Tensions in Practice: Expected payoff in tension with limiting hindsight regret¶
Two plans for a busy or quiet workshop day
Plan A earns 12 value units on a busy day and 0 on a quiet day; Plan B earns 8 and 6. The stipulated chances are 80% busy and 20% quiet. A has the larger expected payoff, 9.6 versus 7.6. But compare each plan with the best plan for the day that actually occurs: A can lag by 6, while B can lag by at most 4. That gap is regret, not a loss relative to zero.
Maximize the weighted mean
Use the declared probabilities to compare expected payoff.
Limit the largest forgone gain
Choose the plan with the smaller worst gap to a state-specific best alternative.
Why these aims pull against each other
The fixed probability forecast favors A, while the worst hindsight comparison favors B. Neither objective determines whether the other is the right concern.
Choose an arrangement to see what changes and what remains difficult.
Busy has probability 0.8 and Quiet 0.2. Each cell shows payoff; Gap is the shortfall from that day’s best available payoff (12 or 6). The last column changes the decision criterion, not the payoffs.
What this choice protects
What it costs
When it fits
Compare the arrangements
Favor the expected payoff
Compute 0.8×12 + 0.2×0 = 9.6 for A and 0.8×8 + 0.2×6 = 7.6 for B; select A.
| Busy | Quiet | Mean | |
|---|---|---|---|
| Plan A | 12Gap 0 | 0Gap 6 | 9.6 |
| Plan B | 8Gap 4 | 6Gap 0 | 7.6 |
- What it protects
- The expected payoff is two units larger under these exact probabilities.
- What it costs
- On a quiet day A earns 0 while the available B would have earned 6.
- When it fits
- Fits a decision that endorses expected payoff and the stipulated probability model.
Illustration note: The mean is an expectation, not a promised payoff on this one day.
Limit worst regret
The best available payoff is 12 when busy and 6 when quiet. Subtract each plan’s payoff from those two benchmarks and minimize the larger gap; select B.
| Busy | Quiet | Max gap | |
|---|---|---|---|
| Plan A | 12Gap 0 | 0Gap 6 | 6 |
| Plan B | 8Gap 4 | 6Gap 0 | 4 |
- What it protects
- No modeled day leaves the chosen plan more than four units behind the best alternative.
- What it costs
- Expected payoff falls from 9.6 to 7.6, and the regret bound depends on the declared alternative plans.
- When it fits
- Fits a decision that explicitly prioritizes limiting forgone opportunity over maximizing the declared mean.
Illustration note: This is worst-case regret, not expected regret. Minimizing expected regret under the same fixed state probabilities would select the expected-payoff maximizer.
What this illustration does—and does not—establish
The source supplies the structural tension. This bounded example makes a particular relation inspectable; the aims, conditions and residual costs are part of the comparison.
- Payoffs, probabilities and the two-option comparison class are invented and held fixed.
- The worst regret is not the worst payoff: the reference is the best alternative within the same state.
- No actual conduct, welfare judgment or recommendation follows from these value units.
Source entries
Regret
The canonical tension motivates this comparison. The setting and arrangements are declared editorial illustrations, not observed findings.
Minimizing regret and maximizing expected value pull apart
Because regret scores against the best hindsight benchmark rather than against an expectation, regret-minimizing acts can be systematically more conservative — or in adversarial settings more robust — than expected-value-maximizing acts.
The source operation
The defining structure is retrospective and relative: the same realized outcome can be coded as a triumph or a regret depending solely on which unchosen path it is compared against. A bet that returns ten percent is a success measured against the bank but a regret measured against the stock that doubled; nothing about the realized payoff has changed, only the benchmark.