How to Use Bayesian Updating as Discovery Unfolds
A method for revising your case's probability dials methodically as evidence arrives, instead of anchoring to your opening read of the case.
Bayesian updating is the process of starting with a prior belief, weighing each new piece of evidence by how strong it actually is, and combining the two into a revised, posterior belief. Discovery produces a natural sequence of these updating events — depositions, document productions, expert reports — each of which should move your probability dials by an amount proportional to what it actually shows, not by how dramatic it felt in the room.
The goal is avoiding two opposite failure modes: overreacting to a single ambiguous document, and anchoring so hard to your opening theory of the case that genuinely contrary evidence fails to move the estimate at all.
Set an honest prior before discovery starts
Your very first probability-of-success estimate — ideally recorded before depositions and major discovery begin — is your prior. Write it down explicitly, along with its basis, so it can later be compared against your posterior estimates rather than being quietly rationalized after the fact to match however the case has gone.
Update after each material discovery event, not continuously
Treat updating as something that happens at discrete points, tied to specific events, rather than as a constantly drifting gut feeling.
- Treat each deposition, each key document production, and each expert report as a distinct update event with its own before-and-after estimate.
- Ask what that specific piece of evidence does to the probability of each contested element individually, rather than adjusting an overall case-wide feeling all at once.
- Log each update with a stated reason — mirroring the kind of provenance a well-run case file keeps, rather than an unexplained shift in the number.
Weight the update by how strong the evidence actually is
Not all evidence deserves the same size of update.
- A single ambiguous email should move a probability dial by less than a sworn admission on a dispositive fact.
- Corroborated testimony should carry more weight than a single uncorroborated account.
- Credibility problems in the source — impeachment material, apparent bias — reduce how much weight a given new fact should carry, even if its content looks favorable on its face.
Watch for anchoring and overcorrection
Two opposite errors are easy to fall into during a long discovery period, and both distort the final estimate.
- Anchoring: letting a strong early theory keep the probability estimate artificially high even after contrary evidence has genuinely accumulated.
- Overcorrection: letting one dramatic deposition swing the estimate further than its actual evidentiary weight supports.
- Periodically re-derive the estimate from the full evidentiary record built up so far, rather than only ever nudging the previous number.
Turn the running posterior into a decision signal
Track the sequence of posterior estimates across the life of the matter as a trend line, not just as a single final figure. A posterior that has stayed roughly stable across several successive discovery events is a meaningful signal that the estimate has genuinely converged and is ready to inform a settlement or trial decision — as opposed to one still swinging with each new document, which suggests it's too early to treat the number as settled.
- How is this different from just 'updating my gut feeling' after each deposition?
- Bayesian updating disciplines the process: an explicit starting prior, an explicit reason and weight assigned to each new piece of evidence, so the resulting estimate is inspectable rather than a vibe that shifted somewhere along the way.
- What counts as a 'prior' if I've already been through some discovery?
- Use your best honest estimate as of the earliest point you can still document, and treat everything that happened after that point as a sequence of updates rather than trying to reconstruct a true starting point retroactively.
- Can Bayesian updating make my probability estimate go down as well as up?
- Yes, and it should when the evidence warrants it. Evidence favorable to the other side is a legitimate update event and should move the dial down, not just be set aside.
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 →