A decision tree turns a tangle of what-ifs into a diagram you can actually reason about. It separates the moves you control from the events you do not, attaches probabilities and payoffs to the uncertain branches, and then collapses the whole thing into a single number you can compare against a settlement offer. Done honestly, it is one of the most clarifying tools in litigation strategy.

The discipline is in the construction. A tree is only as good as the branches you remember to draw and the probabilities you are willing to defend. This guide builds one step by step, shows how to fold it back, and flags the mistakes that quietly corrupt the answer. It is an educational explainer, not legal advice.

Every tree is built from two kinds of forks. A decision node is a choice you make: file the motion or hold it, accept the offer or proceed, go to trial or settle. A chance node is an event outside your control: the judge grants or denies, the jury finds for one side or the other, a key witness is believed or not. Drawing them with different symbols keeps you from treating a coin flip like a choice.

The order matters too. Lay the nodes out in the sequence they actually occur, because a later decision often depends on how an earlier chance resolved. This temporal structure is what makes the tree a model of the litigation rather than a static list.

Start at the present decision and move forward through the case's procedural spine. A typical civil matter branches at the pleading stage, then discovery, then dispositive motions, then trial, and finally any appeal. At each stage, draw the realistic outcomes rather than only the two you care about.

Every chance branch gets a probability, and the branches leaving a single chance node must sum to one. Every terminal leaf gets a payoff: the net money position at the end of that path, after subtracting the legal costs incurred to reach it. Netting out costs at the leaves is essential, because a path that wins a large verdict after an expensive trial is not worth its gross number.

Be candid that these probabilities are estimates. The value of writing them down is that they become debatable and revisable, rather than hidden inside a gut feeling. When you learn something in discovery, you update the number and re-fold the tree.

Now collapse the tree from the leaves toward the root, a process called rollback or averaging-out and folding-back. At each chance node, compute the expected value by multiplying every branch's payoff by its probability and summing. At each decision node, you do not average; you choose the branch with the higher value, because you control that fork. Carry that value backward until you reach the present decision.

The number at the root is the expected value of the case under your current assumptions. Compare it against the settlement offer in hand: if the offer beats the tree's rolled-back value, and you are risk-averse, taking it is rational. The decision node just below the root tells you which action the analysis actually recommends.

The most frequent error is forgetting branches. If you only draw the outcomes you expect, the tree flatters your case. Force yourself to include the adverse denial, the partial ruling, the reduced-damages verdict. The second common error is double-counting or omitting costs, so that leaf payoffs are inconsistent across paths.

A subtler failure is treating an adversary's move as a chance event when it is actually a strategic choice. Opposing counsel is not a random draw; they respond to your position. This is where a plain decision tree reaches its limit and a game-theoretic model earns its keep. Juricratic pairs the tree with an interactive-game view, letting you vary each probability as a dial and re-fold the tree instantly under a seeded simulation, so you can see which single assumption is carrying the recommendation.