Stopping with search costAbout 6 min · Untimed

Discover

The toll booth

When taking an additional roll incurs a search fee or continuation cost c, the effective continuation value is reduced by c. The stopping hurdle for stage k becomes the net continuation value V_{k+1} - c. Paying a search cost lowers the reservation threshold, making it rational to accept lower current outcomes to avoid paying the continuation fee. On a fair six-sided die with a 0.50 reroll cost, net continuation is 3.50 - 0.50 = 3.00, raising the stopping set to include 3 and giving a net expected payout of 4.00.

Words you need

Threshold
The cut-off where you switch from continuing to stopping.e.g. Keep a die roll of 4 or more, reroll 1 to 3.
Continuation value
What carrying on is worth on average. Stop when what you have beats it.
Expected value (EV)
The average result if you could repeat the bet many times: each outcome times its probability, added up.e.g. A die pays its face in dollars: EV = 3.5.
Why it works

Rerolling requires paying cost c, so the hurdle threshold equals future expectation minus the cost.

Threshold = V_{next} - c, Net EV = P(stop) * E[stop] + P(reroll) * (V_{next} - c)

Spot it: Each optional reroll or search step incurs an explicit transaction fee.

Technique reference and sources

Every time you choose to reroll, you must drive through a toll booth costing c. If the destination past the toll is worth $3.50, driving through is only worth $3.50 - $0.50 = $3.00 net. You will accept any first roll of $3.00 or better to avoid paying the toll.

When this method fits

  • Each optional reroll or search step incurs an explicit transaction fee.
  • You need to price options with exercise or carry costs.
  • Evaluating search models where acquiring new offers costs time or money.
  1. Terminal gross value. Find the unpenalized expectation of the final roll: V_2 = (S + 1) / 2.
  2. Net continuation hurdle. Subtract cost c from terminal expectation: Hurdle = V_2 - c.
  3. Determine stopping set. Stop if X_1 > Hurdle. If X_1 = Hurdle, stopping and continuing have identical net EV.
  4. Compute net expected game value. Weight stopping outcomes and net continuation value (V_2 - c) by their branch probabilities.
Six-sided die with 0.50 reroll cost: Final roll averages 3.50. Net continuation is 3.50 - 0.50 = 3.00. Hurdle is 3.00. Net game EV is exactly 4.00.
  • Hurdle: 3.5 - 0.5 = 3.0
  • Net EV: (2/3) * 4.5 + (1/3) * 3.0 = 4.0
Cost indifference at boundary: When a roll exactly equals the net continuation value, stopping for X or paying c to roll again yields the exact same expected net payout.
  • On d6 with c = 0.5, roll 3 equals net continuation 3.0
  • Stopping gives 3; rerolling gives 3.5 - 0.5 = 3

Common traps

Ignoring the reroll cost when deciding whether to keep a first roll. Subtract the reroll cost from future expectation before deciding whether to stop. Failing to subtract cost leads to paying fees for negative-EV rerolls.

Confusing gross expected payout with net expected profit. Gross payout is the face value received; net profit subtracts all fees paid. When roll 1 is free, net EV is gross payout minus P(reroll) * c.

Deducting cost from the first roll when only the reroll incurs a fee. Only deduct costs for the rolls where fees are incurred. If the initial roll is free, no fee is subtracted from stage 1.

Another example: Fair eight-sided die with free first roll and 0.50 reroll cost
  1. Roll 2 has gross expected value V_2 = 4.50
  2. Net value of rerolling is 4.50 - 0.50 = 4.00
  3. Stop on {5, 6, 7, 8} (probability 4 / 8 = 0.5, average 6.50)
  4. Reroll on {1, 2, 3, 4} (probability 0.5, net continuation 4.00)
  5. Net expected value: 0.5 * 6.50 + 0.5 * 4.00 = 3.25 + 2.00 = 5.25

Source notes

These explanations and questions are written for QuantArc. Links show the underlying mathematics and learning design.

Jane Street: Probability & Markets Trading Interview Guide
Interview preparation guide (page 9) testing optimal stopping and market making on a d6 with one optional reroll. Lesson examples and questions are original.

Brainstellar: Probability interview puzzles
Free step-by-step walkthroughs of classic probability interview puzzles by difficulty; companion practice. Lesson examples and questions are original.

IES: Organizing Instruction and Study to Improve Student Learning
Supports worked examples, retrieval, explanations and spaced study. Does not establish QuantArc's exact lesson length, mastery threshold or review intervals.