FICO Algorithm weakness

[remboman]

Since I began monitoring my FICO score, it has ranged from 750 - 768. In December 07, my wife bought my son a pair of slippers for $12.46 at JCP as an "impulse" purchase. For some reason, she forgot about this purchase and assumed there was still a $0 balance on that account, which there had been for a long period of time. Needless to say, 30 days passed and I received MyFico alert that my score had dropped 79 points from 750 to 671, overnight, due to the 30 day late on the $12.46.

The reality is that I am no higher risk to default on a loan now than before the slipper purchase, however, the FICO score algorithm "calculates" that I have gone from great credit (extremely low risk of default) to below average credit (high risk of default).

I recognize how the game is played, and that's not my point. The reality is that the FICO score algorithm is a very poor indicator of actual default risk. This weak algorithm has significant consequences. Since my daughter will be entering college soon, our plan to finance via a 2nd will cost us several thousand more dollars now because of our high-risk (671) status. My next auto lease will be at a higher rate that will cost me hundreds of dollars more.

[Anonymous]

GW JCP -- they can remove the late.

[MidnightVoice]

remboman wrote:

The reality is that the FICO score algorithm is a very poor indicator of actual default risk.

I beg to differ.  Statistically, it is probably a very good indicator.  That means, as in all statistics, that some points (like you) are outliers.  That is terrible for you, but does not invalidate the statistics.

And with scores like yours and history like yours, a GW letter probably has a good chance of success

Message Edited by MidnightVoice on 03-17-2008 11:55 AM

[remboman]

MidnightVoice wrote: Statistically, it is probably a very good indicator. That means, as in all statistics, that some points (like you) are outliers. I believe that the inherent weakness in the algorithm is that it has to be oversimplified to provide risk assessments for hundreds of millions of credit users. The fact that it does not account for the amount of the 30 day late, whether it's $12 or $12,000 is the weakness. I realize that several additional variables would need to be included to more accurately assess the actual risk (ie other payments/amounts made during the same period, etc.) and it's just not feasible when you're assessing the risk of hundreds of millions of credit users. A better, more complex algorithm would enhance the probability of accuracy when calculating actual risk.

[MidnightVoice]

remboman wrote:

I believe that the inherent weakness in the algorithm is that it has to be oversimplified to provide risk assessments for hundreds of millions of credit users. The fact that it does not account for the amount of the 30 day late, whether it's $12 or $12,000 is the weakness. I realize that several additional variables would need to be included to more accurately assess the actual risk (ie other payments/amounts made during the same period, etc.) and it's just not feasible when you're assessing the risk of hundreds of millions of credit users. A better, more complex algorithm would enhance the probability of accuracy when calculating actual risk.

I think you might find it is already very complex.

And it does not take income into account, and there are arguments for and against this.  So a late is a late

I just think that by and large it probably does do a pretty good job - the statistics used in health care decisions are probably both less precise and less accurate.

I am sure it can be improved, but this is a capitalist society.  As long as the model keeps the customers happy there is no need to improve it, and as you and I are not constomers, are opinions are worth not a thing. 

[remboman]

As you said, "I just think that by and large it probably does do a pretty good job" "I am sure it can be improved" By and large, I agree. I give the algorithm about a "C". BTW - I looked at all your posts - how many of you are there?

[MidnightVoice]

---

remboman wrote:
As you said, "I just think that by and large it probably does do a pretty good job" "I am sure it can be improved" By and large, I agree. I give the algorithm about a "C". BTW - I looked at all your posts - how many of you are there?

---

I suffer from MPD

[fishbjc]

call JCP, they will, more than likely, remove the late.  They just removed two of mine and credited me my late fees.

[remboman]

Thanks. I'll give it a shot.

[modernsurrender]

Wasn't there something in some of the recent Fico 08 articles that talked about this very sort of issue?  Where John Doe has great credit, no lates, no negatives at all - then something happens and one minor baddie pops up and the new Fico 08 gives that baddie less of an impact than before?

I haven't read on the news forum in a while but I thought one of the articles discussed this sort of situation and decreasing the huge drops caused by this kind of issue.