Liquidation value only, or probability weighting the workout paths, for non performing notes

I'm getting my first rental ready so I'm out of my depth, but where do the 35 and 25 percent come from? If they're from eleven files then the model isn't wrong, it just has enormous error bars on it. Wouldn't you rather say "somewhere between 20 and 50 percent" and see if the bid survives the low end?

Model A with a lower required multiple. You keep the discipline of the floor and you stop pricing yourself out. A 1.7x on liquidation is a very high bar when the loss cases in your own Model B still net $62,000. Take 1.35x on the collateral and let the workouts be upside you do not pay for.

For anyone reading this who's newer, the two models have plain names. Liquidation underwriting asks what you'd get if the only thing you did was take the house and sell it. Probability weighted underwriting asks what you'd get on average across every way the file might end, with each ending multiplied by how likely you think it is.

A couple of the terms ledger uses. A reinstatement is the borrower catching up the missed payments and going back to paying normally. A modification, or mod, rewrites the loan terms so the payment is affordable again, which turns a defaulted note into a performing one you can hold or sell. A discounted payoff is the borrower paying less than the full balance to settle and be done, usually by refinancing elsewhere or selling the house.

The reason the room splits is that liquidation underwriting can't be wrong about probabilities because it doesn't use any. It uses value, costs, and time. It can still be wrong about the value of the house, the length of the legal process, and how much the property has deteriorated since anyone looked inside.

Probability weighting captures how these files really end, and the accuracy of the answer depends on the accuracy of the inputs, which come from your own history of resolved files. With eleven files, those inputs are early estimates.

One thing both models share: the cushion between what you pay and what the collateral is worth is what protects you when a path goes wrong. That number is worth knowing before anything else.

Your Model B has a hole that has nothing to do with probabilities. The 35 percent mod case assumes you can sell reperforming paper at 75 cents in eight months. Seasoning matters to those buyers, and most want to see 12 payments made rather than 8. So either that case takes longer or it clears at a lower price. Fix that line and your weighted number drops toward $73,000, which changes what the $58,000 ask means.

I'd add a fifth path you didn't model: borrower does nothing, ignores you completely, and the house turns out to be worse inside than the exterior photo. That's not a 10 percent tail in my experience, it's more like a fifth of files, and it puts a repair number into the foreclosure case that your $86,000 assumes away.

To be blunt about the practice rather than the theory, the disciplined approach is to run both models on every file and bid the lower of the two. Model A sets the price. Model B tells you whether to spend attention on the file at all, because a file where the weighted number is barely above the liquidation number has no workout optionality worth chasing, and the one worth buying is the one where the gap is wide. Two models, two different jobs.

The documents decide more of this than the probabilities do. A clean chain with a recorded assignment and the original note in hand supports every path in Model B. A broken chain kills the foreclosure case and the discounted payoff at the same time, because standing cannot be proven to either a court or a title company. It kills them at month nine, with the legal bill already spent. Recording requirements and what a court accepts as proof vary by state, so that check has to be state specific.

Warrant's point is the one that trips up most people new to how this capital moves. Both of ledger's models assume the file is enforceable. Neither has a line for it.

I do a lot of due diligence for passive positions and what I'd want from an operator is exactly the pairing sextant described, plus the actual resolution history. Eleven files isn't a probability set, it's an anecdote. Show me forty and I'll believe the weights. Until then the collateral floor is the only line I'd underwrite to.