Statistics & decisions

Decision Making Under Uncertainty Lab

Compares 2 to 4 strategies against 3 uncertain scenarios with whole-number payoffs and whole-percent chances: expected value, the certainty equivalent of an exponential utility, maximin, maximax, Hurwicz, Laplace, minimax regret and the value of perfect information in closed form; a robust and a least-regret pick over an epsilon-contamination set of distrusted chances, exact at its corners and checked on a 5% grid; a scenario map of the best strategy over every chance with exact break-even lines; and 100 seeded simulated futures. Five scored challenges are included. For learning decision theory; not financial or business advice.

Last updated

Calculator overview

Inputs and outputs

This summary comes from the calculator's published input and output contract.

Inputs

Risk tolerance (required)
About this input

R, the risk tolerance of the exponential utility, a whole number from 10 to 1000 points: the smaller it is, the more a sure payoff is worth against a gamble. Used when risk averse.

Unit points Default 100 Range 10 to 1000
Risk attitude (required)
About this input

Risk neutral ranks the strategies by expected value; Risk averse by the certainty equivalent of an exponential utility with the risk tolerance below.

Default Risk neutral Allowed Risk neutral, Risk averse
Chance of scenario 1 (required)
About this input

P1, the chance of scenario 1, a whole percent from 0 to 100 (the drag point in the probability triangle).

Unit % Default 30 Range 0 to 100
Your prediction (challenges 1 and 5) (required)
About this input

Your prediction for challenges 1 and 5: Not yet predicted, Strategy A to D, or A tie. The results those challenges score stay hidden until you predict. Not used outside challenges 1 and 5.

Default Not yet predicted Allowed Not yet predicted, Strategy A, Strategy B, Strategy C, Strategy D, A tie
Chance of scenario 2 (required)
About this input

P2, the chance of scenario 2, a whole percent from 0 to 100; scenarios 1 and 2 together at most 100, and scenario 3 takes the rest.

Unit % Default 45 Range 0 to 100
Payoffs (row j is strategy A, B, C or D; three whole numbers from -1000 to 1000 points, or leave the row blank for a strategy not in use) (required)
About this input

The decision table: one row per strategy (A, B, C, D) with its payoff in scenarios 1, 2 and 3, whole numbers from -1000 to 1000 points. Leave a row blank for a strategy not in use (rows may have gaps; strategies keep their letters). At least two strategies are needed.

Default 3 rows
ColumnRange or allowed values
Scenario 1 -1000 to 1000
Scenario 2 -1000 to 1000
Scenario 3 -1000 to 1000
Challenge number (required)
About this input

Selects a challenge: 0 for none, 1 to 5 for a predefined challenge, which replaces the payoffs and settings except the input it leaves to you.

Default 0 Range 0 to 5
Distrust of the chances (required)
About this input

e, the distrust of the stated chances, a whole percent from 0 to 100: every mixture of (100 - e)% of your chances with e% of any other chances is considered possible.

Unit % Default 20 Range 0 to 100
Optimism (Hurwicz) (required)
About this input

a, the Hurwicz optimism: the weight on each strategy's best payoff (the worst gets 100 - a), a whole percent from 0 to 100.

Unit % Default 50 Range 0 to 100
Seed of the futures (required)
About this input

The seed of the generator that draws the 100 simulated futures, a whole number from 1 to 2147483646; the same seed gives the same futures.

Default 2026 Range 1 to 2147483646

Outputs

Model version
About this output

The model version.

No unit declared
Model Status
About this output

NOT VALID with the reason when an input cannot be used; CHECK when a challenge caps the scenario 2 chance at 70% or two strategies have identical payoffs; otherwise OK.

No unit declared
Formula module versions
About this output

The formula modules the model is built from, with their versions.

No unit declared
The payoffs used and the regret of each payoff against the best in its scenario
About this output

The payoffs used (a challenge sets its own) and the regret of each payoff against the best payoff of its scenario. A strategy not in use shows only its letter; the regrets are blank while challenge 1 awaits a prediction.

No unit declared
Optimism used
About this output

The Hurwicz optimism used, percent. Blank when the inputs are not valid.

Unit %
Best by Laplace (equal chances)
About this output

The strategy with the best plain average payoff (equal chances), or a tie. Blank when the inputs are not valid or while challenge 1 awaits a prediction.

No unit declared
Best by Hurwicz
About this output

The strategy with the best Hurwicz value (optimism x best + (100 - optimism) x worst), or a tie. Blank when the inputs are not valid or while challenge 1 awaits a prediction.

No unit declared
Best by maximax (best best payoff)
About this output

The strategy with the best best payoff, or a tie. Blank when the inputs are not valid or while challenge 1 awaits a prediction.

No unit declared
Best by minimax regret
About this output

The strategy with the smallest largest regret (shortfall against the best payoff of each scenario), or a tie. Blank when the inputs are not valid or while challenge 1 awaits a prediction.

No unit declared
Best by maximin (best worst payoff)
About this output

The strategy with the best worst payoff, or a tie. Blank when the inputs are not valid or while challenge 1 awaits a prediction.

No unit declared
Value of perfect information
About this output

The value of perfect information, points: the expected payoff if the scenario were known in advance, less the best expected value (equal to the smallest expected regret). Blank when the inputs are not valid or while challenge 1 awaits a prediction.

Unit points
Chance of scenario 3
About this output

P3 = 100 - P1 - P2, the chance of scenario 3, percent. Blank when the inputs are not valid.

Unit %
Chance of scenario 2 used
About this output

The chance of scenario 2 used, percent (challenge 2 caps it at 70). Blank when the inputs are not valid.

Unit %
The scenario map: the best strategy at each point of the probability triangle on a 5% grid (0 = a tie) and its margin over the runner-up
About this output

The scenario map: 231 rows, the points of the probability triangle on a 5% grid (scenario 1 and scenario 2 chances, percent), the best strategy there under the chosen attitude (1 to 4 for A to D, 0 for a tie) and its margin over the runner-up (points of expected value, or of certainty equivalent when risk averse). Best and Margin are blank while challenge 1 awaits a prediction.

No unit declared
Each strategy under every criterion
About this output

Each strategy (rows A to D): expected value, certainty equivalent (the expected value when risk neutral), worst and best payoff, Hurwicz value, average payoff, largest regret, worst case over the distrust set, worst regret over the set and the grid worst case (blank when no grid point is in the set). A strategy not in use shows only its letter.

No unit declared
Strategies in use
About this output

The strategies in use (rows with payoffs), 2 to 4. Blank when the inputs are not valid.

No unit declared
Risk attitude used
About this output

The risk attitude used. Blank when the inputs are not valid.

No unit declared
Least worst regret over the set
About this output

The strategy with the smallest worst regret over the distrust set (payoff units), or a tie. Blank when the inputs are not valid or while challenge 1 or 5 awaits a prediction.

No unit declared
Risk tolerance used
About this output

The risk tolerance used, points. Blank when the inputs are not valid.

Unit points
Chance of scenario 1 used
About this output

The chance of scenario 1 used, percent (a challenge sets its own). Blank when the inputs are not valid.

Unit %
Robust pick (best worst case over the set)
About this output

The strategy with the best worst expected value (or worst certainty equivalent when risk averse) over the distrust set, or a tie. Blank when the inputs are not valid or while challenge 1 awaits a prediction.

No unit declared
Map points inside the distrust set
About this output

The map points (5% grid) that lie inside the distrust set. Blank when the inputs are not valid or while challenge 1 awaits a prediction.

No unit declared
Every criterion: the best strategy (0 = a tie), its margin over the runner-up and how many strategies tie for best
About this output

Every criterion (9 rows): the best strategy (1 to 4, 0 for a tie), its margin over the runner-up (points) and how many strategies tie for best.

No unit declared
Challenge result
About this output

No challenge selected, MET or NOT MET.

No unit declared
Tie band (risk averse)
About this output

The risk-averse tie band, 1E-9 x (1 + the largest absolute payoff), points; 0 when risk neutral (every comparison is then exact). Blank when the inputs are not valid.

Unit points
Distrust used
About this output

The distrust used, percent. Blank when the inputs are not valid.

Unit %
The corners of the distrust set (all the distrusted weight on one scenario)
About this output

The three corners of the distrust set (all the distrusted weight on scenario 1, 2 or 3), as scenario 1 and scenario 2 chances in percent.

No unit declared
Challenge feedback
About this output

Feedback on the selected challenge: the prediction prompt, the met text or the not-met hint (blank with none).

No unit declared
Break-even lines between each pair of strategies, where they cross the triangle
About this output

For each pair of strategies in use (A-B, A-C, A-D, B-C, B-D, C-D), where the line on which the two are equally good (in expected value, or expected utility when risk averse) enters and leaves the triangle, as scenario 1 and scenario 2 chances in percent; the four end columns are blank when the line does not cross the triangle. Every column is blank while challenge 1 awaits a prediction.

No unit declared
Challenge goal
About this output

The goal of the selected challenge (blank with none).

No unit declared
Challenge score
About this output

The challenge score: 100 when met, else 0.

No unit declared
Challenge: the value scored
About this output

The value the selected challenge scores: the prediction code of the expected-value pick (1; 1 to 4 a strategy, 5 a tie), the scenario 3 chance in percent (2), the risk tolerance (3), the distrust in percent (4) or the prediction code of the least-regret pick (5); blank when not valid or while awaiting a prediction.

No unit declared
Dominated strategies
About this output

The strategies dominated by another strategy in use (never better, and worse in at least one scenario), as letters, or None. Blank when the inputs are not valid.

No unit declared
Leader after 100 futures
About this output

The strategy with the highest total payoff after the 100 simulated futures, or a tie. Blank when the inputs are not valid or while challenge 1 awaits a prediction.

No unit declared
Changes of leader over the futures
About this output

How many times the leader changes from one future to the next. Blank when the inputs are not valid or while challenge 1 awaits a prediction.

No unit declared
100 simulated futures: the scenario drawn and each strategy's running average payoff, with the leader (0 = a tie)
About this output

The 100 simulated futures: the future number, the scenario drawn, each strategy's running average payoff (blank for a strategy not in use) and the leader (1 to 4, 0 for a tie). The averages and the leader are blank while challenge 1 awaits a prediction.

No unit declared
Grid never below the exact worst case
About this output

Yes when, for every strategy, the worst case over the grid points in the set is never below the exact worst case over the set; No otherwise. Blank when no grid point is in the set. Blank when the inputs are not valid or while challenge 1 awaits a prediction.

No unit declared
Seed used
About this output

The seed used (always yours). Blank when the inputs are not valid.

No unit declared
Best by expected value
About this output

The strategy with the highest expected value (Strategy A to D), or Tie: and the tied letters. Blank when the inputs are not valid or while challenge 1 awaits a prediction.

No unit declared
Best by expected utility
About this output

The strategy with the highest certainty equivalent when risk averse (the expected value when risk neutral), or a tie; under risk aversion scores within the tie band count as tied. Blank when the inputs are not valid or while challenge 1 awaits a prediction.

No unit declared
Futures in scenario 1
About this output

The simulated futures (of 100) that fell in scenario 1. Blank when the inputs are not valid.

No unit declared
Futures in scenario 3
About this output

The simulated futures that fell in scenario 3. Blank when the inputs are not valid.

No unit declared
Futures in scenario 2
About this output

The simulated futures that fell in scenario 2. Blank when the inputs are not valid.

No unit declared
This page is provided by LogicCommons for informational purposes only. Results are analysis outputs computed from the inputs you supply and are not professional advice. Verify all inputs and results independently.

LogicCommons is in beta. If a result, label, or reference looks wrong, tell us here; we read every message.

Methodology

Purpose and model boundary

This lab sets one decision among 2 to 4 strategies (A to D) against exactly three uncertain scenarios. Each strategy has a whole-number payoff in each scenario, and the three scenarios have whole-percent chances that add to 100. For the table and chances you set, the model ranks the strategies by nine criteria: expected value, expected utility (the certainty equivalent of an exponential utility), maximin, maximax, Hurwicz, Laplace, minimax regret, a robust choice that distrusts the stated chances, and the least worst regret over the same distrusted set. It also returns the value of perfect information, the dominated strategies, a map of the best strategy over every possible set of chances with the exact break-even lines between strategies, and 100 simulated futures drawn with the stated chances. Five scored challenges use the same model.

It is a lab for learning how decision criteria work and why they disagree: a criterion that uses the chances, one that ignores them, and one that trusts them only partly can each pick a different strategy for the same table. Every pick is one criterion's answer for the payoffs and chances entered, not a recommendation. The lab is not financial, investment or business advice. It does not model sequential decisions, learning from new information, more than three scenarios, or risk attitudes other than risk neutral and the exponential utility.

The model is the calculation authority. The page sends the values you set to the model and draws only what it returns: the coloured probability triangle, the distrust region, the break-even lines, the payoff bars with their expected-value and certainty-equivalent marks, and the futures player. The page computes no expected value, certainty equivalent, pick, regret or simulated future itself.

Inputs and units

Payoffs are in points, a unitless score. Chances, optimism and distrust are whole percents. Every input is required and is checked on its entered value, even when the chosen attitude or challenge does not use it.

Input Meaning and unit
Payoffs One row per strategy A, B, C and D, with its payoff in scenarios 1, 2 and 3: whole numbers from -1,000 to 1,000 points. Leave a row blank for a strategy not in use; at least two strategies are needed. Rows may have gaps, and each strategy keeps the letter of its row. Default: A (25, 25, 25), B (-15, 45, 60), C (-70, 15, 140), D blank.
Chance of scenario 1 P1, a whole percent from 0 to 100 (default 30%).
Chance of scenario 2 P2, a whole percent from 0 to 100 (default 45%). Scenarios 1 and 2 together are at most 100%; scenario 3 takes the rest, P3 = 100 - P1 - P2.
Risk attitude Risk neutral (default) or Risk averse. Risk neutral ranks by expected value; Risk averse by the certainty equivalent of an exponential utility.
Risk tolerance R, a whole number from 10 to 1,000 points (default 100). The smaller it is, the more a sure payoff is worth against a gamble. Used when risk averse.
Optimism (Hurwicz) a, a whole percent from 0 to 100 (default 50%): the weight on each strategy's best payoff; its worst payoff gets 100 - a.
Distrust of the chances e, a whole percent from 0 to 100 (default 20%): every mixture of (100 - e)% of your chances with e% of any other chances is considered possible.
Seed of the futures The starting number of the generator that draws the 100 futures, a whole number from 1 to 2,147,483,646 (default 2026). The same seed always gives the same futures.
Challenge number 0 for free exploration (default), or 1 to 5 to load a challenge.
Your prediction Not yet predicted (default), Strategy A to D, or A tie: your answer in challenges 1 and 5. It changes nothing outside those two challenges.

Where each input is set:

  • On the canvas only: the chances of the three scenarios, set by dragging the point in the probability triangle (whole percents; arrow keys move it by 1%, and Enter opens exact values), and your prediction in challenges 1 and 5. They are not in the input panel.
  • In the input panel: the payoffs (one row per strategy, each entry with a slider popup), the risk attitude, the risk tolerance, the optimism, the distrust, the seed and the challenge number.

Governing relationships

Below, x_jk is the payoff of strategy j in scenario k, P1, P2 and P3 are the chances in percent, and only strategies in use take part in any maximum, minimum or comparison. mn_j and mx_j are the worst and best payoffs of strategy j, and cm_k is the best payoff of any strategy in scenario k.

Expected value. EV_j = (P1·x_j1 + P2·x_j2 + P3·x_j3) / 100. The model compares the whole-number sums before the division, so every risk-neutral comparison is exact.

Expected utility and the certainty equivalent. The risk-averse decision maker has the exponential utility u(x) = -exp(-x/R), which has constant absolute risk aversion. The certainty equivalent is the sure payoff with the same expected utility. The model computes it in a shifted form that keeps every exponent between -200 and 0:

E_jk = exp(-(x_jk - mn_j)/R), S_j = P1·E_j1 + P2·E_j2 + P3·E_j3, CE_j = mn_j - R·ln(S_j / 100).

A strategy with the same payoff in every scenario has a certainty equivalent exactly equal to that payoff. The certainty equivalent is never above the expected value. When risk neutral, the certainty equivalent shown is the expected value.

Criteria that ignore the chances.

  • Maximin: the strategy with the best worst payoff mn_j. Maximax: the best best payoff mx_j.
  • Hurwicz: the value (a·mx_j + (100 - a)·mn_j) / 100. At a = 0 it is maximin, at a = 100 maximax.
  • Laplace: the plain average (x_j1 + x_j2 + x_j3) / 3, that is, equal chances.
  • Minimax regret: the regret of a payoff is r_jk = cm_k - x_jk, its shortfall against the best payoff of its scenario. The pick has the smallest largest regret MR_j = max_k r_jk. Regret is measured in payoff points under both attitudes.

The distrust set (epsilon-contamination). With ε = e/100 and the stated chances p, the set of chances considered possible is every mixture (1 - ε)·p + ε·q, where q is any set of chances. Its three corners put all the distrusted weight on one scenario: corner k has chance (100 - e)·P_m/100 for each scenario m, plus e on scenario k itself.

  • Worst expected value over the set (risk neutral). Expected value is linear in the chances, so its minimum is at a corner, and it equals (1 - ε)·EV_j + ε·mn_j.
  • Worst certainty equivalent over the set (risk averse). Expected utility is linear in the chances too, so the worst case is RC_j = mn_j - R·ln(((100 - e)·S_j + 100·e) / 10000).
  • The robust pick has the best worst case. At e = 0 it is the expected-value (or expected-utility) pick; at e = 100 it is the maximin pick.
  • Worst regret over the set. The regret of strategy j at chances q is the best expected value at q minus the expected value of j at q. This is convex in q, so its maximum over the set is also at a corner. The least-regret pick has the smallest worst regret. At e = 0 the worst regret is the shortfall in expected value; at e = 100 it is the largest regret MR_j.

Value of perfect information. The expected payoff if the scenario were known before choosing, less the best expected value:

VPI = (P1·cm_1 + P2·cm_2 + P3·cm_3) / 100 - max_j EV_j.

It is never negative, and it equals the smallest expected regret of any strategy.

Dominance. Strategy j is dominated when another strategy in use pays at least as much in every scenario and more in at least one. Two strategies with identical payoffs do not dominate each other.

Picks and the one tie rule. For each criterion the model takes the best and second-best scores. The pick is tied when the best leads by no more than the tie band; it then reads "Tie:" followed by the tied letters, otherwise "Strategy" and the letter. The band is 0 for every risk-neutral comparison, so ties there are exact. For the two certainty-equivalent scores under risk aversion the band is 1×10⁻⁹ × (1 + the largest absolute payoff), so a difference that small reads as a tie. The same rule and band serve every criterion, the map, the futures leader and the challenges.

The scenario map. The model scores every strategy at each of 231 points of the probability triangle on a 5% grid (scenario 1 and scenario 2 chances 0, 5, … , 100 with their sum at most 100), by expected value or, when risk averse, by the certainty equivalent at those chances. Each point gets its best strategy (or a tie) and the margin over the runner-up. A point lies in the distrust set when 100·Q_k ≥ (100 - e)·P_k for every scenario k, where Q is the point's chances. Because the grid points in the set are part of the set, the worst case of a strategy over them is never below its exact worst case at the corners; the model reports that check as "Grid never below the exact worst case".

Break-even lines. For each pair of strategies in use, the line on which the two are equally good (equal expected value, or equal expected utility when risk averse) crosses the triangle only when one pays more in some scenario and the other in another. With d_k the difference between the two in scenario k (the payoff difference when risk neutral, the difference of exp(-(x - X0)/R) terms when risk averse, with X0 the lower of the two worst payoffs), the line meets each edge where the sign of d changes at the point found by linear interpolation: on the edge P3 = 0 at P1 = 100·d_2 / (d_2 - d_1), on the edge P1 = 0 at P2 = 100·d_3 / (d_3 - d_2), and on the edge P2 = 0 at P1 = 100 - 100·d_1 / (d_1 - d_3). A corner where the two pay the same is an end of the line. The lines are drawn for every pair, not clipped to where that pair is the top two.

The simulated futures. The model implements Park and Miller's minimal standard generator (MINSTD) with the multiplier 48271. Starting from s₀ = the seed, each state is

sₜ = MOD(48271 · sₜ₋₁, 2³¹ - 1), for t = 1 to 100,

so the first state used is 48271 · seed reduced modulo 2,147,483,647, not the seed itself. Every product stays below 2⁴⁷, so the integer sequence is exact. Future t uses one draw, uₜ = sₜ / (2³¹ - 1): it falls in scenario 1 when uₜ < P1/100, in scenario 2 when P1/100 ≤ uₜ < (P1 + P2)/100, and in scenario 3 otherwise (the discrete inverse transform, tested in whole numbers, so no draw can sit exactly on a boundary). Each strategy's running total adds its payoff in the scenario drawn; its running average is that total over t. The leader of future t is the strategy with the highest running total, by the tie rule with a band of 0.

Calculation sequence

The model works through the relationships above in this order:

  1. It checks the entered values in the order of the status list under Errors and warnings.
  2. It fixes the values used: your inputs, or in a challenge the challenge's own values except the inputs that challenge leaves free. In challenge 2 a scenario 2 chance above 70% is used as 70%.
  3. It finds the strategies in use, each strategy's worst and best payoff, and the best payoff of each scenario.
  4. It computes, for each strategy, the expected value, the certainty equivalent, the Hurwicz value, the average, the regrets and largest regret, the three corners of the distrust set, the worst case and worst regret over the set, the value of perfect information and the dominated strategies.
  5. It applies the tie rule to the nine criteria to give each pick, its margin and the number tied.
  6. It scores the 231 map points, marks those inside the distrust set, runs the grid check, and finds the break-even lines.
  7. It draws the 100 futures from the seed and keeps the running averages and the leader.
  8. It sets the status and, in a challenge, scores the challenge.

Outputs and interpretation

  • Picks: best by expected value, expected utility, maximin, maximax, Hurwicz, Laplace and minimax regret; the robust pick (best worst case over the distrust set); the least worst regret over the set. Each reads Strategy A to D, or Tie: and the tied letters.
  • Strategies in use (2 to 4), the chance of scenario 3 in percent, and the dominated strategies as letters, or None.
  • Each strategy under every criterion: expected value, certainty equivalent (the expected value when risk neutral), worst and best payoff, Hurwicz value, average payoff, largest regret, worst case over the set, worst regret over the set and the grid worst case, all in points. A strategy not in use shows only its letter.
  • Every criterion: the best strategy (0 for a tie), its margin over the runner-up in points, and how many strategies tie for best. A wide margin means the pick is safe against small changes.
  • The payoffs used and the regrets: the payoffs (a challenge sets its own) and each payoff's regret.
  • Value of perfect information, points: the most it would be worth to learn the scenario before choosing.
  • Tie band (risk averse), points: 0 when risk neutral.
  • Distrust set: its three corners as scenario 1 and scenario 2 chances, the map points inside it, and "Grid never below the exact worst case", Yes or No (blank when no map point lies in the set).
  • Scenario map: for each of the 231 points, the best strategy (0 for a tie) and its margin.
  • Break-even lines: for each pair of strategies, where the line of equal value enters and leaves the triangle; the ends are blank when the line does not cross it.
  • 100 simulated futures: the scenario drawn for each future, each strategy's running average and the leader; the futures that fell in each scenario; the leader after 100 futures and how many times the leader changed.
  • Settings used: the chances, attitude, tolerance, optimism, distrust and seed of the calculation (a challenge's own where it sets them), and the model and formula module versions.

Reading the canvas: each corner of the triangle is a scenario, and each point inside is one set of chances, coloured by the strategy that is best there; a stronger colour is a wider lead. Your point sits where the chances are; the outlined region around it is the distrust set, and the lines are the break-even lines. The bars show each strategy's payoff in scenarios 1, 2 and 3 with its expected value and certainty equivalent marked; a criterion chip highlights that criterion's pick. The futures player shows the scenario of each future and the running averages settling toward the expected values.

What to take from the lab: an expected value is a long-run average over many futures, not a promise for one; criteria that ignore the chances answer a different question from those that use them; risk aversion and distrust both pull the choice toward strategies with better worst cases; and the value of perfect information bounds what any forecast could be worth.

Challenges

Every challenge uses the same plant table: A (20, 20, 20), B (-10, 40, 50) and C (-60, 10, 120), with no strategy D, at chances 30%, 40% and 30%, risk neutral, a risk tolerance of 50, an optimism of 50% and a distrust of 0%, except where noted. Each challenge leaves you only the input it names, plus the seed of the futures, which is never scored. The model scores each challenge MET (score 100) or NOT MET (score 0), with the goal, the value scored and a one-line feedback or hint.

  1. Predict the expected-value pick. Predict which strategy has the highest expected value when the chances are 30%, 40% and 30%. Open: your prediction. MET when your prediction names the expected-value pick (A tie for a tied pick).
  2. Tip the balance. With scenario 1 fixed at 30%, find the smallest scenario 3 chance at which C has the highest expected value. Open: the chance of scenario 2 (the drag point moves along the line where scenario 1 is 30%). MET when the scenario 3 chance is at least 1%, C is the expected-value pick alone, and with one percent less in scenario 3 it is not.
  3. How cautious must you be? Risk averse. Find the largest risk tolerance, in steps of 5, at which the safe strategy A is the expected-utility choice. Open: the risk tolerance. MET when A is the expected-utility pick and, at a tolerance 5 points higher, it is not.
  4. Distrust your forecast. Find the smallest distrust, in steps of 5%, at which the robust choice becomes the safe strategy A. Open: the distrust. MET when the distrust is at least 5%, A is the robust pick, and at 5% less distrust it is not.
  5. Least regret over the set. With 30% distrust, predict which strategy has the smallest worst regret over the whole set of chances. Open: your prediction. MET when your prediction names the least-regret pick.

Challenges 1 and 5 hold back the answer until you predict. While your prediction reads Not yet predicted, the challenge reads NOT MET with the prompt to predict first and no value scored. In challenge 1 every pick, the value of perfect information, the grid check, the futures leader and its changes, the criterion results, every summary column except the worst and best payoff, the regrets, the map colours and margins, the break-even lines and the futures averages are held back. In challenge 5 only the least worst regret pick, the worst regret over the set and that criterion's row are held back. Once you predict, every result appears and the challenge is scored.

Validation and status logic

The status reads the first match in the list under Errors and warnings. The NOT VALID checks apply to the entered values, also inside a challenge. The two CHECK items apply to the values used, and they also apply inside a challenge. A CHECK does not stop the calculation: every result is still shown. While challenge 1 or 5 awaits a prediction, the status is computed in the usual way. Ties, dominated strategies and criteria that disagree are results, reported by the picks, never warnings.

Assumptions and limitations

  • Two to four strategies and exactly three scenarios; whole-number payoffs from -1,000 to 1,000 points and whole-percent chances.
  • Risk neutral, or risk averse with the exponential utility (constant absolute risk aversion) only; no wealth-based risk aversion.
  • Regret is measured in payoff points under both attitudes.
  • The distrust set is the epsilon-contamination set only. Its worst case and worst regret are exact, taken at its three corners; the 5% grid only checks them.
  • The map has 5% resolution; the exact borders are the break-even lines, which are drawn for every pair and not clipped to where that pair decides.
  • Every risk-neutral comparison is exact. Under risk aversion, scores within the tie band read as a tie, so a real difference that small is not shown.
  • The 100 futures come from one small linear congruential generator and illustrate sampling only. They settle nothing and are not a forecast. The same seed always gives the same futures.
  • One decision made once: no sequential decisions and no updating of the chances from new information.

Restrictions and non-computing states

Under NOT VALID every result and every table is blank, except the model and module versions, the challenge result (NOT MET when a challenge is selected, otherwise No challenge selected), a challenge score of 0, and the goal and hint of a selected challenge. A challenge number that is not a whole number from 0 to 5 counts as no challenge selected.

In challenges 1 and 5, until you predict, the results named under Challenges are held back.

Errors and warnings

The status messages, first match wins:

  • NOT VALID: strategy A needs three whole-number payoffs from -1000 to 1000, or leave its row blank: a row in use has a payoff that is blank, not whole, or out of range. The letter names the first such row.
  • NOT VALID: enter payoffs for at least two strategies: fewer than two rows are in use.
  • NOT VALID: the scenario chances must be whole percents from 0 to 100, and scenarios 1 and 2 together at most 100: a chance is blank, not whole or out of range, or the first two add to more than 100.
  • NOT VALID: choose the risk attitude from the list
  • NOT VALID: the risk tolerance must be a whole number from 10 to 1000
  • NOT VALID: the optimism and the distrust must be whole percents from 0 to 100
  • NOT VALID: the seed must be a whole number from 1 to 2147483646
  • NOT VALID: the challenge number must be a whole number from 0 to 5, and the prediction one of the listed choices
  • CHECK: this challenge fixes scenario 1 at 30%, so scenario 2 can be at most 70%; it is used as 70%: challenge 2 with an entered scenario 2 chance above 70%.
  • CHECK: strategies A and B have identical payoffs, so every criterion ties them; change one of them or clear its row: two strategies in use have the same three payoffs; the letters name the first such pair.
  • OK: the inputs are valid with nothing to flag.

References

This laboratory is an educational model of decision making under uncertainty. It ranks a few strategies against three scenarios by expected value, by the certainty equivalent of an exponential utility (constant absolute risk aversion), and by the classic criteria that ignore the chances: maximin, maximax, Hurwicz's weighting of the best and worst payoffs, Laplace's equal chances and minimax regret. It gives the value of perfect information, and a robust choice over an epsilon-contamination set of chances, every mixture of the stated chances with any others, whose worst case lies at the set's corners. Its simulated futures come from the Park-Miller minimal standard (MINSTD) generator with the multiplier 48271.

Sources and further reading

  • The decision criteria under uncertainty (maximin, maximax, Hurwicz, Laplace, minimax regret), expected monetary value and the expected value of perfect information: INFLIBNET Centre, Managerial Economics, chapter 3, Elements of Decision Theory and Decision Analysis, sections Decision Making Criteria, Decision Making under Uncertainty and Decision Making under Risk. INFLIBNET
  • Solved decision-table problems with these criteria, regret tables, expected values, the expected value of perfect information and the break-even chances between alternatives: Stevenson, Operations Management, 7th edition, McGraw-Hill, Online Learning Center, chapter 5 supplement, S-Problems/Solutions. McGraw-Hill
  • Risk aversion, constant absolute risk aversion and the certainty equivalent: J. W. Pratt, "Risk aversion in the small and in the large", Econometrica 32(1/2), 1964. doi:10.2307/1913738
  • The exponential utility function and its constant absolute risk aversion: Wikipedia, Exponential utility, section Risk aversion characteristic. Wikipedia
  • The epsilon-contamination class of distributions: J. Berger and L. M. Berliner, "Robust Bayes and empirical Bayes analysis with epsilon-contaminated priors", The Annals of Statistics 14(2), 1986, pages 461-486. doi:10.1214/aos/1176349933
  • The minimal standard generator: S. K. Park and K. W. Miller, "Random number generators: good ones are hard to find", Communications of the ACM 31(10), 1988, pages 1192-1201. doi:10.1145/63039.63042

No table or text from any of these publications is reproduced, and none of their authors or publishers has reviewed or endorses this laboratory.

This laboratory is for learning. Its results hold only within its stated ranges; every pick is one criterion's answer for the payoffs and chances entered, not financial, investment or business advice.

Links were current at build time. If one moves, search the document's title rather than reporting a broken link as a defect.

Results are informational and not professional advice. See our Terms of Use. Powered by SpreadsheetWeb · About · Privacy · Terms