FICO to Vantage Score Ranges Comparison

[GregB]

The entire concept is a complete waste of time.

One change on a report will often move them in opposite directions.

[cashnocredit]

VantageScore and FICO are both validated scores based on large samples even though VantageScore is perhaps 10% of FICO's market share. As to which is more predictive, who knows but there are some clues. FICO taylors each score to an individual CRA since CRAs historically have different levels of reporting completeness. This means that different algorithms are in place at each CRA even for the same generation of FICO score. This produces the most accurate risk estimate for the folks in each CRA database but has the counterintuitive effect of having more variation between each CRA for the same person. VantageScore tries to make scores more closely match between CRAs by ignoring data that often differs between CRAs. For instance, a year ago my VantageScore was at 92% (better than) while FICO was at %39 (better than).  This was due to VantageScore not factoring in a lien that FICO did use. I verified this when my Vantage Score didn't change after the lien dropped. This cumulative percentile variance is roughly the difference between 680 and 810 on the FICO scale.

So even if you use the most accurate way to compare scores, the cumulative percentile, they can differ wildly.

[Cdnewmanpac]

Sorry. As others have said, you cannot do it. Their correlation coefficient is close to zero. In addition to differences in how credit limit is factored in, vantage ignores authorized user accounts. Neither scale is linear and neither publishes their curve. This is like trying to determine the relationship between the stock market and television ratings. If you make the sample size small enough, a correlation will appear. But it will be the result of random chance, not because an actual relationship exists.

[hateAshbury]

Someone explain, either logically or empirically, how two systems purportedly designed to measure the same thing -- a borrower's creditworthiness -- could possibly be uncorrelated. That makes no sense at all. Celsius and Fahrenheit are two different scalings of the same measure: temperature. Does anyone doubt that the two are highly correlated?

But again, if people were to post their two scores, the question of correlation could be answered empirically.

[cashnocredit]

---

@hateAshbury wrote:

Someone explain, either logically or empirically, how two systems purportedly designed to measure the same thing -- a borrower's creditworthiness -- could possibly be uncorrelated. That makes no sense at all. Celsius and Fahrenheit are two different scalings of the same measure: temperature. Does anyone doubt that the two are highly correlated?

But again, if people were to post their two scores, the question of correlation could be answered empirically.

---

There is little doubt VantageScore correlates, in the mathematical sense, with FICO scores but the general public uses the term more loosely. People often confuse differences of scale and offset with being uncorrelated. For instance the C and F temperature scales, while different, are mathematically perfectly correlated (R==1). Since both FICO and VantageScore are validated credit scores that predict serious delinquency risk on the basis of extremely large numbers of CRA report samples they simply have to be "correlated."  But here's the rub. Scores can be highly correlated yet vary tremendously for any given individual. These variations are, for the largest, due to corner cases where the statistical sample is small and major factors are either included or not. Vantagescore methodology seeks to ignore factors that are likely to differ significantly in credit reports. For instance older liens and AUs may not be factored into VantageScores but are in FICO scores. This can distort VantageScores in the 10% or so of folks with PR liens.

That long history AUs tend to raise FICO scores spawned a cottage industry of folks selling AUs to people in need of higher credit scores. This is done with zero risk exposure since the AU never even gets their card let alone can charge anything. It really is gaming the system. Newer FICO scores are dealing with this to a degree but the older ones are the main ones still in use and are still being gamed. Since FICO scores are the mainstay of the mortgage industry, FICO risk model change is hard and slow. Do the widely used "custom scores" that FICO helps create with CC lenders look for AU gaming? I don't know but I can guess.

Anyway, both of these are corner cases and people with one or the other can expect significant difference in scores even after rescaling. But there are other differences as well such as the fact that FICO doesnt factor in total CLs but only ratios. VantageScore, and some of the FAKOs do and high limits bumps these scores. Is it really predictive? I don't know. There is another problem. The fact scores are 3 digit numbers conveys an unrealistic expection that a score is some sort of precise measure. By it's very nature it isn't. It attempts to be an estimate of your future risk based solely on some information in your credit report. It has been a long slow slog to produce good predictive scores. FICO has decades of experience developing and refining these models and there are some measures of how well they work. For instance, the better a model is the higher the ratio of projected risk is between the top quartile and the bottom quartile when the algorithms are applied to large, seperately selected samples. These have increased as FICO comes out with new generations.

So again, when I was in the 90% on VantageScore and 40% on FICO it was because of factors that made me an outlier statistically. As for which was right, I can say the risk of me defaulting in the next 2 years is way, way under 1% since I'm retired and have income and savings that far exceed my yearly expenses. These, however, are factors neither FICO nor VantageScore can glean from my credit report or anyone elses.

None the less FICO remains the gold standard. Remember, when you see a score go down or up 3 points, or even 10 or 20 don't ascribe a lot of significance to it. It's not a label on how good you are as a human being. It's just a very fuzzy estimate of risk using large numbers of other people's histories.

[crunching_numbers]

There is definately some factoring of AU in Vantage.  I removed $29K of AU debt from my TU CR, creditkarma gave me a 61 point jump in my Vantage score, from 804 to 865.  This was entirely the dropping of the 3 accts.

[cashnocredit]

---

@crunching_numbers wrote:

There is definately some factoring of AU in Vantage.  I removed $29K of AU debt from my TU CR, creditkarma gave me a 61 point jump in my Vantage score, from 804 to 865.  This was entirely the dropping of the 3 accts.

---

Could be. The original VantageScore didn't include AU accounts but apparently the most recent version 2.0, does. Nice to know CreditKarma's VantageScores are the latest version.

http://www.creditcards.com/credit-card-news/jeremy-simon-bad-credit-authorized-user-card-1508.php

"Due to requests from the market, the company began including authorized user  accounts when it launched the second version of its score, known as VantageScore  2.0, in October 2010, Burns says."

[marty56]

---

@GregB wrote:

The entire concept is a complete waste of time.

One change on a report will often move them in opposite directions.

---

+1

[MattH]

---

@hateAshbury wrote:

Someone explain, either logically or empirically, how two systems purportedly designed to measure the same thing -- a borrower's creditworthiness -- could possibly be uncorrelated. That makes no sense at all. Celsius and Fahrenheit are two different scalings of the same measure: temperature. Does anyone doubt that the two are highly correlated?

But again, if people were to post their two scores, the question of correlation could be answered empirically.

---

A full answer to your question would get into fairly advanced statistics such as covariance and correlation, but here is a simplified explanation. Suppose FICO scores and Vantage scores each captured about 80 percent of the overall variation of credit risk among individuals. I dunno the actual degree of correlation, what a statistician would call "r-squared," but I doubt it could be higher than 80 percent and it might be a lot lower. Well, if the r-squared for each model is about 80 percent then since they are independent implementations the r-squared FOR PREDICTING ONE SCORE GIVEN THE OTHER would be about 64 percent. Now if you are a lender using Vantage Scores, you don't really care how well Vantage scores predict FICO scores, you only care how well Vantage scores predict credit risk. But if you are a consumer trying to predict how a lender who uses FICO scores will judge you, then your Vantage score is of very little value for that purpose. "A=B" and "B=C" imply that "A=C" because equality is transitive, but APPROXIMATE EQUALITY MAY NOT BE TRANSITIVE.

[cashnocredit]

Lol.

And I have yet to see someone say: "my FICO is 750, what would my VantageScore be?"