Table of contents
1. How to play
Goal: move your asset along the spiral while maximizing value and cash, managing failures, repeats, dilution, and (very rare) jumps such as repurposing.
Board and node selection
- The board is a spiral: each circle is a project stage.
- The circle with a yellow border is your current stage.
- To play, select your stage (click the yellow circle or use the “Select my stage” button).
Actions and bet (“Budget”)
- Each stage has 1+ available tasks (e.g., “Target validation → Assay”).
- You set a Budget ($M): this is your bet for that attempt.
- The budget is first subtracted from cash; then the outcome (success/fail) is sampled based on the probability.
- Shortcuts: Min (minimum cost), 2×, Max (all‑in).
Task types
step
A “classic” step with base probability Pbase at minimum cost baseCost. Spending above the minimum increases the probability with diminishing returns.
perMol
A “screening” task: the budget is converted into y (number of molecules/candidates tested). Probability grows with y, but if pMol is small, success remains difficult.
Finance: Sell and Raise
- Sell %: sell a fraction of your asset and receive
proceeds = sellPct × assetValue(stage). - Sell 100% & Restart: full exit; you cash out and restart from stage 0 with a new asset.
- Raise (dilution): raise capital but dilute equity. Default:
equity_sold = R / (value + R), and your equity is multiplied by(1 - equity_sold).
“Pure pre‑seed” mode: it is normal to be unable to self‑fund advanced ADME/PK, GLP tox, or clinical work.
The game is designed to force decisions on fundraising, partnerships, or exit.
2. What all stages represent
Stages represent typical “inflection points” in a drug discovery program: at each transition, evidence, derisking, and therefore the expected value of the asset should increase.
| ID | Stage | Meaning (practically) |
|---|---|---|
| 0 | Start / TPP | Define the Target Product Profile: population, endpoints, route, dosing, safety/CMC constraints, and development strategy. |
| 1 | Disease & Biomarkers | Understand the biological context and measurable translational biomarkers (mechanism signal and/or response). |
| 2 | Target ID | Select the target (genetics/omics/literature) and assess causal plausibility vs. the disease. |
| 3 | Target Validation | Experimental validation (in appropriate models) + tractability and risk assessment (on‑target/off‑target). |
| 4 | Assay Platform | Build robust, reproducible assays (HTS‑ready) with controls, QC, and reliable readouts. |
| 5 | Primary Screen | Screen libraries: many molecules are tested; the real hit rate may be extremely low. |
| 6 | Hit Confirmation | Confirm hits, counterscreen, eliminate artifacts, and start assembling a “hit series”. |
| 7 | IP & Triage | Assess novelty and freedom‑to‑operate; prioritize chemical series and define the patent strategy. |
| 8 | MoA / Engagement | Demonstrate target engagement and a mechanistic rationale (MoA) consistent with biomarkers. |
| 9 | Safety Profiling | Early liabilities: off‑target profile, cardiotoxicity, preliminary genotoxicity, etc. |
| 10 | ADME/PK | Absorption, metabolism, clearance; exploratory in vivo PK and compatibility with the TPP dosing. |
| 11 | Hit-to-Lead | Fast iterations: potency, selectivity, solubility, metabolic stability; reduce ADME risk. |
| 12 | Lead Optimization | Multi‑objective optimization (efficacy/PK/safety/CMC). Often the toughest part of preclinical work. |
| 13 | Candidate Selection | Select candidate (and backup) with Go/No‑Go criteria: exposure, safety window, manufacturability. |
| 14 | CMC Scale-up | Process chemistry, solid form, formulation, larger batches, and quality controls. |
| 15 | IND-enabling (GLP) Tox | Regulatory studies (GLP tox + safety pharmacology) to support IND/CTA submission. |
| 16 | Phase I | First‑in‑human: safety, tolerability, PK; only rarely leads directly to “Launch”. |
| 17 | Phase II/III + Launch | Proof‑of‑Concept, pivotal trials, approval and launch. High spend and residual clinical risk. |
“Jumps” (repurposing) represent situations where a known asset or existing molecule enables skipping some stages,
but in the game they are intentionally rare and expensive.
3. Costs & probabilities
The values below are tuned for a pure pre‑seed mode (low starting cash) and to make progression harder than the “easy” demo: many phases require decisions on fundraising, partnerships, or exit.
Mathematical model used by the game
// TASK "step"
p = 1 - (1 - pBase)^(spend / baseCost) // spend >= baseCost
p capped at 0.99
// TASK "perMol"
y = floor((budget - baseCost) / perMolCost), y >= 1
p = 1 - (1 - pMol)^y // prob. of ≥ 1 success
// RAISE (dilution)
equity_sold = R / (value + R)
new_equity = old_equity * (1 - equity_sold)
Interpretation: Pbase is the success probability at the minimum spend. Spending more increases p with diminishing returns.
For perMol tasks, pMol is often extremely small: even thousands of tests do not guarantee success.
Default parameters (pure pre‑seed)
Starting cash
- startCashM = 1.5 ($1.5M)
Asset value (for Sell/Raise)
“Value” is a simplified proxy of asset value as derisking progresses.
| ID | Stage | Asset value (M$) |
|---|---|---|
| 0 | Start / TPP | 1.5 |
| 1 | Disease & Biomarkers | 3 |
| 2 | Target ID | 6 |
| 3 | Target Validation | 12 |
| 4 | Assay Platform | 18 |
| 5 | Primary Screen | 25 |
| 6 | Hit Confirmation | 35 |
| 7 | IP & Triage | 45 |
| 8 | MoA / Engagement | 60 |
| 9 | Safety Profiling | 80 |
| 10 | ADME/PK | 110 |
| 11 | Hit-to-Lead | 160 |
| 12 | Lead Optimization | 240 |
| 13 | Candidate Selection | 320 |
| 14 | CMC Scale-up | 450 |
| 15 | IND-enabling (GLP) Tox | 650 |
| 16 | Phase I | 900 |
| 17 | Phase II/III + Launch | 2500 |
Minimum costs and base probabilities (“step” tasks)
| From → To | Task | baseCost (M$) | Pbase | Fail → |
|---|---|---|---|---|
| 0 → 1 | Define TPP & strategy | 0.10 | 0.85 | 0 |
| 1 → 2 | Biomarkers + disease map | 0.25 | 0.75 | 1 |
| 2 → 3 | Target ID package | 0.45 | 0.65 | 1 |
| 3 → 4 | Target validation | 0.80 | 0.55 | 2 |
| 4 → 5 | Assay dev + QC | 0.50 | 0.75 | 4 |
| 7 → 8 | IP/triage + series choice | 0.20 | 0.80 | 6 |
| 8 → 9 | MoA/engagement package | 0.60 | 0.60 | 7 |
| 9 → 10 | Early safety profiling | 0.90 | 0.55 | 8 |
| 10 → 11 | ADME/PK cycle | 1.80 | 0.55 | 9 |
| 11 → 12 | Hit-to-lead iteration | 3.50 | 0.50 | 10 |
| 12 → 13 | Lead optimization | 6.00 | 0.35 | 11 |
| 13 → 14 | Candidate selection | 1.50 | 0.55 | 12 |
| 14 → 15 | CMC scale-up readiness | 4.00 | 0.50 | 13 |
| 15 → 16 | IND-enabling (GLP) tox | 12.0 | 0.40 | 14 |
| 16 → 17 | Phase I (FIH) → Launch | 20.0 | 0.06 | 15 |
| 17 → 17 | Lifecycle / RWE (repeat) | 40.0 | 0.85 | 17 |
“perMol” tasks (screen and confirm)
| From → To | Task | baseCost (M$) | perMolCost (M$) | pMol | Fail → |
|---|---|---|---|---|---|
| 5 → 6 | Primary screen (y molecules) | 0.35 | 0.00001 (=$10) | 1.5e-7 | 5 |
| 6 → 7 | Confirm hits (y candidates) | 0.25 | 0.0008 (=$800) | 4e-5 | 5 |
Repurposing (jump)
Repurposing is treated as an “expensive lottery”: it can skip stages, but with very low probability. In the game it is implemented as multiple outcomes that sum to 1.
| From | Minimum cost (M$) | Outcomes |
|---|---|---|
| 0 (Start) | 1.20 | 0.3% → 17 (late) • 0.9% → 16 • 1.8% → 15 • 97.0% → fail (stay at 0) |
| 1 (Disease) | 1.20 | 0.2% → 17 (late) • 1.0% → 16 • 2.0% → 15 • 96.8% → fail (stay at 1) |
Important: these are “didactic” numbers. They are not intended as predictions for any single program,
but to reproduce orders of magnitude and the need for capital/partnerships in later stages.
4. Strategy suggestions
Practical goal: maximize value per $ spent in early phases, then use Raise/Sell
intelligently to overcome bottlenecks without collapsing your equity.
1) Do not overspend where P(success) is already high
- If a step task has P > 70% at minimum cost, spending 3× is rarely efficient.
- Use Min (or slightly above Min) often in early transitions: you need runway for the truly hard steps.
2) Recognize “walls” and plan multiple attempts
- “Walls” are points where you are not blocked by a wrong choice or rule, but where progress depends mostly on luck because the success probability is low.
- For low‑probability tasks (e.g., screening), think in terms of a series of attempts, not a single roll.
- Rule of thumb: if success probability is p, the average attempts per success is about 1/p.
- Before starting a screen, estimate how much runway you have for 5–15 attempts and decide whether you need capital.
3) Raise vs Sell: use each at the right time
- Raise is preferable when you need substantial runway and expect to increase value later (but you dilute).
- Sell % is closer to a partial deal/exit: you monetize value now, but reduce future upside.
- Avoid using Sell 10% repeatedly as an ATM: you can quickly end up with too little equity.
4) “Derisk then raise”: increase value before diluting
- If possible, advance 1–2 stages with contained costs before a major raise.
- Raising too early (low value) costs a lot of equity for little cash.
5) Repurposing: rare move, only if you can afford all‑in
- Repurposing is implemented as a low‑probability event: it often fails.
- Use it when you have a strategic reason (or for a controlled “high‑variance” bet), not as a default on every run.
Fast heuristic (pre‑seed): build value up to an inflection point (e.g., target validation or a robust assay),
then decide whether to Sell (deal) or Raise (round) to finance screening and beyond.
5. Typical numbers (ranges by therapeutic area)
The figures below are orders of magnitude, not “laws of nature”.
They vary with: time period, definitions (program vs indication), modality mix (small molecule/biologic),
target quality, study design, and dataset bias.
Probability of approval: differences across areas
The table uses a practical indicator: LOA from Phase I to Approval (Likelihood of Approval starting from Phase I). It is one of the most intuitive ways to compare how “hard” clinical development is across areas.
| Area | LOA Phase I→Approval | Why it is higher or lower (on average) |
|---|---|---|
| Oncology | ~5% | Heterogeneous biology, hard endpoints (OS) often long, fragmented populations, high complexity of combinations. Some sub‑areas can perform better (e.g., certain immuno‑oncologies). |
| Neurology / CNS | ~6% | Endpoints often clinical/behavioral and noisy, strong placebo effects, slow progression, difficulty measuring target engagement and reaching the target (BBB), high patient heterogeneity. |
| Psychiatry | ~7% | Heterogeneous diagnoses and phenotypes, subjective outcomes, high inter‑site variability, strong placebo effects. |
| Infectious diseases | ~13% | Often clear endpoints (viral load/clearance, microbiology), shorter trials, stronger cause‑effect link. Risks: resistance, episodic enrollment, fast changes in standard‑of‑care. |
| Hematology | ~24% | Biomarkers and response measures are often more direct (e.g., hematological parameters), defined populations, earlier signals and trials that are comparatively more informative. |
| Metabolic | ~16% | Standardized biomarkers (HbA1c, lipids, etc.) and large populations; however, safety requirements and large trials can increase cost and operational risk. |
| Autoimmune | ~11% | Good translatability in some indications and available biomarkers in subgroups; but heterogeneity, comorbidities and competition with effective therapies can make trials difficult. |
| Ophthalmology | ~12% | Often measurable and localized endpoints, imaging, relatively controllable trials; constraints on local safety and specific populations. |
| Cardiovascular | ~5% | Trials are often very large and long, hard clinical endpoints, high safety requirements, strong standard‑of‑care: incremental benefit must be compelling. |
| Rare diseases | ~8–17% | When targets are genetically validated and effects are large, success can be higher; but enrollment and natural history are hard, trials are small, and design errors are costly. Orphan incentives help. |
| Vaccines | ~30% (order of magnitude) | Often immunological endpoints and correlates of protection, consolidated development strategies; average probability can be higher than many small molecules, but depends on pathogen and context. |
Why PoS changes so much
Scientific factors
- Target validation: human genetics and causality increase PoS; “fragile” targets decrease it.
- Translational biomarkers: engagement/response measurements improve Go/No‑Go decisions.
- Biological heterogeneity: more heterogeneity → more noise → less statistical power.
- Safety window: some classes/targets have intrinsically higher risk.
Clinical & operational factors
- Endpoints: objective and fast (e.g., load) tend to improve PoS; subjective/long endpoints reduce it.
- Placebo and variability: very high in CNS/psychiatry, complicating signal detection.
- Enrollment: rare diseases and some rare cancers create operational bottlenecks.
- Standard‑of‑care: the stronger it is, the harder it becomes to demonstrate incremental benefit.
How to use these numbers in the game
- If you choose an “oncology/CNS mode”, make clinical stages harder (and/or increase costs) and value biomarkers more.
- If you choose an “infectious/vaccines mode”, increase average PoS but accept exogenous shocks (standard‑of‑care, resistance).
- For rare diseases, increase PoS when genetics/target validation is strong, but make enrollment more costly/fragile.
Design suggestion: instead of a single PoS for all, select a therapeutic profile at game start that applies multipliers
to probabilities and costs (e.g., Oncology: PoS↓, costs↑; Rare genetic: PoS↑ for MoA/engagement).
6) Bibliographic sources
The difficulty choices (high attrition, increasing costs, rarity of “miracle” repurposing, and low overall clinical PoS) are consistent with established results on costs/timelines and success probabilities in pharma R&D. Below is a minimal bibliography supporting the orders of magnitude used.
R&D costs and timelines
- DiMasi JA, Grabowski HG, Hansen RW. Innovation in the pharmaceutical industry: New estimates of R&D costs. Journal of Health Economics, 2016. doi:10.1016/j.jhealeco.2016.01.012
- Wouters OJ, McKee M, Luyten J. Estimated Research and Development Investment Needed to Bring a New Medicine to Market, 2009–2018. JAMA, 2020. doi:10.1001/jama.2020.1166
- Schuhmacher A, Gassmann O, Hinder M. Changing R&D models in research-based pharmaceutical companies. Journal of Translational Medicine, 2016. doi:10.1186/s12967-016-0838-4
Success probabilities and attrition (preclinical → clinical → approval)
- Hay M, Thomas DW, Craighead JL, Economides C, Rosenthal J. Clinical development success rates for investigational drugs. Nature Biotechnology, 2014. doi:10.1038/nbt.2786
- Wong CH, Siah KW, Lo AW. Estimation of clinical trial success rates and related parameters. Biostatistics, 2019. doi:10.1093/biostatistics/kxx069
- Paul SM et al. How to improve R&D productivity: the pharmaceutical industry's grand challenge. Nature Reviews Drug Discovery, 2010. doi:10.1038/nrd3078
- Kola I, Landis J. Can the pharmaceutical industry reduce attrition rates? Nature Reviews Drug Discovery, 2004. doi:10.1038/nrd1470
- Scannell JW, Blanckley A, Boldon H, Warrington B. Diagnosing the decline in pharmaceutical R&D efficiency. Nature Reviews Drug Discovery, 2012. doi:10.1038/nrd3681
Repurposing
- Pushpakom S et al. Drug repurposing: progress, challenges and recommendations. Nature Reviews Drug Discovery, 2019. doi:10.1038/nrd.2018.168