POKER SAMPLE SIZE

How Many Poker Hands Do You Need Before the Data Means Anything?

Enough for what?

50,000 hands can be plenty for a common preflop stat and still be thin for a rare river spot. Confidence belongs to the finding, not the file size.

There is no magic hand count. Total hands are not the same as opportunities for a specific decision.

EDUCATIONAL MODEL · NO UNIVERSAL THRESHOLD

NO UNIVERSAL THRESHOLD

There is no single “enough hands” number

The useful question is never “how many hands is enough?” It is always enough for what. Different metrics accumulate evidence at different rates.

DIFFERENT METRICS, DIFFERENT SAMPLES

Winrate needs a very different sample from VPIP

VPIP is observed in a large share of hands. Winrate is a noisy result metric. The same total-hand count can be useful for one and almost useless for the other.

Winrate

Result metric. High variance. Slow to stabilize.

VPIP

Common aggregate frequency. Accumulates in most hands.

COMMON DECISIONS

Common preflop decisions accumulate quickly

Positions that occur every orbit, and preflop choices that appear often, build opportunity counts faster than later-street families.

Speed of accumulation is not a verdict. It only tells you how soon a common frequency can become worth inspecting.

RARE DECISIONS

Rare river spots accumulate slowly

A river decision that requires a specific board, position, and action line appears in a small subset of the file. Total hands grow while opportunities stay thin.

Treat rare families as slow evidence, not as automatic leaks hiding in a large database.

DATABASE SIZE VS DECISION OPPORTUNITIES

Why opportunities matter more than total hands

Total hands describe the file. Opportunities describe how often the defined decision actually occurred.

COMMON SPOTAppears often
RARE SPOTAppears rarely

Qualitative frequency, not a measured rate and not a finding.

A leak is judged in a defined spot. Only the times that spot occurs count as evidence for that leak.

WHAT VOLUME CAN AND CANNOT DO

These ranges describe typical uses of a file size. They are not confidence grades and they do not declare a database reliable.

10k hands

What this volume can help with

  • Broad sanity checks
  • Very common aggregate frequencies

What it still cannot prove

  • Rare decision families
  • Short-term winrate conclusions

25k hands

What this volume can help with

  • More aggregate stability than a tiny sample
  • Common preflop frequencies with more context

What it still cannot prove

  • Spot-independent certainty
  • Rare later-street families

50k hands

What this volume can help with

  • Substantial evidence for some common spots
  • A file with room to separate volume from opportunity

What it still cannot prove

  • Universal certainty across every street
  • Thin river contexts that still occur rarely

100k+ hands

What this volume can help with

  • A large database for common frequencies
  • Room to inspect repeating decision families

What it still cannot prove

  • Universal certainty
  • Proof that every rare spot is well evidenced
FINDING-LEVEL EVIDENCE

How Performance Coach calculates confidence

The product evaluates evidence at the finding level. Sample strength is not the same as strategic confidence.

  • SAMPLE STRENGTH

    How much opportunity volume exists for a defined decision.

  • STRATEGIC CONFIDENCE

    A later judgment about decision quality. Not implied by file size.

  • INSUFFICIENT EVIDENCE

    A valid result when the spot is still too thin to support a claim.

QUESTIONS

Sample size, asked directly

Are 50,000 hands enough?

Enough for what? Common preflop frequencies and rare river decisions do not share a sample-size threshold.

Why do you talk about opportunities instead of total hands?

A leak is judged in a defined spot. Only the times that spot occurs count as evidence for that leak.

Check the sample
before trusting the number.

Use the calculator to see how total hands and decision opportunities diverge.

Educational estimate. Not a production confidence model.