it may not be active at Experian or it may be weighted less.
As for number two I would say known thresholds because it is possible another could be found and I think the working belief would be that yes, there should be no utilization changes from crossing 6, 7 or 8. And as stated, these are small changes, so 5 point changes would afford even further evidence, actually.
For number three, I would say for those with very high total credit limits, yes, utilization changes could be confused with raw dollar changes at low aggregate utilizations, i. e., where the 0 - 9% interval falls around $0 - $25,000 - $50,000, thereby placing it right in the middle of probable raw dollar thresholds.
Another major concern respectfully is when you rely upon a frontend that calculates utilization one way versus the algorithm calculating utilization by a different method.
Consequently, we don’t have the true utilization numbers, so there’s no way to make any conclusion for paragraph number four by anyone!
Well you don’t have to think the reason codes are absolute, they are; the only two pieces of information you get directly from the algorithm are your negative reason codes and your score change. There is nothing else; they are absolute.
Of course the wording is not precise, they have a team of lawyers there for obfuscation to keep people from suing. It’s up to us to discover what each reason code means; that’s part of the puzzle. If it were that easy we wouldn’t need any of these forums or any experiments or research. Again it’s purposefully obtuse that’s why we’ve been studying it for years to figure out what each code correlates to, but some are as plain as day.
No you need to study the reason codes, there’s different reason codes for different versions I believe and if you actually examine the chart, it indicates which ones are for industry options.
JMHO as always.Well, it seems that we agree that it is up to us to figure out what the precise meanings of the reasoning codes are and how much weight they carry on a given Fico version.
But based on that I would call them "relative" and "evidence" and not "absolute" and "proof"
I certainly agree that reasonings code are very valuable, but I do prefer actual data points. @SouthJamaica should have enough data to determine whether the score changes could be due to raw dollar amounts, and if yes, what the thresholds are.
@Anonymous negative reason codes are data points. The only thing we get from the algorithm is score change and negative reason codes; everything else is fluff from a CMS that cannot be relied upon, so what else could you want?
We have to agree to disagree as to the reason codes because they are proof; they are the only thing that is absolute. It's the only output we have from the algorithm other than score changes. There is no other output from the algorithm, so where else can we get data from that can be relied upon?
how can you hold that data output from the algorithm is not absolute? what else do we have?