CapEx Intelligence — OilyGiant

CapEx Budget

$100M

Development capital allocated

200 wells · $500K each

Recommended Region

Region 1

Only region within risk mandate

▼ 0.20% loss risk

Expected Profit

$6.65M

Bootstrap mean · 1,000 simulations

CI: [$1.81M, $12.06M]

Best REQM

0.89

Region 1 · Linear Regression

42x more precise than Region 0

Risk Mandate

< 2.5%

Max tolerated loss probability

2 of 3 regions failed

🛢️

Final Recommendation — OilyGiant

After predictive modeling and 1,000 bootstrap simulations, Region 1 is the only viable candidate for CapEx deployment. It delivers the highest expected profit at $6.65M with a loss probability of just 0.20% — 12.5x below the 2.5% mandate. Regions 0 and 2 were disqualified on risk grounds.

Region 0 DISCARDED

$6.06M

Bootstrap Profit

$6.06M avg

Loss Risk

2.50%

95% CI

[$0.10M, $12.46M]

Model REQM

37.76

Region 1 RECOMMENDED ✓

$6.65M

Bootstrap Profit

$6.65M avg

Loss Risk

0.20%

95% CI

[$1.81M, $12.06M]

Model REQM

0.89

Region 2 DISCARDED

$5.85M

Bootstrap Profit

$5.85M avg

Loss Risk

2.60%

95% CI

[$-0.01M, $12.12M]

Model REQM

40.15

Bootstrap Profit Distribution

Expected profit per region · 1,000 simulations

BOOTSTRAP

$6.65M

Region 1 expected profit
highest among viable regions

$0.80M

Margin over Region 0
+13.2% better outcome

Loss Risk vs. Mandate

% probability of negative return · threshold: 2.5%

RISK

0.20%

Region 1 risk
12.5x below mandate

2/3

Regions eliminated
failed risk threshold

Linear Regression — REQM per Region

Root Mean Squared Error · lower = more precise predictions · 75/25 train-validation split

MODEL ACCURACY

Region 0

37.76

Avg predicted: 92.40K barrels

Region 1

0.89

Avg predicted: 68.71K barrels

Region 2

40.15

Avg predicted: 94.77K barrels

REQM Comparison

Region 1 is 42x more precise than Region 0

CHART

Predicted vs. Real Volume

Average reserves per region (K barrels) · minimum breakeven = 111.1K

VOLUME

111.1K

Minimum barrels/well
to break even at $500K cost

All 3

Regions below breakeven
selective ML choice is essential

Why Region 1's Model is Exceptional

REQM 0.89 vs industry-typical 35–40 · a structural advantage in predictive precision

INSIGHT

Prediction Error

±0.89K

barrels per well

vs Region 0

42x

more precise

vs Region 2

45x

more precise

Signal Type

Linear

f2 drives volume

Bootstrap Simulation Results — 1,000 Iterations

500 random wells sampled per iteration · 200 best selected · risk = % of negative outcomes

STOCHASTIC

Region 0

$6.06M

95% CI: [$0.10M → $12.46M]

Risk: 2.50%threshold 2.5%

⚠ At the limit — discarded

Region 1 — SELECTED

$6.65M

95% CI: [$1.81M → $12.06M]

Risk: 0.20%12.5x below limit

✓ Approved — invest here

Region 2

$5.85M

95% CI: [$-0.01M → $12.12M]

Risk: 2.60%exceeds threshold

✗ Discarded — risk too high

Confidence Intervals — All Regions

Lower bound · Mean · Upper bound of bootstrap distribution

95% CI

Risk Threshold Analysis

Loss probability vs 2.5% mandate · red line = disqualification boundary

THRESHOLD

Profit Without Model vs With Model

Naive random selection vs ML-guided top-200 well selection

MODEL VALUE

Negative

All regions unprofitable
with random well selection

ML Required

Model selection is not optional —
it is the business case

$33.59M

Best raw profit estimate (R0)
before risk adjustment

✅

Investment Decision — Region 1 Approved

The stochastic analysis confirms that Region 1 is the only region meeting OilyGiant's risk mandate. With a loss probability of 0.20% — against the 2.5% ceiling — and the highest expected return of $6.65M, it satisfies both the financial and the risk objectives simultaneously. The model's REQM of 0.89 provides an additional layer of confidence: predictions are near-deterministic for this region, making well selection highly reliable.

Data Ingestion

300K well records across 3 regions. Zero duplicates, zero missing values. Features: f0, f1, f2 (geological signatures) + product (reserve volume in K barrels).

Predictive Model

Linear Regression trained on 75K records per region, validated on 25K. Region 1 achieved REQM 0.89 — indicating a near-linear relationship between features and reserves.

Bootstrapping

1,000 iterations sampling 500 wells each. Top 200 selected per iteration by predicted volume. Profit and loss probability computed across the full distribution.

Risk Filter

Regions with loss probability ≥ 2.5% automatically disqualified. Only Region 1 cleared the threshold. Final recommendation: deploy $100M CapEx in Region 1.

Decision Matrix

Profit vs Risk · Region 1 dominates the efficient frontier

DECISION

Summary Scorecard

All criteria evaluated per region

SCORECARD

Criterion	Region 0	Region 1	Region 2

Oil Well SelectionRisk-Adjusted CapEx Strategy

Oil Well Selection
Risk-Adjusted CapEx Strategy