Score gains from aging

@Justaguy yes its scorecard change. On the mortgage scores you go from dirty to clean at a certain threshold, commonly believed to be 54 months, as it seems you just confirmed. Furthermore 18 months is the threshold on a clean profile for whether or not it is in a new account Scorecard. And yes the absence of the new account code does mean you are not in a new account scorecard if you have no chance of it being hidden, on the mortgage scores anyway. Fico eight codes are screwy at times but you can usually count on it there. What type of delinquency was it? Just a Delinquency or derogatory? We’re trying to determine if the threshold is the same for PR’s?

Birdman wrote: ↑Sun May 08, 2022 1:12 am @Justaguy yes its scorecard change. On the mortgage scores you go from dirty to clean at a certain threshold, commonly believed to be 54 months, as it seems you just confirmed. Furthermore 18 months is the threshold on a clean profile for whether or not it is in a new account Scorecard. And yes the absence of the new account code does mean you are not in a new account scorecard if you have no chance of it being hidden, on the mortgage scores anyway. Fico eight codes are screwy at times but you can usually count on it there. What type of delinquency was it? Just a Delinquency or derogatory? We’re trying to determine if the threshold is the same for PR’s?

It’s a lone 30D. Are you saying that was enough to put me on a dirty scorecard on EX2 for the past 4.5 years?

Justaguy wrote: ↑Sun May 08, 2022 2:30 am

Birdman wrote: ↑Sun May 08, 2022 1:12 am @Justaguy yes its scorecard change. On the mortgage scores you go from dirty to clean at a certain threshold, commonly believed to be 54 months, as it seems you just confirmed. Furthermore 18 months is the threshold on a clean profile for whether or not it is in a new account Scorecard. And yes the absence of the new account code does mean you are not in a new account scorecard if you have no chance of it being hidden, on the mortgage scores anyway. Fico eight codes are screwy at times but you can usually count on it there. What type of delinquency was it? Just a Delinquency or derogatory? We’re trying to determine if the threshold is the same for PR’s?

It’s a lone 30D. Are you saying that was enough to put me on a dirty scorecard on EX2 for the past 4.5 years?

No definitely not. I mispoke without all of the information. But there is a threshold ~ 54 months for delinquencies, apparently you just crossed it. We’re trying to confirm where the scorecards go from recent to mature on the dirty cards on the new versions. On old versions you go back to a clean card after like two years, but with penalty. You just found a threshold where it gives back points, as evidenced by reason code shift. But a 30 day isn’t sufficient.

I think there's more going on here than just a diminishing penalty for the baddie. I described my EX AU2 scores and reasons above in Post 5, but here are screenshots to reinforce the point. As you can see, the score went up 14 points. So we would expect to see some factors that used to be relatively high either vanish or move down. But look what moved down: AWB. It went from first position to fourth. I can only think of three possible explanations for that factor moving down. 1. If the number of accounts with balance got better. But that is emphatically not what happened. Literally nothing changed with my balances. Not by even one account or one dollar. 2. If other factors got worse and moved ahead of AWB. This also did not happen. The only -- and I mean only -- thing that changed from 4/30 to 5/1 was aging. Nothing in FICO's eyes gets worse with another month of aging, so it's not possible for scoring penalties to increase with aging, except... 3. If I moved to a new scorecard. This can alter the signal strength of many things, and produce big changes in the score with comparatively minor changes in the data. If anyone sees another explanation for AWB moving down, I'd love to hear it.

I'm confused as to how AWB under any circumstances (even 100% AWB) would be more impactful and thus placed higher up on the list than bad payment history.

BrutalBodyShots wrote: ↑Mon May 09, 2022 10:02 pm I'm confused as to how AWB under any circumstances (even 100% AWB) would be more impactful and thus placed higher up on the list than bad payment history.

In the abstract, I get that. But with details added it seems reasonable. The totality of my bad payment history is one lone 30D from 4.5 years ago. My number of accounts with balance is six. Given how few cards people had back when the scoring model behind EX2, FICO 98, was being developed, I find it easy to believe that a whopping six accounts presently having a balance was more highly correlated with eventual default than a single missed payment from 4.5 years back (or, put differently, perfect payment history for 4.5 years).

That's a reasonable assessment.

Justaguy wrote: ↑Mon May 09, 2022 1:39 am I think there's more going on here than just a diminishing penalty for the baddie. I described my EX AU2 scores and reasons above in Post 5, but here are screenshots to reinforce the point. EX AU2 Apr 30 2022.JPG EX AU2 May 1 2022.JPG As you can see, the score went up 14 points. So we would expect to see some factors that used to be relatively high either vanish or move down. But look what moved down: AWB. It went from first position to fourth. I can only think of three possible explanations for that factor moving down. 1. If the number of accounts with balance got better. But that is emphatically not what happened. Literally nothing changed with my balances. Not by even one account or one dollar. 2. If other factors got worse and moved ahead of AWB. This also did not happen. The only -- and I mean only -- thing that changed from 4/30 to 5/1 was aging. Nothing in FICO's eyes gets worse with another month of aging, so it's not possible for scoring penalties to increase with aging, except... 3. If I moved to a new scorecard. This can alter the signal strength of many things, and produce big changes in the score with comparatively minor changes in the data. If anyone sees another explanation for AWB moving down, I'd love to hear it.

Let me answer your question the best I can based on my theory. On the old scores, a person with a 60D or worse went back to a clean card at two years. You never went to a dirty card on New or old scores because it was just a 30 day late. Either way, You get points back at 6 months 12 months 18 months 24 months and I don’t know all the other ones. But there is definitely a separate system whereby the old scores had a way to penalize derogatories after two years while still being in a clean scorecard as evidenced by your many new accounts code. I do not even pretend to know how their penalty system for derogatories after two years works, not that I pretend about anything anyway, but just saying. But what we do know is there are certain thresholds along the way and this may be an example of something @Cassie mentioned once about second layer binning or like a separate scorecard within a scorecard. In other words, you didn’t change main scorecards, but you crossed a threshold that moved you out of a certain penalty section of your scorecard, I believe. I would like to confirm Delinquencies change Scorecards at 24 months or if not when for the new scores. I would also like to verify PR‘s reassigne to mature at 54 months. I think I may have spoken incorrectly earlier, we think delinquencies reassign to mature at 24 months on the new scores but think it’s around 54 months for PR, but it would be nice to have tests & multiple confirmations. Edited first paragraph for errors @Justaguy

This “scorecard within a scorecard” notion sounds like a very plausible — and very interesting — way of explaining what happened to me on May 1.

I would love to hear/learn more about it. Thanks for introducing the concept to me!

Justaguy wrote: ↑Thu May 12, 2022 2:48 am This “scorecard within a scorecard” notion sounds like a very plausible — and very interesting — way of explaining what happened to me on May 1. I would love to hear/learn more about it. Thanks for introducing the concept to me!

Here's a nice introduction to it: Credit Models and Binning Variables are Winning and I’m Keeping Score! TransUnion has some whitepapers where they discuss sub-binning. Searching for those keywords should bring it up on the first page. There are a few links to articles about logistic regression (what credit score models are based on) in the Primer bibliography as well.

Cassie wrote: ↑Thu May 12, 2022 3:07 pm

Justaguy wrote: ↑Thu May 12, 2022 2:48 am This “scorecard within a scorecard” notion sounds like a very plausible — and very interesting — way of explaining what happened to me on May 1. I would love to hear/learn more about it. Thanks for introducing the concept to me!

Here's a nice introduction to it: Credit Models and Binning Variables are Winning and I’m Keeping Score! TransUnion has some whitepapers where they discuss sub-binning. Searching for those keywords should bring it up on the first page. There are a few links to articles about logistic regression (what credit score models are based on) in the Primer bibliography as well.

Thanks! I read the piece you linked to. I can’t pretend to fully understand everything in it, but what it’s describing sounds to me like changing signal strength within a scorecard. Applying it to my case, with hypothetical numbers, when I had a 30D aged < 54m, let’s say Bad Payment History had a relative weight of 0.5, AWB had a relative weight of 0.2, and all other factors have a combined relative weight of 0.3. (I like to use relative weights between 0 and 1, with the total summing to 1, for ease of interpretation.) When my baddie hit 54m, it might have entered a new bin with a new relative weight of 0.6 and now AWB’s relative weight drops to 0.1 while all other factors’ weights stay unchanged. This, or something like it, could explain why I’m losing fewer points at the same AWB without changing scorecards. @Birdman: is this how you understand it too? Does this soften your resistance to the notion of changing signal strength within a scorecard? @BrutalBodyShots: is this similar what you had in mind as an explanation? @Cassie: Have I completely missed the point of the piece?

Justaguy wrote: ↑Thu May 12, 2022 4:34 pm @Cassie: Have I completely missed the point of the piece?

Oh no, and in fact normalizing the weighting to a [0,1] range is common practice. I don't know exactly how it's done, but with only 8 clean scorecards, there has to be some sort of sub-binning going on in each scorecard. Otherwise, everyone should see the same results on that Clean/Thick/Mature/No-New-Account scorecard, but we know they don't. Length of credit history still seems to have an effect after the binary switch to 'Mature'.

I will preface this with her haven’t read the article linked yet. I do believe subbinning is possible for 30 day lates because that is a special circumstance where a derogatory does not cause Scorecard reassignment. There has to be some function. Likewise for after two years for derogatories on the old scores. That’s my best guess yes. It could limit maximum contribution for like category well within that intro. It could limit maximum contribution for that category while within that interval? I do not have any evidence of signal strength necessarily being affected, I think it limits the maximum contribution while in that bin. I don’t believe inquiries or other things affect it, although I cannot rule it out 100%. Even if so, I do not believe signal strength is affected across categories. PH is 35%, there is plenty of room to limit/penalize. @Cassie maybe right on point for aging because it does seem that people in a younger mature scorecard are treated differently than those who are in a older mature scorecard. I think it is limited to the length of time category’s withholding part of its contribution also, not causing subbinning across categories. But, CRADIS will resolve those questions. Also I edited the above first paragraph couple typos.