How to Use Backward Induction to Plan Trial Strategy
How to work from the trial outcome backward through each procedural stage so today's decisions are anchored to where they actually lead.
Backward induction is the game-theory method of solving a sequential decision problem by starting at the last decision point and working backward, so each earlier choice is made with full knowledge of how later stages will play out. Litigation is naturally sequential — pleadings, discovery, dispositive motions, trial — which makes it a strong fit for the technique.
The practical payoff is a way to answer 'should I file this motion' or 'is this deposition worth the cost' as a question about how it changes a value that has already been folded back from the trial-stage estimate, rather than as an isolated tactical call made in a vacuum.
Map the sequence of decision points to the end
Start by laying out the matter as a simple sequence of stages and branches: a motion to dismiss that is granted or denied, a discovery period, a summary judgment motion that succeeds or fails, and ultimately a trial. Each branch point is a node where the case either ends or continues, and each continuing branch eventually leads to a trial-stage outcome.
Solve the last node first
Backward induction starts at the end of the chain, not the beginning.
- Estimate the expected value at trial — damages multiplied by the probability of prevailing, net of the cost of getting there — conditional on the case actually reaching that point.
- That trial-stage expected value becomes the payoff attached to the 'proceed to trial' branch at the prior node, such as a denied summary judgment motion.
Fold that value backward one stage at a time
With the trial-stage value established, work back through each earlier node, comparing outcomes at that point using the values already solved further down the chain.
- At the summary judgment node, compare the value of dismissal against the probability-weighted value of proceeding to trial you already solved.
- At the discovery node, compare the cost of further investigation against the incremental value it is expected to unlock at the later nodes it feeds into.
- Each fold produces a value for the current node that already accounts for optimal play at every node after it — that is what makes it backward induction rather than a series of isolated guesses.
What backward induction tells you today
The method reframes today's tactical questions. A motion or a deposition is only worth its cost if it measurably shifts a downstream probability or cost figure that has already been folded back through the chain — not because it feels like the aggressive or thorough move in isolation.
Where backward induction assumptions can mislead
The method is only as reliable as its inputs and its assumptions about how both sides behave.
- It assumes both sides play optimally at every future node — a real opponent may not, which can make the folded values overstate or understate what actually happens.
- It assumes dials stay stable across stages, when new evidence at any point can change the probabilities the whole chain was built on.
- The entire fold is only as good as the trial-stage estimate at the root of it — an unreliable trial estimate propagates errors backward through every earlier node.
- Is backward induction the same as a decision tree?
- They're closely related — the decision tree is the structure of branches and nodes, and backward induction is the method used to solve it from the end back to the start.
- Do I need exact numbers at every stage?
- No. Even rough dial estimates let you compare which branch has higher folded value. Use a sensitivity analysis to find out which stage's estimate is worth refining first.
- Does backward induction assume the other side is rational?
- Yes, the classic version assumes both sides play optimally at every future node. If you doubt that assumption for a specific opponent, it may be worth modeling the game with an exploitability gap instead of relying on a clean equilibrium fold.
This page is an educational explainer, not legal advice, and creates no attorney–client relationship. Juricratic is a simulation engine: every probability-like figure is a dial you set, not a calibrated prediction. Verify every rule, deadline, and figure against the authorities and orders that govern your matter.
Stop estimating one number at a time.
Juricratic models the whole matter as a solvable game and runs it thousands of times — so the settlement value, the risk, and the optimal line all move together when the facts do.
Request access →