Dividing the piano, the clock and the ring: a fair-division tool for heirs

A new free tool divides an estate's belongings between heirs: sealed bids with envy-free cash, or a proven max-Nash-welfare split with no cash at all.

Money in an estate splits itself. The piano, the grandfather clock and a wedding ring do not, and they are where families fall out. The usual answers are taking turns (whoever picks first does best) or appraisal value (which says nothing about who cares about what). The fair-division literature has better ones, so we built them into a free page: estate item divider.

Each heir privately writes down what every item is worth to them. Then:

Sealed bids, cash evens it up. Each item goes to whoever values it most, and cash moves so that no heir would rather have another heir's items plus that heir's cash. In the worked example, three heirs split six items and every one of them ends $1,255.56 above a fair third of the estate as they value it. The payments come from a short argument: envy-freeness is a set of difference constraints between heirs, longest paths give the smallest payments that satisfy them, and those payments move one-for-one with the margin over a fair share, so the best margin has a closed form. With unequal shares (a will saying 50/25/25) it switches to Knaster's 1946 sealed-bid procedure with percentages.

Items only, no cash. Many families will not pay each other. Then the tool maximises the product of what each heir gets (maximum Nash welfare), which is Pareto-optimal and envy-free up to one item. That is NP-hard in general, so the interesting part was proving answers optimal in a browser. A plain branch and bound with an "everyone gets everything left" bound proved 0 of 10 random estates with 4 heirs and 20 items. Adding a Lagrangian dual bound, with item prices read off the best division found so far, proved 10 of 10 at that size in about 2 ms, 10 of 10 with 6 heirs and 60 items, and 7 to 8 of 10 at the 8-heir, 80-item limit. When the proof does not finish, the page says "best found" rather than "proven". Local search alone already matched brute force on 400 of 400 small estates; the proof is what lets the page say so for yours.

The suite checks the claims rather than the code paths: 500 random estates where payments must sum to zero and nobody may envy anyone, a random search for envy-free payments that would treat the worst-off heir better (none found), brute-force comparisons for the Nash-welfare answers, and the guide's three tables recomputed from the engine.

It is arithmetic on the values people type, not legal advice; the will and the executor decide. Nothing is uploaded. Source and measurements: ventures/estate-item-divider.

All writing · More in Tools and projects · All tools and research