Nicely done experiment and well written report! It supports what I've often seen stated on this forum, and also experienced myself in the past in a similar experiment: EQ in particular hates (or at least used to hate) all-zero balances. In my little experiment, done about two years ago, I alternated several times between total balances of $0 and $10 (on one card, didn't matter which), and got my EQ and TU FICO scores after each change. It was quite reliable: with the $10 balance total, I had 813 FICO scores on both EQ and TU; with zero balance, TU went up to 819, EQ down to 809. That was exactly what I had seen predicted here.
But recently, it seems to have changed. I prefer not to bother with the small residual balance unless I have some reason to maximize my score, so I've been at all zero for a number of months. I got an offer from FICO for reduced-rate scores, and I realized I hadn't actually pulled mine for some time, so I took the offer even though I hadn't had time to get the $10 residual in place. (Yeah, sucker for a bargain.) And I found just the opposite behavior as previously: EQ was higher (807) than TU (799). The lower scores than previously didn't surprise me;, since I recently took on a larger mortgage since the 813 etc. reported above. But I was surprised that TU seemed more sensitive to the zero balances than EQ. And the infamous "factors" text confirmed this: For TU, "Your FICO score was hurt because you are not currently demonstrating active revolving credit management.", while EQ thought I was good on that (and the other) criteria. So I'm wondering whether anyone else has noticed this sort of change.
By the way, these scores, just above and below 800, give an interesting view of how nonsensical the "Your risk to the lender" report can be. Supposedly, TU thinks I'm twice as likely to default (2%) than EQ does (1%). I know, it's all statistical, and there have to be boundaries, but such a proportionally enormous difference can't be realistic in an individual case.
Starting Score: EQ 804 - (April 2009) 