Testing greater than 8.99% but less than 9.00%

[Revelate]

Yeah, not only that it's one of the places where you can see the translation that myFICO does to make all the reason codes line up.

Frankly for reason code analysis for FICO 8 the TU monitoring here is second to none.

Sadly everywhere else including the myFICO main interface truncate them for FICO 8... it's one of the things I abolutely hate about CMS's these days, but I guess not everyone wants as much detail as I do and I'm not their target demographic.

[Anonymous]

---

@Revelate wrote:

Sadly everywhere else including the myFICO main interface truncate them for FICO 8... it's one of the things I abolutely hate about CMS's these days, but I guess not everyone wants as much detail as I do and I'm not their target demographic.

---

I wanted to make a post titled 'Misunderstanding FICO Scoring' about the various mismatches in data I have seen across 8 CMS front-ends.

Then I realized I don't have enough bandwidth for all the snapshots of the utilization rounding errors.

You're right - we're not the target demographic and most people are just happy to see all the scores.

[Anonymous]

Here's another data point proving that utilization does not have to be less than 'x.99'.

(I'm under 3 years total credit history, so these changes may not happen on other scorecards.)

EX 8 will go up +3pts when all cards are at 4% or below. EQ/TU 8 go up +3/+1 with just aggregate at or below 4% - an individual card can be above 4%. Proof here.

Right now, I have 4 cards reporting at 4% utilization each, with various percentages calculated to 3 decimal places. If one of these reported above 4%, I would lose 3 points on EX 8, same as I gained when I switched to all at 4%.

Two of these cards are at 4.000 and 3.991. I did not lose those points with that card reporting >3.99 percent.

[Revelate]

I really want someone to test that installment loan metric of 8.99% too honestly, I suspect 9.5% would be fine too based on your findings but unfortunately I can't chuck a half million playing reindeer games right now, or probably ever.

[Anonymous]

---

@Revelate wrote:

I really want someone to test that installment loan metric of 8.99% too honestly, I suspect 9.5% would be fine too based on your findings but unfortunately I can't chuck a half million playing reindeer games right now, or probably ever.

---

I would expect 9.5 to be taken as 10% by the algorithm, which would put util in the next higher interval and score lower.

Do you think the rounding is different for installment and revolving utilization calculations? I never even considered that. Everything points to a ceiling function being used, such as anything slightly above $0 becoming 1%.

[Revelate]

---

@Anonymous wrote:

---

@Revelate wrote:

I really want someone to test that installment loan metric of 8.99% too honestly, I suspect 9.5% would be fine too based on your findings but unfortunately I can't chuck a half million playing reindeer games right now, or probably ever.

---

I would expect 9.5 to be taken as 10% by the algorithm, which would put util in the next higher interval and score lower.

Do you think the rounding is different for installment and revolving utilization calculations? I never even considered that. Everything points to a ceiling function being used, such as anything slightly above $0 becoming 1%.

---

Maybe I'm confused by your datapoints, but I would have expected them to be the same too TBH.

I have to go re-read, I'm in a serious fugue state at this point which I'm pretty sure is back to working hard and eating like crap again because I don't have a working refrigerator.  I'm starting to loathe new construction nothing ever works correctly the first time anymore it seems =/.

[Anonymous]

---

@Revelate wrote:

---

@Anonymous wrote:

---

@Revelate wrote:

I really want someone to test that installment loan metric of 8.99% too honestly, I suspect 9.5% would be fine too based on your findings but unfortunately I can't chuck a half million playing reindeer games right now, or probably ever.

---

I would expect 9.5 to be taken as 10% by the algorithm, which would put util in the next higher interval and score lower.

Do you think the rounding is different for installment and revolving utilization calculations? I never even considered that. Everything points to a ceiling function being used, such as anything slightly above $0 becoming 1%.

---

Maybe I'm confused by your datapoints, but I would have expected them to be the same too TBH.

I believe the internal code breakpoints are 0, 5, 10, 30, 50, 70, and 90%. (Aggregate for example.)

They take the ceiling of (total balance/total limit), and then compare to one of those.

Ceiling(9.01) = 10 which is not less than 10, so it's in the next higher interval.

This also makes sense when considering all the articles out there that talk about 'less than 10% utilization'. Many of them have quotes or comments from someone at FICO (VP of credit scoring), or some expert like John Ulzheimer.

On the forums, it's simplified to < 8.9, 28.9, 48.9, 68.9, and 88.9. This is ok to use, of course. But it's not technically correct, as I have found with these tests to see if I went over the 9% and 4% thresholds on my scorecard.

Util > x.99 isn't causing some rounding error to (x+1).01 (8.99 to 9.01) or something like that. This may have been a guess by someone familiar with floating point rounding errors.

I have to go re-read, I'm in a serious fugue state at this point which I'm pretty sure is back to working hard and eating like crap again because I don't have a working refrigerator.  I'm starting to loathe new construction nothing ever works correctly the first time anymore it seems =/.

---

Sorry about that.