There are many ways. There are also many different "real" FICO models. Don't just rely on oversimplifcations such as "real". Which specific FICO mode do you need/want?
rohnin wrote:seems they are all much higher than the ones the actual lenders get when they do a hard pull
It's not given that a creditor will pull a higher number. Again, there are many scoring models used by creditors/products -- not just one. The specific model will impact the number generated. You cannot assume that all models are equivalent and must always consider the specific model and CRA when referencing scores.
flan wrote:If two models are reasonably accurate
"Reasonably accurate" is arbitrary. Each model evaluates report data differently. One cannot assume that one model is accurate and another is not. They're just different interpretations of report data.
flan wrote:they will produce similar scores at the high and low ends, because that's the people least and most likely to default, which is what the scores are measuring.
I see what you're getting at. Sorry, I tend to end up in autoreply mode after hitting MF as everyone asks the same questions over and over and over.
However, even your use of accuracy isn't about the similarities or differences in scoring models. it's about how well the model predicts whether or not one will default. That is entirely separate from how similar two models may or may not be unless one model has been determine to be able to actually predict default. Do you have studies to cite that indicate accuracy of the scoring models or is this just academic?
From a practical standpoint, accuracy doesn't mean anything when comparing models. In any case, it's popular usage that is my concern here. One model is not accurate while all others are not despite the tendency of people to assume that there is a single accurate model. Proving whether or not a given model can predict default is not what I'm addressing. Ideally a model should be accurate.
What people are really asking when they inquire about accuracy is "Can I rely solely on this score?" and the answer is no. The credit world doesn't rely on just one model and no one model will accurately represent every scoring model in use today.