Credit Score Spreadsheets

[Anonymous]

---

@Anonymous wrote:

Thanks for catching that -- this is a bug in Google Sheets when I add a new row and it doesn't update formulae.  I told google about it last week and hopefully they fix it.  I'll correct the spreadsheet when I get to WiFi next!

---

Sweet.  Besides that, it's nice.  Thanks for putting it together.  Helpful.

[Anonymous]

I just started porting it to a website instead, one that isn't tied to your email address or identity.  Just a username and password and if you lose it, oh well.

The website will be way better because you can just copy and paste data from your creditors and when an equation updates on the back end you don't need to create a brand new copy of the spreadsheet and repopulate it.

Plus it'll be fun to let people enter more data points optionally and start looking for some connections.  If they apply for new credit, all the data should already be in the system so they can submit a new credit pull based on what their data shows and share it with others.

Also will have a "rebuilding" section to list your collections, charge-offs, what state you're in, what the SOL is, and maybe even help people write disputes based on crowd sourced data.  I also think I can make an actual REAL FICO simulator that will let you know pretty accurately what disputing off derogatories may do for you over time.

My PHP and JS skill set sucks but I think I can do it pretty quickly since it's all simple math and lookups and database stores.  Since I will never ask for an email address, account numbers, etc, it should be relatively anonymous and if people want to give fake account names, they can do that as well (although real account names will help with crowdsourcing data like CLIs, CLDs, AA, etc).

[Anonymous]

---

@Anonymous wrote:

I just started porting it to a website instead, one that isn't tied to your email address or identity.  Just a username and password and if you lose it, oh well.

The website will be way better because you can just copy and paste data from your creditors and when an equation updates on the back end you don't need to create a brand new copy of the spreadsheet and repopulate it.

Plus it'll be fun to let people enter more data points optionally and start looking for some connections.  If they apply for new credit, all the data should already be in the system so they can submit a new credit pull based on what their data shows and share it with others.

Also will have a "rebuilding" section to list your collections, charge-offs, what state you're in, what the SOL is, and maybe even help people write disputes based on crowd sourced data.  I also think I can make an actual REAL FICO simulator that will let you know pretty accurately what disputing off derogatories may do for you over time.

My PHP and JS skill set sucks but I think I can do it pretty quickly since it's all simple math and lookups and database stores.  Since I will never ask for an email address, account numbers, etc, it should be relatively anonymous and if people want to give fake account names, they can do that as well (although real account names will help with crowdsourcing data like CLIs, CLDs, AA, etc).

---

Feel free to ping me if you need any help on the database side.  I'm a SQL and DB guru fortunately.

[Anonymous]

Awesome I may do that, thanks!  I am the world's worst programmer but I have made some really popular web apps in the past that I hear are still used today.  It's fun for me to write things that I will use and then just let other people use them as well.  This one could be really exciting because the crowdsourced (optional) data might unlock some new tricks no one realized, plus the option to let people peek at your data with your approval (again, anonymized, etc).  A temporary share link (say 24 hours) to show people your data so they can give you advice might be really useful on this and other forums, too, but without forcing people to put in any data they don't want public or even private.

Plus if people are really accurate on entering data including derogatories, and they share their MyFico or CCT pull scores with that data, we might end up having a remarkable FICO score simulator without having to do any real tracking work.

[Adidas]

This is a really interesting idea making a website & database to store credit scores and credit profiles. It would be beneficial to figure out more details on how scoring works. I wonder how FICO will respond to you trying to reverse engineer their scores though. Just a word of caution. I still think you should go through with it but they may eventually shut you down depending on how you go about it.

[Anonymous]

I already spoke with my IP lawyer about it and he said that caselaw should side with me since reverse engineering something I use and that effects my life should be well covered.

The reality is that reverse engineering FICO has never been publicly attempted AFAIK. I would believe one would not only be victorious in litigation but would end up forcing disclosure of trade secrets during discovery.

In the event of litigation I bet crowdsourcing funds for defense would be realistic, too. Anyone harmed by an opaque FICO score would love to stab back and pitch in $20.

[Revelate]

---

@Adidas wrote:
This is a really interesting idea making a website & database to store credit scores and credit profiles. It would be beneficial to figure out more details on how scoring works. I wonder how FICO will respond to you trying to reverse engineer their scores though. Just a word of caution. I still think you should go through with it but they may eventually shut you down depending on how you go about it.

---

It's really not any different than the work I and a few others did trying to figure out installment utilization.  That was only a small part of the algorithm, but we effectively did figure it out, and it's since been advertised to the advantage of many with the thread CGID put together on the SSL trick.

I had looked at doing a credit pulls database but got distracted, as long as there's no money being made off this it's OK or at least that was what I last got back from FICO  when I'd inquiried  (note this was a year or two ago, and I'm not a FICO employee just a volunteer forum moderator so my opinion means nothing officially, I'm just relating what was stated to me).

FWIW I don't think FICO much cares, it's not unlike what CreditXpert and everyone else that has tried to figure out the algorithm for a simulation purpose, and that's what we're doing here.  It's just not a threat to their business model.

Also and this may just be the negative nelly in me, I'm not certain we're going to get a substantial amount out of this unless we can correlate a lot of raw data... we've had some goofy things said on the forums and some data of questionable quality in the past, not certain if this is going to be able to weed that out.

[Anonymous]

Humans shouldn't actually touch crowdsourced data, though.  Collaborative filters deal with it -- and they work by making non-obvious connections to disparate data sets with structures.  So if 10 people submit data on their CRAs with a given FICO, if some of that data is impure or just submitted wrong, the filters should still work as long as MOST of the data is correct.

It would just be interesting to SEE what kinds of data are submitted and see what error rate the collaborative filter algorithm objects to.

I've done collaborative filtering for things like diet and food intake (you can't truly measure a calorie and food labels are fraudulent guesses) and those worked great.  I've also done collaborative filtering for things like vacation/travel pricing and that information is even more disparate and that worked really well for the market targetted.

Since the data sets are all unique but structured well (1 account, here's an array of lates and dates), I think crowdsourced data would spit out some relatively useful data.  Not perfect, of course, but useful.