04-08-2017 11:16 AM - edited 04-28-2017 07:53 AM
Starting a new thread as it turned out to be unanticipatedly more interesting than I expected rather than a confirmation of what I already allegedly knew and I caught an Amex charge card datapoint too. I also had pulled an EX 1B report after my late went away to see if I got anything (didn't) so I had a decent comparison set which I decided to simply follow. Posting this part of it now before I catch a flight tomorrow.
Also a much larger drop than my last test (-11 or -12 on all 3, this one EX FICO 8 dropped 23, and EQ FICO 8 dropped 18), may have shifted scorecards in that period of time with some of the changes I had (dropped a 30D from April so presumably the pattern lates are now no longer a possible factor in FICO 8 for example, pure conjecture btw, no real way to tease that out at this point in time).
1) Test HELOC on all models
2) Test PLOC on all models
3) Test Amex charge card on all models
Charge card has a high balance of $11K when we're talking CL on some models
HELOC has a limit of 27K.
PLOC has a limit of 20K
Aggregate revolving discounting the above: 163K
1) Pay off all non-HELOC/Amex revolvers to zero, HELOC / Amex reporting for datapoint. Done 4/7
2) Pay off Amex, statement date 4/11, only HELOC reporting balance reporting for datapoint (also DCU Beacon 5.0) . Done 4/15
3) Transfer $20 out of the PLOC, reports end of month, HELOC + PLOC reporting for datapoint
FICO 8 all models: Appears to discount both the HELOC and Amex charge card from the number of revolvers with balances calculation.
FICO 9 all models: Identical to FICO 8 in behavior
FICO 04: Only the one FICO 3 score in this current dataset (I will snag DCU's EQ towards the end of the month) and apparently is counting either the HELOC or the charge card or both. Prior data from others suggests the charge card isn't factored, which implies the HELOC change wasn't backed into this algorithm as I had understood or size of the HELOC does matter and my 27K isn't enough to get above the hurdle.
Update: Appears the HELOC is counting on EX FICO Risk Model v3 which is a FICO 04 algorithm, checking DCU near end of month - Amex now zerod, no change in score.
Update 4/28: The HELOC does not appear to factor into EQ Beacon 5.0 based on DCU's provided scores: 693 -> 684, -7 on my dirty scorecard. That's strange, I wonder if FICO may well have backed it into this model as I had thought they did, and possibly TU 04 too on the premise that they're actually used whereas EX FICO Risk Model v3 made next to zero market penetration as I understand it.
Update 4/28 #2: The HELOC is not counting under TU FICO 4 either, looks like it was backed out on both of the FICO 04 models used for mortgage underwriting.
FICO 98: Not sure what's going on there, this is a pretty clean dataset pulled within a short time period and just balance changes, it appears more sensitive than FICO 8 to numbers of cards with balances, and can sort of conjecture the magnitude is more on the BC industry option, but I am at a loss as to the AU drop. To be fair I only care about the base EX FICO 2 score as that's relevant for mortgages and a bunch of CU's, other two scores = whatever.
I will update the analysis as more data comes in or if others analyzing the data see something I missed.
|EX FICO 8||732||709||-23||709|
|EX FICO 2||719||723||4||723|
|EX FICO 3||678||679||1||679|
|EX FICO 9||768||733||-35||733|
|EX FICO 8 AU||729||714||-15||714|
|EX FICO 2 AU||701||694||-7||694|
|EX FICO 9 AU||771||740||-31||740|
|EX FICO 8 BC||745||729||-16||729|
|EX FICO 2 BC||701||717||16||717|
|EX FICO 9 BC||759||730||-29||730|
04-15-2017 06:44 AM
4/15 Update: Amex -> $0, and literally nothing moved.
FICO 9: Discounts both HELOC and Amex charge card (Term = 1 month)
FICO 8: Discounts both HELOC and Amex charge card (Term = 1 month)
FICO 04: EX appears to count the HELOC absolutely, waiting on DCU for EQ Beacon 5.0 at end of month
FICO 98: counts the HELOC as we pretty much all knew.
04-19-2017 02:53 AM - edited 04-19-2017 02:55 AM
Update from 4/14 from CK's Vantage score.
Balances: Mortgage, HELOC, $22/500 Alliant loan, and still $10 on the Amex at this point.
Changes from last (according to CK, I will verify this eventually): DCU / Elan / Chase -> $0
EQ VS 3 +7 to 802 of all things, which is a new record for me.
TU VS 3 +7 to 707 (my file scores really awkwardly on this bureau)
04-28-2017 07:34 AM
4/28 Update: DCU provided EQ FICO 5: 693 -> 686 (-7)
Balances: Mortgage, HELOC, $22/500 Alliant loan
Wasn't expecting this after the EX FICO 3 data but it doesn't appear that the HELOC is counting under EQ Beacon 5.0.
Going to also try zeroing out the HELOC but unfortunately I can't get DCU to report that for another 32ish days, and I have inquiries passing the 1 year mark in May from my aborted refinance and HELOC and that may skew results somewhat though until my 60D or tax lien falls off it shouldn't be substantial.
04-28-2017 07:51 AM - edited 04-28-2017 08:00 AM
Well I will be darned. Bought the 1B report for TU just to see, it's not factored into TU 04 either.
4. There are no recent balances on your revolving credit accounts.
That's on every model provided except for FICO 8 AU where it's complaining about no recent auto loan activity
Actually, and this one's interesting to me, base FICO 9 reason codes:
The heck? Evidence now of revolving accounts not all being created equally????? Guess I should've pulled TU as well back when the Amex was reporting but since I don't really care about the bureau, le sigh. Also interesting it isn't complaining about recently missed payments which TU 8 and TU 4 all industry options are. FICO 9 BC mentions the late, FICO 9 AU doesn't like the baseline classic score.
Scores for future analysis post PLOC report. Balances = Mortgage, HELOC
|TU FICO 8||724|
|TU FICO 4||724|
|TU FICO 9||740|
|TU FICO 8 AU||737|
|TU FICO 4 AU||734|
|TU FICO 9 AU||743|
|TU FICO 8 BC||741|
|TU FICO 4 BC||708|
|TU FICO 9 BC||732|