Truth: It feels like just yesterday I launched this blog. Dare: But several years have passed, and it’s time to face the music and reveal how our bank probability of failure predictions have held up. When we played this game as a kid, didn’t it always seem we learned a lot more about people in a dare?
Since the last update, we’ve seen a handful of bank failures—thankfully, a very limited number. Good news for Americans, but still enough to keep this game interesting. My goal remains the same: to keep you informed through regular back‑testing of our ratings and predictions.
In earlier posts, I’ve shared the truth about our historical numbers: most failures come from D‑rated institutions (we don’t give out “F” ratings—if a bank earns an F, it’s already in receivership and has failed the dare). I’ve also talked about the few banks that managed to slip through the cracks.
Now it’s time to spin the bottle and see where it lands.
After months of compiling data and running regression analysis, my first back‑test was a thrill—I had pinpointed the worst bank in America at that moment. I checked the failure list… and sure enough, it had gone down. A direct hit. The chart below shows how things have gone since then. Sure, we’d love a perfect score, but anomalies happen. And every miss deserves an explanation.
From our dataset of 385 banks—random selections plus known failures—90% of failures came from D‑rated banks. Six failures came from banks rated in the “A (+/-)” range, several tied to fraud. I’ll walk you through where we stumbled, why it happened, and what we’re doing to avoid repeating those mistakes.
Truth: I wasn’t thrilled with the results. Dare: Let’s dig in anyway.
Round 1: The Easy Truths
The first two banks? Slam dunks. Over 99% probability of failure, “D” ratings, and composite estimated CAMELS of “4”—the next to worst possible. These were the “glaring truths” of the game: obvious, straightforward, no surprises. I included one of the bank’s charts below.
Round 2: The First National Bank of Lindsay — The Trick Question
Then came The First National Bank of Lindsay—the dare disguised as a truth.
For a long time, their ratings looked great. The oddity? A negative Allowance for Credit Losses (ACL). That’s like someone saying, “I dare you to believe everything is fine,” while hiding a ticking bomb behind their back. Negative ACLs are almost unheard of—AI searches show maybe 0–3 banks nationwide with this anomaly. In decades as an examiner, I had never seen it.
Regulators didn’t have any formal or informal actions in place, which would have raised their failure probability from 3.9% to around 54%. They missed it too.
And then came the truth behind the dare: Fraud.
Copilot searches revealed:
- Fraudulent loans to friends and associates
- Manipulated bank records
- Obstruction of federal examiners
- Failure to implement anti–money laundering controls
This wasn’t a miss on our part—it was a rigged game.
Round 3: Pulaski Savings Bank — Déjà Vu Dare
Pulaski Savings Bank felt like the sequel nobody asked for. Fraud hasn’t been officially confirmed, but the irregularities were suspicious enough to raise eyebrows. Despite an estimated CAMELS of “2,” the bank had:
- $20.7 million in deposits missing from the core system
- Large off‑system deposits
- Suspense account discrepancies big enough to swallow the bank whole
Once the missing deposits were finally recorded, equity evaporated instantly. Another dare wrapped in a truth.
Round 4: Metropolitan Capital Bank & Trust — The Obvious Dare
This one was easy. High bank probability of failure, “D” rating, no fraud—just a straightforward collapse. No plot twist.
Round 5: Community Bank & Trust — The Dare That Lied
Community Bank & Trust of West Georgia looked stressed for several quarters, then suddenly cleaned up its balance sheet. It was the classic “I dare you to trust me” moment.
We took the bait.
But like SVB and others with large unrealized investment losses, the bank’s apparent improvement masked deeper issues. When the truth came out, it wasn’t pretty. And it wasn’t a good look for our analysis.
Truth: We missed it. Dare: We learn from it.
In summary, we have some work to do in order to accurately predict bank probability of failure. I’m not going to lie. I am not surprised by the anomalies, and eventually the numbers will “bare out” the truth. Because isn’t that what we are looking for when we play truth or dare, the bare[naked] truth. We plan on incorporating a negative ACL into our analysis. As for the SVB affect, we may actually have a bit too much emphasis on unrealized losses. However, if this is the case I think this anomaly will work itself out over time as the data set increases and the number of failures due to this anomaly smooths out. And, of course the big one is fraud. We are working on this as well. Forensic accounting is one method, and we have already adopted a tool we feel will be helpful. We will talk more about this in the next blog.
Leave a Reply