01
EXPERIMENT INPUT
Choose the question
%
0 / IMPOSSIBLE100 / CERTAIN
Applied independently to every game in the experiment.
3×
Three layers of doubt.
Every number gets a best guess, how wrong that guess could be, and how wrong that could be. Inside each drawn world the game is solved exactly, so the error bars come only from what you said you did not know.
1,089×
No probability required.
The engine tests every 0.01% from 0.01% to 1%, then every 0.1% from 1.1% to 99.9%.
TOTAL GAMES60,000
NO PROJECT LIMIT · CAPACITY DEPENDS ON THIS MACHINE
SAMPLING EFFORT
WORLDS EXPLORED—
THE ENGINE MEASURES ITS OWN ERROR AND KEEPS WORKING UNTIL THE ANSWER IS STEADY
01B
What you think,
STATE YOUR BELIEFS
What you think,
and how sure you are.
Each parameter takes three numbers. The best guess is the single value you would name if forced. The uncertainty is how far off that guess could plausibly be, as a percentage of it. The doubt about the doubt asks how much you trust that error bar — leave it at zero when you are confident in it, raise it when the error bar is itself a guess.
HOW LEARNING WORKS
PARAMETERBEST GUESSUNCERTAINTYDOUBT ABOUT THE DOUBT
02
OUTCOME FIELD
Run results
SUCCESS
FAILURE
—
MINI AGENTS →
PROBABILITY GAMES →
FIVE ANALYTICAL PORTRAITS
The whole field,
seen five ways.
Each view compresses the complete experiment into a different signal: convergence over time, game-to-game variation, the distribution between agents, and final performance ranking. Hover to inspect; click to pin a value.
Cumulative convergence
Does the observed success rate settle toward the probability you chose?
Game pulse
How many agents succeeded in each probability game?
Agent distribution
How many agents ended with zero, one, two, or more successes?
Agent ranking
Every agent ordered from the highest success count to the lowest.
Time to first success
How many games did each agent play before achieving its first success?
VALUE
SCALE
LAYERS
ACTIONS
DRAG ACROSS THE PLOT TO SELECT A RANGEVIEWING ALL GAMES
SELECTED RANGEALL GAMES
AGENTS—
SHARE—
EXPECTED—
OBS − EXP—
AVG WAIT—
MEDIAN WAIT—
EXPERIMENT ECONOMICS
The price
of probability.
Cost is charged for every game played. Reward is earned for every success. These views show whether the observed field created or destroyed value—and where the economics break even.
TOTAL COST—
GROSS REWARD—
NET P&L—
ROI—
BREAK-EVEN RATE—
Cumulative P&L
Observed and statistically expected net value as each game round completes.
Agent profit distribution
How profit and loss are distributed across the complete agent population.
Profitability sensitivity
Expected experiment value at every possible success probability.
02
UNKNOWN-ODDS LAB
The probability atlas
One identical experiment across 1,089 odds: 0.01% steps through 1%, then 0.1% steps through 99.9%. The decision lens automatically finds the first profitable odds and its 90% move-on boundary.
PROBABILITIES TESTED1,089
TOTAL TRIALS—
BREAK-EVEN ODDS—
COMPUTE TIME—
DECISION LENS
When should I move on?
The engine will find the first tested probability with positive expected value, then calculate the attempts needed for at least a 90% chance of one win.
AUTOMATIC STOP POLICY
FIRST POSITIVE-EV ODDS
90% CHANCE OF ≥1 WIN
MOVE-ON BOUNDARY——
FIRST WINNING ODDS—
MAXIMUM SPEND—
VALUE / NEXT TRY—
Profit-target deadline
Choose the return you require. The planner calculates the last attempt where one win would still preserve that target.
0—
STOP AFTER——
CUMULATIVE COST—
PROFIT IF WIN LANDS THERE—
CHANCE AT INSPECTED ODDS—
Wins against losses
The full outcome composition at every tested probability.
Profit frontier
Observed and expected value reveal the odds where this game becomes financially rational.
Agent coverage
What share of agents achieved at least one win before the game limit?
Move-on boundary
Attempts needed to reach the automatic 90% chance of seeing at least one win.
Cost to 90% coverage
The maximum spend required to reach that move-on boundary.
First-win waiting bands
Median and 90th-percentile waits, including agents that never won within the limit.
Zero-win risk
The observed and theoretical chance of reaching the limit without a single success.
Outcome volatility
How widely the win count varies between agents at each probability.
Attempt decision curve
For the inspected probability: cumulative chance of a first win, remaining miss risk, and the fixed 90% boundary.
02
UNCERTAIN-WORLD LAB
The shape of not knowing
Each world draws a different set of error bars; each universe inside it draws different parameters. Within a universe the game is solved in closed form rather than played out, so everything below is exact given the beliefs you stated — the only sampling error lies in exploring your uncertainty.
UNIVERSES EXPLORED—
ATTEMPT HORIZON—
LEARNING LAW—
COMPUTE TIME—
PRECISION REACHED—
THE ANSWER
How many times
should I try?
—
STOP AFTER
—
—
EXPECTED PROFIT—
EXPECTED OUTLAY—
RETURN ON OUTLAY—
CHANCE YOU END UP AHEAD—
—
TYPICAL CASE (P50)—
BAD CASE (P05)—
WORST 5% AVERAGE—
Profit against where you stop
The expected result of quitting after each number of attempts, with the middle half and the middle 90% of outcomes behind it.
Is one more try worth it?
The value of playing each attempt rather than stopping just before it. Once this is below zero for good, you are done.
Getting better as you go
Your chance of winning on each attempt, fanned out across everything you might be. The dashed line is the chance you would need just to break even.
Where you might land
The full spread of outcomes if you follow the recommendation. Everything left of the line is a loss.
Chance of ending ahead
The probability of finishing in profit for each place you could stop. Its peak sits elsewhere than average profit, and that gap is the risk you are choosing.
Why the outcome is uncertain
Your risk split three ways: the luck of the game itself, not knowing the parameters, and not knowing how badly you know them. Only the first is unavoidable — the other two shrink if you go and find something out.
Which unknown moves the needle
How strongly each parameter drives the result, measured on ranks so a strong but curved relationship still registers.
What an answer would be worth
How much better you could do by learning one parameter exactly before deciding. This is the most it is worth paying to find out.
Chance of a win against money spent
How the odds of having won at least once build up, and what you will have spent by the time they do.