Score
0.50
Kelly
0.0%
1 / 12
Overseas STEM degree
Education / degree · 24-month horizon
Score 0.50 · expected value is negative under current assumptions (NO_GO)
With the resources you committed ($71,800 of equivalent capital) over a 24-month horizon, the prior win rate is 23.0%, the payoff ratio is b=2.16, giving an expected value of EV=-$19,616 and a risk-adjusted return of RAROC=-0.34. Under your current assumptions this is not worth acting on. The Kelly-optimal stake of 0.0% sits inside the 20% risk budget for your Working professional profile.
A two-year STEM master's at out-of-state international rates, plus living costs. The win rate is programme completion multiplied by the 34.9% H-1B selection rate, so it turns on a lottery you do not control — and from FY2027 that lottery is wage-weighted against entry-level salaries.
2 / 12
The three numbers
Win rate, payoff ratio and expected value decide most of the recommendation; Kelly, RAROC and score turn them into an action.
Win rate p
23.0%
Prior probability of success
Payoff ratio b
2.16 ×
b = E[win] / E[loss]
Expected value
-$19,616
EV = capital × (p × b − (1 − p))
Kelly
0.0%
(b·p − (1 − p)) / b
RAROC
-0.34
EV / (capital × σ)
Score
0.50
score = p × b
3 / 12
What you are committing
Equivalent capital $71,800, converted from five kinds of resource.
This opportunity carries $71,800 of equivalent capital. The largest single component is cash, at 95% of the total ($68,000). Cash has a clear floor, but keep 30% in reserve. Never put every liquid resource behind one decision.
How non-cash resources are priced
One hour of your own time is priced at $38 and one person-month of hired labour at $8,080. These are market shadow prices, not what you personally earn; changing them changes only the scale, not the ranking.
4 / 12
How this compares with the category
Your assumptions against the Beta(α, β) prior for this kind of decision.
Category mean win rate
56.7%
Beta(α=17, β=13)
Category payoff P50
1.50 ×
P90 4.70×
Your win rate of 23.0% is below the category baseline of 56.7% (-33.7%). That usually reflects conservative estimating rather than a weak opportunity, but it does pull the expected value down. The payoff ratio of 2.16× is above the baseline median of 1.50× — check that your success case has not quietly flattened the downside tail. Baseline source: FREOPP lifetime return by programme (US-EDU-12..17) + BLS CPS earnings by attainment (US-EDU-06..09).
Baseline source: FREOPP lifetime return by programme (US-EDU-12..17) + BLS CPS earnings by attainment (US-EDU-06..09)
5 / 12
Action plan
Sized to the risk budget of your profile: Working professional.
Recommendation: NO_GO — under your current win-rate and payoff assumptions the expected value is negative. This is not where your money and time should go.
- The expected value is negative. That is arithmetic, not a lack of nerve.
- Two ways forward: (a) restructure the inputs — less cash, more time, or a smaller scope — until the expected value turns positive; or (b) move the resources to a higher-ranked GO opportunity in your portfolio.
- Save this decision and recompute every three months. Category win rates and payoff ratios move over time.
6 / 12
Sensitivity analysis
How much the expected value moves when each variable is swung ±20%, ranked by impact. The top two are the ones worth researching before you commit.
- · The tornado analysis identifies Win rate (p) as the variable the conclusion is most sensitive to: a ±20% swing moves the expected value by $20,874. Spend your research budget here first.
- · The second most sensitive variable is Payoff ratio (b) (expected value swings $14,268); the remaining inputs matter markedly less.
- · A quick test for whether to keep researching: buy the information only if it costs less than one fifth of the expected-value swing it would resolve.
7 / 12
Monte Carlo distribution
10,000 simulated outcomes, seed demo-study-abroad-stem — rerunning with the same inputs gives the same distribution.
P5
-$133,738
Median (P50)
-$60,986
P95
$290,303
Probability of a positive outcome
23.5%
Across 10,000 Monte Carlo runs: P5 = -$133,738, P50 = -$60,986, P95 = $290,303, and the probability of a positive outcome is 23.5%. The distribution is right-skewed — a long upside tail carries most of the value.
8 / 12
Scenarios
Bull, base and bear, moving win rate and payoff ratio together by ±15%.
Scenario analysis (win rate and payoff ratio moved together by ±15%): bull = -$5,635, base = -$19,616, bear = -$31,991. The $26,356 gap between bull and bear measures how fragile your assumptions are — the wider it is, the more due diligence pays for itself.
9 / 12
Counterfactual matrix
What the expected value becomes at other combinations of win rate and payoff ratio — how wrong your assumptions can be before the answer changes.
The counterfactual grid shows that expected value turns non-negative once the win rate reaches 20%, or once the payoff ratio reaches 0.5×. Treat those two lines as the minimum bar any further evidence has to clear.
10 / 12
Assumptions
Every input behind the numbers above, in one place.
11 / 12
Reproducibility appendix
Everything needed to recompute this report yourself.
Core formulas
EV = capital × (p × b − (1 − p)) f* = (b·p − (1 − p)) / b RAROC = EV / (capital × σ)Python
from models.feasibility_model import expected_value, kelly_fraction, raroc, go_no_go
p, b, sigma, capital = 0.23, 2.16, 0.8, 71800
ev = expected_value(p, b, capital)
kelly = kelly_fraction(p, b)
rr = raroc(ev, capital, sigma)
print(ev, kelly, rr, go_no_go(ev, kelly, 0.20, rr))TypeScript
import { expectedValue, kellyFraction, raroc, goNoGo } from "decision-lab/lib/feasibility";
const p = 0.23, b = 2.16, sigma = 0.8, capital = 71800;
const ev = expectedValue(p, b, capital);
const kelly = kellyFraction(p, b);
const rr = raroc(ev, capital, sigma);
console.log(ev, kelly, rr, goNoGo(ev, kelly, 0.20, rr));Both implementations are in the repository and are kept in agreement by golden-parity tests: models/feasibility_model.py and lib/feasibility/.
12 / 12
Sources and disclaimer
- · Kelly J.L. (1956). A New Interpretation of Information Rate. Bell System Technical Journal.
- · Markowitz H. (1952). Portfolio Selection. The Journal of Finance.
- · Sharpe W.F. (1966). Mutual Fund Performance. The Journal of Business.
- · Kahneman D. & Tversky A. (1979). Prospect Theory. Econometrica.
- · Industry baseline: FREOPP lifetime return by programme (US-EDU-12..17) + BLS CPS earnings by attainment (US-EDU-06..09)
- · Decision Lab methodology
- · Full business plan (PDF)