What is Ollin Study
A GTO study product for heads-up MTT poker, built on exact solver output — no interpolation, no fabricated numbers.
Ollin Study is the user-facing study product for Ollin's solver data. It covers heads-up MTT poker at 10–100bb effective, and every number on every screen traces back to an exact solve:
- Preflop — the CEO's HRC (HoldemResources Calculator) solves: 8-max MTT with antes, ChipEV metric, ten symmetric stack depths (10 / 15 / 20 / 25 / 30 / 40 / 50 / 60 / 75 / 100bb). You walk the real action tree; nothing is abstracted between depths.
- Postflop — the sealed
berserk_v1backbone: 44,528 solves (242 heads-up arriving spots × 184 canonical flops), each converged to a Nash distance of ≤ 0.3% of the pot, published under the immutable dataset stamp20260721T131354Z.
The product's design rule is honesty about coverage: what is solved is browsable, what is not solved is visibly locked with the reason stated — never interpolated, never guessed.
The experience in one sentence
Pick a stack depth, walk the 8-max preflop tree action by action, and when a line closes to a heads-up flop, cross the bridge into the solved postflop family — 184 flops, per-combo mixed strategies, EVs, and pot geometry.
Where to go next
The 60-second tour
From /study to a per-combo postflop drilldown in six clicks.
Run it locally
The two-process dev stack, fixtures vs GCS mode, and the 127.0.0.1-not-localhost gotcha.
User guide
One page per screen: Study, Library, family view, strategy browser.
Architecture
Teacher solver → sealed backbone → extraction → serving API → this app, with the AGPL boundary drawn where it actually is.
API reference
Every endpoint of ollin-study-api with real response JSON.
Roadmap
Serving tiers root → tree → live, and the coaching-SaaS slots.
Current serving status
| Layer | Status |
|---|---|
| Preflop trees (10 depths, full 8-max) | Extracted (preflop-v0 tier) |
| Postflop flop root nodes | Serving now (v0-roots tier, packs upload continuously) |
| Postflop full trees (turn/river nodes) | Roadmap — Tier-A extraction campaign (draft, pending pilot) |
| Real-time solves | Roadmap — live tier |
The app itself never breaks on missing data: with no API configured (or an unreachable one) every screen renders from a deterministic local mock dataset and shows an amber "Demo data" ribbon so fabricated numbers can never be mistaken for solves.