Documentation
The Algorithm
How we turned “should I do this?” into math. Every score on this site is computed from peer-reviewed research, actuarial data, and neuroscience — then compressed into two numbers you can actually argue about at dinner.
136
Activities
1,088
Data Points
4
Risk Factors
4
Fun Factors
Overview
Every activity gets two scores: Fun (0–100) and Risk (0–100). These combine into two output metrics:
Worth It (0–100)
Our primary ranking metric. 50 = break-even, 100 = amazing deal, 0 = terrible idea.
FRR (Fun÷Risk Ratio)
A secondary badge. Above 2.0× = great deal, below 0.5× = reconsider your life choices.
The system uses fixed anchor points — not dataset-relative normalization — so scores never shift when we add new activities. Walking stays walking whether or not Everest is in the database.
Risk Score
Risk captures four dimensions: acute mortality, chronic health burden, life disruption, and legal exposure. The first two are fused into a single Health Score using a calibrated log scale.
Step 1: Unified Health Burden (H_eq)
Acute risk (micromorts) and chronic risk (microlives) are measured in fundamentally different units. We convert both to a common scale using an empirically-derived equivalence:
Where:
Micromort (μmt)
A one-in-a-million chance of death per exposure. Unit introduced by Ronald Howard (Stanford, 1980).
Microlife
30 minutes of life expectancy lost through chronic behavior. Unit introduced by David Spiegelhalter (Cambridge, 2012).
1.25 conversion factor: derived from the observation that 1 microlife (30 min life expectancy) ≈ 1.25 micromort-equivalents when assuming ~45 years remaining life expectancy.
Step 2: Anchored Log Normalization
The raw Heq ranges from 0.001 (walking) to 37,932 (Everest) — a 38-million-fold range. Linear scaling would make everything except Everest look like zero. We use a fixed-anchor logarithmic transform:
This produces intuitive, stable anchor points:
| H_eq (micromort-equiv) | Health Score | Example Activity |
|---|---|---|
| 0.001 | ~1 | Walking around the city |
| 0.01 | ~6 | Taking a shower |
| 0.1 | ~21 | Sauna session |
| 1 | ~40 | Scuba diving |
| 10 | ~60 | Skydiving, bungee jumping |
| 100 | ~80 | BASE jumping |
| 1,000+ | ~100 | Wingsuit flying, Everest |
Rationale: Log Scale
Humans perceive risk logarithmically (Weber-Fechner law). The difference between 1 and 10 micromorts “feels” similar to the difference between 100 and 1,000. Our log transform matches this psychological reality.
Step 3: Weighted Risk Composite
Weight rationale (70/20/10):
Fun Score
Fun captures four dimensions from distinct scientific frameworks, each measuring a different aspect of “this feels good.”
The Four Dimensions
| Dimension | Raw Scale | Normalized | Weight | Source |
|---|---|---|---|---|
| Subjective Pleasure | 0–5 (DRM) | × 20 → 0–100 | 45% | Kahneman's DRM |
| Flow State | 1–10 | (x−1)/9 × 100 | 25% | Csikszentmihalyi |
| Hedonic Purity | 0–10 | × 10 → 0–100 | 20% | Bentham's Calculus |
| Dopamine | 0–10 | × 10 → 0–100 | 10% | Neuroscience (NAc) |
Subjective Pleasure (45%)
Highest weight because it's what people actually report experiencing. Kahneman's DRM is the gold standard for experienced utility.
Flow State (25%)
Deep engagement isn't the same as pleasure, but it's a powerful predictor of satisfaction. Rock climbing scores low on raw pleasure but high on flow.
Hedonic Purity (20%)
The “aftertaste” question. Meth scores 10/10 on acute dopamine but ~0 on purity because tomorrow is hell. This dimension penalizes activities with harsh comedowns or regret.
Dopamine (10%)
Lowest weight intentionally. Dopamine measurements are noisy, lab-derived, and not directly comparable across categories (substance vs. activity vs. social).
The Fun Equation
Output Metrics
Primary: "Worth It" Index
This linear rescaling centers at 50 (break-even) and spreads symmetrically:
100
Fun=100, Risk=0
Best possible
50
Fun = Risk
Break-even
70
Fun=80, Risk=40
Good deal
0
Fun=0, Risk=100
Worst possible
Why not a raw ratio?
Because Fun÷Risk can't distinguish between “80 fun / 75 risk” (exciting but dangerous) and “20 fun / 15 risk” (boring but safe) — both give ≈1.07×. The difference metric captures the magnitudeof the deal, not just the ratio. “80/10” = 8× vs “20/2.5” = 8× — same ratio, but WorthIt is 85 vs 59.
Secondary: Fun-to-Risk Ratio (FRR)
The +15 additive smoothing solves three edge cases:
Division by zero
Activities with Risk=0 (reading a book) would give infinite FRR. The +15 keeps it bounded.
Extreme sensitivity
Without smoothing, Risk=0.1 vs Risk=0.2 would double the ratio. The +15 dampens this noise.
Boring-safe inflation
Reading (Fun=40, Risk=0) shouldn't outrank skydiving (Fun=85, Risk=42). Smoothing prevents this.
Tier Classification
Risk Tiers
| Score | Tier |
|---|---|
| 0–9 | Negligible |
| 10–29 | Low |
| 30–54 | Moderate |
| 55–79 | High |
| 80–100 | Extreme |
Fun Tiers
| Score | Tier |
|---|---|
| 0–14 | Boring |
| 15–34 | Mild |
| 35–59 | Fun |
| 60–79 | Thrilling |
| 80–100 | Euphoric |
The "Condom Problem" & Other Edge Cases
The best test of any scoring system is whether it handles the tricky cases intuitively. Here are the ones that shaped our formula:
Sex: Protected vs. Unprotected
The fun difference is small (maybe 80 vs 75 on pleasure), but the risk difference is massive (STI exposure, pregnancy risk). Does the formula capture this?
| Metric | With Condom | Without Condom |
|---|---|---|
| Fun Score | ~68 | ~72 |
| Risk Score | ~4 | ~13 |
| Worth It | ~82 | ~80 |
| FRR | 4.37× | 3.11× |
The small fun bump from no condom does NOT compensate for the risk jump. FRR drops from 4.37× to 3.11×. The math agrees with your doctor.
Smoking: The Chronic Risk Trap
A single cigarette has near-zero acute mortality (0.01 micromorts). Without chronic risk modeling, it would score as “safe.” But the microlife cost (0.4 per cigarette) feeds into Heq via the 1.25× conversion, giving a Health Score of ~21 — properly flagging it as a real risk.
Near-Zero Risk Activities
Reading a book: Fun=40, Risk=0. Without FRR smoothing, the ratio would be infinity. With the +15 offset: FRR = (40+15)/(0+15) = 3.67×— a reasonable “good deal” that doesn't break the leaderboard.
Scientific Foundations
Every dimension in our scoring system maps to established academic research. Here are the key papers and frameworks, organized by which part of the algorithm they inform.
Risk: Mortality Quantification
On Making Certain Decisions in an Uncertain World
Ronald A. Howard (1980) — Stanford University
Application: Introduced the micromort (μmt) unit — a one-in-a-million chance of death — enabling direct comparison of mortality risk across radically different activities.
Using Speed of Ageing and Microlives to Communicate the Effects of Lifetime Habits and Environment
David Spiegelhalter (2012) — BMJ 345:e8223
Application: Introduced the microlife concept (30 minutes of life expectancy). Provided conversion tables for chronic behaviors: 2 cigarettes = −1 microlife, 20 min exercise = +2 microlives.
Global Health Estimates: Leading Causes of DALYs
World Health Organization (2024) — WHO Global Health Observatory
Application: The DALY (Disability-Adjusted Life Year) framework: DALY = YLL + YLD. Provides disability weights for hundreds of conditions, enabling our chronic health scoring.
Risk: Perception vs. Reality
Perception of Risk
Paul Slovic (1987) — Science 236(4799):280–285
Application: Identified "dread" and "unknown" as the two primary factors driving risk perception. Explains why people fear flying (high dread, low control) more than driving despite 100× lower actual risk.
Prospect Theory: An Analysis of Decision Under Risk
Daniel Kahneman & Amos Tversky (1979) — Econometrica 47(2):263–291
Application: Demonstrated loss aversion: losses weigh ~2.25× heavier than equivalent gains. Informs why our risk weights are intentionally higher than fun weights in the composite formula.
Fun: Subjective Pleasure
A Survey Method for Characterizing Daily Life Experience: The Day Reconstruction Method
Daniel Kahneman et al. (2004) — Science 306(5702):1776–1780
Application: The DRM measures moment-to-moment net affect during activities. Provided our primary pleasure scale: sex = 4.7/5, socializing = 4.0, eating = 3.8, commuting = 2.6.
An Introduction to the Principles of Morals and Legislation
Jeremy Bentham (1789) — T. Payne and Son, London
Application: The Felicific Calculus: seven dimensions for quantifying pleasure — intensity, duration, certainty, propinquity, fecundity, purity, extent. Our 'Hedonic Purity' dimension derives directly from Bentham's 'purity' and 'fecundity' measures.
Fun: Flow & Engagement
Flow: The Psychology of Optimal Experience
Mihaly Csikszentmihalyi (1990) — Harper & Row
Application: Defined the flow state as total absorption where challenge matches skill. Activities inducing deep flow (climbing, music, surgery) score differently from pure sensory pleasure (eating, drugs). Our Flow State dimension captures this "engagement ≠ pleasure but matters" distinction.
Fun: Neuroscience
Drug Harms in the UK: A Multicriteria Decision Analysis
David J. Nutt, Leslie A. King, Lawrence D. Phillips (2010) — The Lancet 376(9752):1558–1565
Application: Used MCDA to score 20 drugs on 16 harm criteria plus intensity of pleasure. Alcohol ranked most harmful overall (72/100) despite lower pleasure intensity than heroin or cocaine. This framework directly inspired our multi-dimensional approach.
Effort Boosts Value of Subsequent Reward
Nature (2026) — s41586-025-10046-6
Application: Demonstrated that dopamine release for a reward increases if the preceding effort was higher. The "effort paradox" — explains why hard-earned fun (climbing summit, marathon finish) feels more rewarding than passive fun (scrolling social media).
Dopamine Baseline Data
| Activity/Substance | Dopamine Spike | Our Score (0–10) |
|---|---|---|
| Food (palatable meal) | 50–150% | 4–5 |
| Sex | 100–200% | 7–8 |
| Nicotine | ~150% | 5 |
| Exercise (vigorous) | 100–200% | 5–7 |
| Cocaine | ~1,000% | 8 |
| Methamphetamine | 1,000–10,000% | 10 |
Source: ScienceInsights 2025, Nature 2026. Dopamine is weighted at only 10% because these measurements are lab-derived, noisy, and not directly comparable across categories.
Data Sources
Activity-specific risk and fun values come from the following data repositories:
Micromort tables: Ronald Howard (1984), David Spiegelhalter (Winton Centre for Risk and Evidence Communication), and the Wikipedia Micromort article (comprehensive compilation).
GBD Results Tool (IHME): vizhub.healthdata.org/gbd-results/ — raw DALY/mortality data for habits like smoking, drinking, unsafe sex.
USPA (United States Parachute Association) — annual skydiving fatality reports.
DAN (Divers Alert Network) — annual diving fatality and injury reports.
NHTSA (National Highway Traffic Safety Administration) — motor vehicle crash data.
CDC — STI transmission rates, drowning statistics, smoking health effects.
UK ONS (Office for National Statistics) — drug poisoning deaths in England and Wales.
Nutt et al. 2010 (Lancet) — MCDA drug harm and pleasure scoring for 20 substances.
Himalayan Database — Everest summit fatality rates.
Kahneman DRM studies (2004) — subjective pleasure ratings for daily activities.
Limitations & Caveats
Not personalized
Your actual risk depends on age, health, skill level, location, and a hundred other factors. A 25-year-old experienced skydiver faces very different risk than a 60-year-old first-timer.
Jurisdiction-dependent
Cannabis is legal in some US states and carries severe penalties in Singapore. Our legal scores use a rough global average.
Fun is subjective
One person's "euphoric" is another's "meh." We use population-level research averages, not individual preference.
Dopamine data is noisy
Lab measurements of dopamine release are from animal studies and small human samples. That's why it gets only 10% weight.
Chronic risk assumes default frequency
The risk score for "smoking a cigarette" reflects one cigarette. A pack-a-day habit is 20× that. We note default frequencies where relevant.
Not advice
This is not medical, legal, or financial advice. It's a data visualization project. Consult actual professionals before making life decisions.
The Complete Formula
Built with obsessive attention to detail and questionable life choices.
Questions? Think our numbers are wrong? We probably agree. This is v1.