Expectancy in Wicked Environments

There is a usable estimate inside almost any messy choice, and this page is how to get it out. The exact version needs posted odds — chance times size on each side, read the sign — and the choices that matter almost never post them. So the method is three cuts: delete everything small, weigh only the heavyweights that survive, rate those in bands — imperative down to insignificant for size, guaranteed down to low for chance. Crossing the two bands is the whole calculation, and it is small enough to hold in a head.

The game, then the setting that will not post odds

A game that publishes its rules lets the arithmetic run in the open. Chance of the good times its size, minus chance of the bad times its size: read the sign. A positive sign is a bet worth repeating. A negative sign is a bet to refuse.

Take a posted game. Win $10 at 75 percent, lose $50 at 25 percent, one hundred rounds. The wins add to $750. The losses add to $1,250. Expectancy is −$500. Do not play. Shrink the loss to $20 and the same odds flip: $750 of wins against $500 of losses, +$250. Same two levers, one number changed. The sign turns on chance and magnitude on each side.

That arithmetic assumes a posted-odds setting — one that publishes what winning and losing mean, so chances and sizes can actually be measured. Chess and golf do this. Some other scored games do too; “most sports” is looser once judging, weather, and opponents enter. The rest of life will not post those numbers. The win is undefined. The loss is undefined. Sometimes only a less-bad loss exists. More parts move than a head can multiply. No head computes expectancy over that list, and waiting for the perfect calculation means never deciding.

Three cuts

Three cuts shrink that kind of choice until the game logic applies again.

Cut the small fry. Low-chance, low-magnitude, or both — including outcomes that are guaranteed but too insignificant to matter — come off the list wholesale. Almost all of any long list is this residue. Small positives and small negatives roughly cancel; the rest can be handled as it arises. The move is delete the small. A count of how many is theatre.

Weigh only the heavyweights. After the first cut, judge on the few most significant positives against the few most significant negatives. The situation is now small enough to plug into the two-lever comparison.

Rate the survivors in broad bands. Fake precision — 78 percent against 79.3 — is the failure mode here. Size runs imperative, significant, convenient, insignificant; for a risk, imperative means unacceptable. Chance runs guaranteed, very high, somewhat, low. Crossing the two bands over the surviving variables is the whole calculation.

What keeps the cuts honest

Discovery comes before deletion. The danger is deleting a major downside that was never noticed. A thorough information pass is first, so a reader cannot “simplify” a harm they never listed. Speed compounds the benefit: the faster that pass runs, the cheaper good decisions become.

Then engineer the biggest downside. Drop its chance or its magnitude. Either lever alone can flip the sign — the $50 loss to $20 was one lever; a 25 percent chance cut to 5 percent is the other. Do the mirror work on the upside.

Ratings ride on the information behind them. A band after extensive discovery is an informed estimate. The same band assigned cold is a guess wearing the framework’s clothes. That is the quit signal: bands assigned before the pass.

The estimate will sometimes be wrong. The decision to estimate still pays. Running a complex choice through this protocol produces better decisions on average than skipping it — as a reason to run it, not as a measured trial of these labels.

Where a genuinely posted-odds pocket appears — real statistics, explicit win conditions — drop the bands and do the exact arithmetic. Positional Decisions and Expected Value owns the position-improving move and the downside protection across repeats; this page is why that calculation must be banded until a pocket like that shows up.

The survivors now fit the two-lever comparison, or a posted-odds pocket appeared and the bands dropped. Either way the estimate is one a person can actually run. A cold band is still a guess in this framework’s clothes. Stop faking the precision the setting never offered.

Links into the knowledge base

• Positional Decisions and Expected Value — position-improving move and downside protection across repeats; this page is why the calculation must be banded.
• Decision Making — hub operating loop (variable-naming, downside); the cuts slot into those steps.
• judging a decision by its process — process vs outcome after the estimate.

Open questions

When a band stays “I don’t know” after the information pass, is the next move delay or a no?

Sources

• Expected value as definitional: chance of the good times its size, minus chance of the bad times its size. Any finite-sample bet.
• Tetlock / superforecasting craft — coarse, calibrated probabilities beat fake fine ones for many tasks. Bound: not a trial of these four size labels or these four chance labels.
• Hogarth, Lejarraga & Soyer 2015, Current Directions in Psychological Science — a kind learning environment is one whose feedback is accurate, plentiful, and tightly linked to the action; a wicked one has missing, delayed, or misleading feedback. Bound only: that paper measures feedback quality. This page’s “wicked” means the odds are not posted and the win is not defined.