@Thomas_Thumb wrote:
@SouthJamaica wrote:
@Thomas_Thumb wrote:I started trending the difference between my Fico 8 and Fico 9 scores to see if anything sticks out. My profile is clean so I don't see any positive impact with Fico 9 relating to paid collections. I have no inquiries or lates so those don't factor into score differences. My youngest account is over 4 years old.
Best I can tell is Bankcard enhanced Fico 9 puts significantly more weight on depth of file - meaning # of accounts than does Fico 8. My file is rather thin (9 accounts total with 7 open. One open account is a mortgage - 11 years into a 15 year loan). Auto enhanced and Classic Fico 9 don't appear to put added weight on # open accounts relative to their Fico 8 counterparts.
In August 2016 I had 0% utilization with no revolving accounts reporting a balance. An AU card did report a balance but, I have confirmed it is not counting toward my Fico 8 score. The zero revolving accounts reporting caused a significant drop in Fico 8 scores. Minimal, if any drop on Fico 9 suggests Fico 9 is either counting the AU card or it weighs 0 % utilization/zero cards reporting very little. This is illustrated by dip in the (F8 - F9) score difference.
1. What I can't get over about your graphs is that most of the action takes place over zero, whereas mine would take place entirely under zero. My FICO 9 scores are always significantly higher than my FICO 8 scores. Your August 2016 disturbance would be the norm for me for all 3 scores in all 3 scoring model families.
2, The other thing I can't get my head around is the flat lining, signifying a total absence of difference between FiC0 8 and FICO 9, represented by the blue line.... i.e. your Auto & Bankcard scores bouncing around, while your classic scores hardly budge.
3. Since you attribute the differences, whatever they are, to number of accounts reporting balances, my questions are: (a) is there a way to simply create a more or less balanced optimization plan for all 6 scores and (b) if so what is it?
Hello SJ:
To try and answer your questions.
1) I believe Fico 9 wants to see more open and closed accounts in ones file than I have. I rate only very good for credit mix due to too few accounts in total - it's not the # of accounts showing activity. The lack of accounts is not hurting my Fico 8 bankcard score - still can reach 900 on TU and EX. However, on Fico 9 max I can get is 885. So, the limitation in Fico 9 is lack of critical mass. I see the same thing with VantageScore 3.0.
2) Reason my Classic scores show zero difference is the buffer I have at 850 - IMO. I did have TU Classic Fico 8 drop to 845 once while Classic Fico 9 stayed at 850. Thus, the -5 points.
3) When I reduce # cards reporting balances to 2 or 3 from 4, 5 or 6 all my all my Fico enhanced scores (8 and 9) go up a bit. It's just that the Fico 9 scores bump up against some kind of ceiling - which I attribute to unattainable mix points for my file due to too few accounts.
You have over 20 accounts as I recall and that may be optimizing your Fico 9 "mix points" - by comparison I have dropped from 11 - 12 to 9 total which hurts my mix.
Yes I have a lot of accounts, around 25.
Since the most notable thing about my current profile is its newness and aggressiveness -- it shows lots of inquiries and recent accounts, and none of my old stuff -- it seems likely that my consistently higher scores on the model 9 score versions are attributable to that. It seems to me that inquiries and recent accounts are less concerning to the model 9 versions than to the model 8 versions, and less concerning to the 8 versions than to versions 2 to 7, which are consistently lower than my FICO 8 classic and enhanced scores.
FYI - Below are some Fico model comparison charts from a TU article:
As a whole I suspect clean, somewhat thin, profiles with no new accounts and Fico 8 scores in the 800s may experience lower Fico 9 scores. Conversely takne as a whole, clean/thicker profiles with scores in the 700s may see a score increase with Fico 9 (relative to Fico 8). Mildly dirty profiles with paid collections likely represent the +50 point increase in Fico 9 scores relative to Fico 8 and relative to Fico 4.
@Thomas_Thumb wrote:FYI - Below are some Fico model comparison charts from a TU article:
As a whole I suspect clean, somewhat thin, profiles with no new accounts and Fico 8 scores in the 800s may experience lower Fico 9 scores. Conversely takne as a whole, clean/thicker profiles with scores in the 700s may see a score increase with Fico 9 (relative to Fico 8). Mildly dirty profiles with paid collections likely represent the +50 point increase in Fico 9 scores relative to Fico 8 and relative to Fico 4.
I'm no expert in statistics, and try as I might, I can't understand what these charts say.
With a gun to my head my guess would be:
(a) they shed no light on the algorithm differences among the 3 scoring models portrayed
(b) they show that sometimes FICO 9 is higher than FICO 8 and FICO 4, and sometimes it's lower, and that it's slightly more likely to be higher than lower on TU data
(c) the higher your overall scores, the higher your FICO 4 is likely to be relative to the others
@Thomas_Thumb wrote:SJ - you are correct in that the difference graph shows Fico 9 score can be quite a bit higher or lower than Fico 8 and Fico 4 scores. It clearly suggests that results are highly profile dependent so testing of an individual profile can not be extrapolated to the general population. That being said, I do believe the 0 point difference distribution interval is twice as high for TU (Fico 9 - Fico 8) relative to TU (Fico 9 - Fico 4) is due to profiles that have both Fico 8 and Fico 9 850 scores. An 850 score is not possible with TU Fico 4.
Here is what the interval score distribution graphs indicates:
1) Scores above 839 have not been achieved on TU Fico 4. Scoring 820 or above on TU Fico 4 is rare - less likely than scoring above 840 on TU Fico 8 or TU Fico 9.
2) About 4% of the population achieve Fico 9 scores at 840 and above compared to less than 2% for Fico 8 and 0% for Fico 4. So Fico 9 overlooks more shortcomings.
3) Major shortcomings (derogs) are scored more harshly on Fico 9 as evidenced by the higher % of scores under 500 compared to Fico 8 and Fico 4.
3) The general population is more likely to have a score in the 800 to 819 range than in any other interval score segment. This is true for all three Fico models.
4) TU Fico 4 has a soft ceiling under 820. Thus, the high percentage of scores in the 800 to 819 range.
P.S. A key point about the difference chart is it clearly shows scores being skewed upward with Fico 9. This anticipated outcome was publicized when Fico 9 was initially rolled out.
Thank you T_T!
I'm always learning from you.
That's an interesting find, thank you TT!
