What is the FICO score baseline?

I was reading this thread EX8 AU Termination Point Loss and saw this post where the subject of whether FICO scores start at 250/300 or 850/900 was broached by @Justaguy . I am a bit befuddled by the idea that a credit file would start at either extreme of the scale. It would make more sense to me to start the score at a set baseline then add and/or remove appropriate points as dictated by the individual factors of the record. Otherwise, we would need to identify where, literally, hundreds of points come from when a new credit file hits 6 months and is scored for the first time. Is "first account" worth 200 points? "First inquiry" worth another 100 points? "First on time payment" worth 50 points? Or, maybe 6 months AoOA is worth 300 points? If there are some baked in points gained for "first" this and that or a huge point allocation for hitting 6 months, isn't that really just saying "all files begin with this many points then lose and gain from there"? The same argument can be made from the perspective of starting with a perfect score and losing ~150 points off the bat. What do we know about this?

This is definitely a classic glass half full / glass half empty argument.

I tend to go with the top score argument then reduced to whatever it ends in due to profile data. The reductions are score reducing factors, highlighted by the top 4 (or however many) negative reason statements provided. Since we focus on the negative reason statements, that's why I view it as a high score starting point being suppressed. We know that positive reason statements are just CMS fluff and while they may tell some of the story they aren't directly tied to Fico scoring.

BrutalBodyShots wrote: ↑Sat Jul 02, 2022 10:33 am This is definitely a classic glass half full / glass half empty argument. I tend to go with the top score argument then reduced to whatever it ends in due to profile data. The reductions are score reducing factors, highlighted by the top 4 (or however many) negative reason statements provided. Since we focus on the negative reason statements, that's why I view it as a high score starting point being suppressed. We know that positive reason statements are just CMS fluff and while they may tell some of the story they aren't directly tied to Fico scoring.

I agree this is a fascinating thing to contemplate. @Birdman has strong opinions, much knowledge and keen insights on the subject (shocking, right? ). I am drawn to the idea the FICO scoring algorithms start at the top score and subtract points for flaws, but I don’t think the evidence is strong enough to say we know for sure. I don’t necessarily share the conventional view that positive reason statements are meaningless fluff. MF user IV (who is awesome!) found some contrary evidence when hacking the MF reason codes (positive and negative). I’ll try to find and link to that fascinating thread. @Have1, I love how your mind works! I had always thought about this issue as only having two possibilities. I assumed the algorithms must either start at the bottom and only add points for good things, or start at the top and only subtract points for bad things. You are suggesting a third option: Start somewhere near the middle and do some adding and some subtracting. Brilliant!

Justaguy wrote: ↑Sat Jul 02, 2022 12:32 pm I don’t necessarily share the conventional view that positive reason statements are meaningless fluff. MF user IV (who is awesome!) found some contrary evidence when hacking the MF reason codes (positive and negative). I’ll try to find and link to that fascinating thread.

Please note that I didn't use the word "meaningless" before fluff While I do feel that much of CMS fluff is indeed meaningless, I don't consider positive reason statements to be completely meaningless. More often than not from my experience the positive statements seem to align with the converse of the negative ones. For example on a clean file instead of seeing a negative reason statement pointing to a recently missed payment one may see a positive statement saying no missed payments or something similar. I don't see that as a bonus for on time payments, but rather the lack of a penalty/points being lost for the presence of a negative item. That thread does sound interesting and I'd enjoy checking it out!

This is a slide that was shown to lenders during a presentation by FICO itself. The presenter was quick to point out that this is an example only, and not necessarily the actual number of points the algorithm is using. Note that each attribute value has an optimal maximum value, and it's the difference between that value and the points received that contributes to reason statement order. ( #1, #2, #3, #4 in the margin.) Personally, I think the algorithm starts at 0 (null or void) and starts running through the characteristics depending on what's on file. (A mortgage is not required to hit 850 - no need to even look at those attributes if a mortgage isn't on file.) ( This is also in the Credit Scoring Primer here. )

Cassie wrote: ↑Sat Jul 02, 2022 1:44 pm Personally, I think the algorithm starts at 0 (null or void) and starts running through the characteristics depending on what's on file. (A mortgage is not required to hit 850 - no need to even look at those attributes if a mortgage isn't on file.)

I've seen that slide and found it very interesting. It makes me want to get my pencil out and try to predict my score change when I lose my Amex account. But are there enough points to build from the bottom to where new scores usually appear? That would have to be a lot of points. Don't most new scores appear in the 680 range? If it does start from null, wouldn't a lot of those points have to be gimmees? So every new record would be guaranteed to get the same ones simply for existing? Then the profile dependent adjustments would take place?

Have1 wrote: ↑Sat Jul 02, 2022 2:19 pm

Cassie wrote: ↑Sat Jul 02, 2022 1:44 pm Personally, I think the algorithm starts at 0 (null or void) and starts running through the characteristics depending on what's on file. (A mortgage is not required to hit 850 - no need to even look at those attributes if a mortgage isn't on file.)

I've seen that slide and found it very interesting. It makes me want to get my pencil out and try to predict my score change when I lose my Amex account. But are there enough points to build from the bottom to where new scores usually appear? That would have to be a lot of points. Don't most new scores appear in the 680 range? If it does start from null, wouldn't a lot of those points have to be gimmees? So every new record would be guaranteed to get the same ones simply for existing? Then the profile dependent adjustments would take place?

The points can add up quickly when there are big chunks awarded for lack of bad things. No public record? +100 No collections? +80 No serious delinquencies? +70 No high util cards? +40 No missed payments? +30 No accounts ever late? +25 No recent missed payments? +20 No high util loans? +15 No inquiries? +10 No new accounts? +10 I totally made up those numbers, but that would be +400. Starting from a baseline of 300, that gets to 700 without any actual accounts or payments. Seems close enough to make the concept plausible, no?

Justaguy wrote: ↑Sat Jul 02, 2022 3:10 pm

Have1 wrote: ↑Sat Jul 02, 2022 2:19 pm

Cassie wrote: ↑Sat Jul 02, 2022 1:44 pm Personally, I think the algorithm starts at 0 (null or void) and starts running through the characteristics depending on what's on file. (A mortgage is not required to hit 850 - no need to even look at those attributes if a mortgage isn't on file.)

I've seen that slide and found it very interesting. It makes me want to get my pencil out and try to predict my score change when I lose my Amex account. But are there enough points to build from the bottom to where new scores usually appear? That would have to be a lot of points. Don't most new scores appear in the 680 range? If it does start from null, wouldn't a lot of those points have to be gimmees? So every new record would be guaranteed to get the same ones simply for existing? Then the profile dependent adjustments would take place?

The points can add up quickly when there are big chunks awarded for lack of bad things. No public record? +100 No collections? +80 No serious delinquencies? +70 No high util cards? +40 No missed payments? +30 No accounts ever late? +25 No recent missed payments? +20 No high util loans? +15 No inquiries? +10 No new accounts? +10 I totally made up those numbers, but that would be +400. Starting from a baseline of 300, that gets to 700 without any actual accounts or payments. Seems close enough to make the concept plausible, no?

But wouldn't that be a baseline score? One from which the actual score would be derived after adjustments per individual record? "You get all these to start and the credit world is your oyster."

Maybe I misunderstood what you meant by “baseline.” I was using it as the starting point for the calculation within the FICO formulas. The “score” before any data from a particular consumer’s file is added. No one other than the software people would ever see this number. I’m tempted to call this an input to the algorithm, but maybe that’s not quite right since it doesn’t exist anywhere except right at the starting point. It sounds like you might be using “baseline” to describe the score of a squeaky clean, just-enough-data-to-get-a-score, consumer profile. I think of that as an output of the algorithm. It’s a score that an attentive consumer can see. There are some data points around, including one from @Cassie’s friend, showing what this score can be. Sorry if I’m misunderstanding you.

I'm making the point, poorly, that regardless of the algorithm's intent a baseline would exist if all files start from the same number based on the implied assumption that all new files have a lack of most of the scoring factors.

Even if the scoring started at zero, you get all these points because you haven't had a chance to screw up, or excel, yet. That makes for a baseline score.

You will start from there. Gimmees aren't earned, they are free.

I may be making arguments to further the development of theories, here...

Justaguy wrote: ↑Sat Jul 02, 2022 4:18 pm It sounds like you might be using “baseline” to describe the score of a squeaky clean, just-enough-data-to-get-a-score, consumer profile. I think of that as an output of the algorithm. It’s a score that an attentive consumer can see. There are some data points around, including one from @Cassie’s friend, showing what this score can be.

Her first credit score was 750 at 6 months for TransUnion FICO 8. That's coming from a non-existent file. Also, TransUnion was showing a single inquiry, and inquiries cost a lot more on a new file! I saw -15 points for 1 on mine! 1 revolving account - a regular, unsecured Visa bankcard from Bank of America. ( On my brand-new file, with a single creditbuilder loan on file, TWO inquiries, and at 11 months: TU8: 666 !!! Mark of the Beast! lol ) So, yeah, you could call that a 'baseline' score. I like your example, @Justaguy .

Have1 wrote: ↑Sat Jul 02, 2022 5:50 pm I'm making the point, poorly, that regardless of the algorithm's intent a baseline would exist if all files start from the same number based on the implied assumption that all new files have a lack of most of the scoring factors. Even if the scoring started at zero, you get all these points because you haven't had a chance to screw up, or excel, yet. That makes for a baseline score. You will start from there. Gimmees aren't earned, they are free. I may be making arguments to further the development of theories, here...

Taking the last point first: understood. Same here. None of this is personal or acrimonious. We’re just playing with ideas and I love having smart people like you to learn from. Are you perhaps assuming that all profiles that have just enough data for FICO to generate a score are clean? I’m pretty sure they aren’t. Plenty of lenders don’t report to the credit bureaus, but if you don’t pay them back, they will take action that will eventually result in a public record. If that happens before the consumer has a credit bureau file, and they later establish a file, the baddie will be there waiting for them in their first FICO score. So instead of thinking of this in terms of a new profile, whose characteristics we cannot know, I prefer to think of it from the algorithm’s perspective. At the start of the calculation, the score is effectively X. As pieces of data are factored in, the score will change. In which direction? That’s a mystery to me.

Yes as @Justaguy said, I have strong opinions and theories on this LOL. Timmy logic dictates my theory. We are dealing with a mathematical algorithm. Think of a formula in your head. There is no output until there’s input! Zero input could be Noel could be zero could be a lot of things but unless there’s meaningful data input there is no output. Therefore it doesn’t start at zero it doesn’t start at 850. It evaluates the data, iit chooses the scorecard, runs the data through it and there’s an output. Your score is what it is. but it’s also programmed to give up to five negative reason codes if available and yes the positive codes are useful in identifying the negative codes if we don’t know them. They are useful for identifying scoring factors. Lenders are required to give you reasons for denial so fico prepares these negative reason statements with their legal team to pass legal mustar and lender give you the reasons your score is not higher as the reasons for denial. So I Kinda have to agree with @Have1 and his initial idea of it. Now, baseline again is ambiguous as you have noted. If you mean the six-month first score one will get, it depends on how many bankcards/Revolvers they got when they started or loans, along with all the other scoring factors. But assuming everything is optimized, it will depend on the number of tradelines and bankcards. If one starts with three bankcards and a loan, they will have the highest first score possible at six months if they pull the SSL trick. (assuming one cd not start with five bankcards and a loan, which would be too much of a stretch to consider viable, imo, except maybe for an Outlier, lol!) Anyway that’s my belief & my 2 cent theory! And by the way the 250/350 & 850/900 are the differences in the classic and industry standard scores.