Optiver challenge · AdaHack 2026

Cheap carbon credits can fail together.

Our goal

Deliver 100,000 tCO₂e

Rules we used

budget: max $1,000,000
success goal: at least 95.0%
tests: 10,000 each

Our spread-out portfolio

13 projects · 166,671 tonnes bought

99.3% of simulated tests reached the goal

$181,659

We compare the cheapest ways to buy credits with a portfolio spread across more projects. See what the extra cost buys when projects fail.

Explore the comparison Download summary

Our result

Build in room for failure.

13 projects · 166,671 tonnes bought · $1,000,000 spending limit · met our chosen success rate at every risk level

Tonnes we need to deliver

100.0KtCO₂e

Our chosen success goal: 95.0%

Cost to buy these credits

$181,659

Budget: $1,000,000

Tests that reached the goal

99.3%

Estimated range from test sampling: 99.08–99.42%

What if projects fail together?

This changes how often projects fail together in our simulation. It is an assumption, not a measured rate. The credits bought and their cost stay the same.

Where too many credits share a risk

Registry concentrationVCS
60.0%
Country concentrationChina
35.0%
Project-type concentrationSolar - Centralized
35.0%

These percentages show where the credits are concentrated. They do not show how likely failure is.

Compare the options

Why buy more than the cheapest option?

All three options aim to deliver 100,000 tonnes. The cheapest buys only enough credits to reach that number if nothing fails. The second accounts for average losses. Our option buys extra credits from more projects so one failure does less damage.

StrategyCost to buyTests that reached the goal

Lowest upfront cost2 projects · 100.0K tonnes bought

$94,277

68.3%

Lowest cost for average delivery1 project · 117.6K tonnes bought

$115,294

85.1%

Our spread-out portfolio13 projects · 166.7K tonnes bought

$181,659

99.3%

These percentages come from simulated failures at the selected shared-risk level. Changing the level does not change what each option buys or costs.

Cost of confidence

How much does a better chance cost?

Each point shows the cheapest portfolio our search found for a different success goal. A higher goal usually means buying more credits in case some fail.

Download chart data

Results from new tests at the selected shared-risk level. A portfolio must meet its goal at every risk level to count as a pass.
Success goal	Cost	Projects	Tests that reached 100,000	Conservative estimate	Result
80.0%	$140,197.45	9	83.7%	83.00%	Met success rate at all risk levels
85.0%	$148,534.58	13	90.7%	90.10%	Met success rate at all risk levels
90.0%	$161,727.02	8	94.5%	93.98%	Met success rate at all risk levels
95.0%	$181,658.56	13	99.3%	99.08%	Met success rate at all risk levels
99.0%	$282,130.45	15	100.0%	99.96%	Met success rate at all risk levels

We ran 10,000 new tests for each portfolio at each risk level. Our search tried a limited set of options, so a cheaper solution may exist. Points that missed their goal are left off the line. The conservative estimate allows for uncertainty from testing a sample of possible outcomes.

Exposure

Where the tonnes come from

Circle area represents purchased tonnes. Select a country to inspect its projects.

Portfolio

166,671 t

across 6 countries

Country-level locations; the dataset has no project coordinates. Public-domain outline: Natural Earth.

Country exposure

China

58,335t

Share: 35.0%
Projects: 4
Avg per project: 14,584

Ningxia Angli Lingwu Photovoltaic Grid Connected Power Plant Project
VCS1143 · VCS · 16,667 t
Solar - Centralized
CGN Hami Phase I 20MWp Grid-connected PV Power Plant Project
VCS1248 · VCS · 8,334 t
Solar - Centralized
Inner Mongolia Sunjiaying 50.25MW Wind Power Project
VCS395 · VCS · 16,667 t
Wind
Xundian Jinfeng 12.6MW Hydropower Project
VCS1032 · VCS · 16,667 t
Hydropower

Shared risks

Where we still rely on the same groups

Our portfolio uses several countries, companies, registries and project types. But 60% of the tonnes still come from one registry. If many projects in that group fail together, we could miss the goal.

country

Top 6 by purchased tonnes

Worst-case check

What if every project in one group fails?

Pick a group to see how many tonnes would remain if all its projects failed at once. Some losses are covered by buffer pools. This shows the size of the damage, not how likely it is.

If all projects in VCS failed

66,668 tCO₂eTarget missed

0Target: 100,000166,671 bought

Tonnes lost after buffer recovery: 100,003 t
Projects affected: 8
Below target: 33,332 t

This assumes every project in the selected group fails at once, with some tonnes returned by buffer pools. It shows how bad that event would be, not how likely it is. The portfolio does not change.

Test results

How often did we reach the goal?

We simulated 10,000 possible outcomes for each shared-risk level. The figures below describe those tests, not real-world guarantees.

Results for our portfolio

Metric		Definition
Reached the goal	99.3%	Share of tests delivering at least 100,000 tonnes
	99.08%–99.42%	Range caused by testing a sample of possible outcomes
	116.7K tonnes	Only 5% of tests delivered less than this
	77 tonnes	Average gap below the goal across all tests

Projects

What we would buy

Search and filter the projects in our portfolio. The map and totals still show every project, even when you filter this table.

Projects in our portfolio

Search by project, company or ID. You can combine the filters.

Search projectsFilter by countryFilter by project typeSort projects

13 of 13 projects · 166,671 tonnes · $181,658.56 in this view

Project	Country	Type	Registry	Tonnes	Cost
Xundian Jinfeng 12.6MW Hydropower Project VCS1032	China	Hydropower	VCS	16,667	$20,000.40
Canakkale WPP GS906	Türkiye	Wind	GOLD	16,667	$19,833.73
Inner Mongolia Sunjiaying 50.25MW Wind Power Project VCS395	China	Wind	VCS	16,667	$18,333.70
Monjolinho Energética S/A Hydropower Plant Project (Alzir dos Santos Antunes) ACR0177	Brazil	Hydropower	ACR	16,295	$17,598.60
2x50 MW Orange Suvaan Solar Photovoltaic Power Project in Maharashtra India GS5928	India	Solar - Centralized	GOLD	16,667	$17,500.35
Ningxia Angli Lingwu Photovoltaic Grid Connected Power Plant Project VCS1143	China	Solar - Centralized	VCS	16,667	$17,333.68
Vishnuprayag Hydro-electric Project (VHEP) by Jaiprakash Power Ventures Ltd.(JPVL) VCS173	India	Hydropower	VCS	16,667	$16,333.66
Srepok 1 Solar Power Project VCS1974	Vietnam	Solar - Centralized	VCS	16,667	$15,500.31
InfraVest Changbin and Taichung bundled Wind Farms Project - Taiwan (300190) GS472	Taiwan	Wind	GOLD	11,339	$15,307.65
CGN Hami Phase I 20MWp Grid-connected PV Power Plant Project VCS1248	China	Solar - Centralized	VCS	8,334	$8,750.70
GUNAYSE HPP GS636	Türkiye	Hydropower	GOLD	5,700	$6,441.00
Energy from renewables VCS296	India	RE Bundled	VCS	5,328	$5,328.00
2 x 3.5 MW Ullunkal Hydro Power Project in Kerala, India. VCS869	India	Hydropower	VCS	3,006	$3,396.78

Behind the numbers

How we tested this

What these results mean

Data source

We used the challenge dataset, which combines carbon-credit projects from UC Berkeley with made-up prices and failure ratings. The prices are not real market prices.

Failure model

We assume an entire project can fail. A buffer pool returns half its lost tonnes; a past reversal increases its failure chance by 50%. Projects can fail together when they share a country, developer, registry or type. We chose how strong that shared risk is because the challenge does not say.

Evaluation

We tried a small set of portfolio rules using 2,000 simulated outcomes, then checked the chosen portfolio on 10,000 new outcomes at each risk level. We cannot claim it is the cheapest possible portfolio. The percentages describe our simulation, not guaranteed delivery.

No real-world claims

This is an offline challenge demo. We have not bought or retired any credits, and we have not measured an emissions reduction. The results show what happened under our assumed risks.

PDF summary report.json portfolio.csv