Most people evaluate a settlement offer the way they'd evaluate a price tag: is it high enough, yes or no. That instinct treats the offer as a final answer rather than what it actually is — one branch on a much larger tree of possible futures. The offer arrived because the other side ran some version of a calculation about what happens if you don't settle. Reading it well means running that same calculation yourself, in reverse.

A game tree forces you to make explicit what your gut is doing implicitly: mapping out the sequence of decisions and chance events that lie ahead — motion practice, discovery, trial, appeal — and attaching rough probabilities and costs to each branch. The offer sitting in your inbox is only meaningful in comparison to the weighted average of everything that happens if you decline it.

The most reliable way to evaluate an offer is to work backward from the endpoint. Start with a plausible trial outcome and its rough odds under something like the preponderance standard, then attach a damages range. Discount that figure by the probability the case even reaches a verdict — most don't, because dispositive motions like a motion to dismiss or summary judgment prune branches long before trial. Each of those procedural stages is itself a probabilistic fork: win the motion and the tree shrinks favorably; lose it and settlement leverage often drops.

Layer in the cost of getting from here to there. Discovery is rarely a flat fee; it's a branching set of disputes over scope, cost, and time, each with its own drag on the client's economics. A contingency-fee arrangement changes whose risk tolerance matters most at each node, since the lawyer and client may value delay and uncertainty differently. Once the tree is built, you're not asking 'is this offer good?' You're asking 'is this offer better than the probability-weighted, cost-adjusted value of everything downstream?'

An offer is also a signal. In a setting of imperfect information, neither party can see the other's private assessment of liability or damages, so offers function as noisy hints about what the other side believes their own case is worth. A low early offer might mean genuine confidence in their defense, or it might mean they're testing your resolve before investing further. A number that arrives right before a costly discovery deadline often has less to do with case merits and more to do with the other side's own budget dial.

This is where bayesian updating becomes useful even informally: each offer, each delay, each unexpected motion result should shift your estimate of the other side's private view, not just your own. If an offer moves substantially after a single piece of discovery, that tells you which fact mattered to them. Reading the offer's timing and size together, rather than the number alone, is often more informative than the number by itself.

Every settlement evaluation ultimately reduces to a comparison against your best alternative to a negotiated agreement — your BATRA, the litigation-specific cousin of BATNA. If the offer exceeds the risk-adjusted value of continuing to litigate, decline is expensive even if the case feels emotionally strong. If it falls short, accepting may be leaving value on the table, but only if your estimate of the downstream tree is realistic rather than hopeful.

This is also where a prisoner's-dilemma-settlement dynamic shows up: both sides might be individually better off pushing forward to extract more information, yet collectively worse off doing so once fees and delay are counted. Recognizing that trap doesn't mean always taking the first reasonable number — it means being honest about whether further conflict is being driven by expected value or by momentum.

Building this tree by hand is useful but limited — most people quietly round toward whatever number confirms their existing position. A structured simulation approach, running a decision-tree analysis with many seeded variations of the key uncertain inputs, exposes how sensitive your valuation actually is to the assumptions you're least sure about. A sensitivity analysis on damages range, liability probability, and cost-per-stage often reveals that the 'right' settlement number isn't a point at all — it's a band, and the offer's position inside or outside that band is what matters.

This is close to how Juricratic models a matter: as a solvable game with dials for the variables litigants actually argue about, run through many simulated paths rather than one intuition-driven guess. The output isn't a prediction of what a court will do — it's a map of where your position is exploitable and where it's well-defended, which is exactly the information you need before deciding whether a number on the table is worth taking.