Reading Poker Analysis Reports: HUD Stats, Sample Size, and Leak Triage

VPIP 34% is not automatically a leak—it might be a 40-hand sample from a loose live table. Poker analysis reports—tracker dashboards, session summaries, AI leak flags—require sample calibration and cross-stat reads before you change strategy.

Most players treat reports reactively: one red number, one drastic range change, same leak next month. RiverOdds, nlh.poker, TPP Academy, and Seeker Start converge on a cluster read—VPIP with PFR, WTSD with W$SD, c-bet with fold-to-c-bet—plus minimum hand thresholds before any stat counts as reliable.

Primary Stats (What Each Measures)

Stat	Full name	What it tells you	Common misread
VPIP	Voluntarily Put $ In Pot	How wide you enter	"High = bad" without position context
PFR	Preflop Raise	How often you raise vs limp/call	Ignoring limp-heavy pools
VPIP − PFR gap	—	Passivity index	Gap >15 = call/limp heavy
3-Bet%	3-bet frequency	Preflop aggression vs opens	Read at 500+ hands minimum
C-Bet%	Continuation bet flop	Postflop initiative after raising	Auto-bet leak or missed value
Fold to C-Bet	Fold vs flop c-bet	Flop defend tightness	Overfold exploit target
AF / AFq	Aggression factor / freq	Bet-raise vs call ratio	High AF with low WTSD = bluff-heavy
WTSD	Went to Showdown	Calling river frequency	High WTSD + low W$SD = pay-off station
W$SD	Won $ at Showdown	Showdown win rate	Low W$SD with high WTSD = calling losers
BB/100	Big blinds per 100 hands	Profit rate	Swings until 20k+ hands in one format

RiverOdds frames online 6-max reg territory roughly 22–28% VPIP / 18–22% PFR—a band, not a commandment. Live $1/$3 plays wider; PKO MTT plays different again. Compare you vs you by format, not vs a generic chart.

Sample Size Thresholds (When Stats Lie)

Stats need volume before they stabilize. Seeker Start and nlh.poker hand thresholds:

Stat	Rough minimum hands	Why
VPIP / PFR	~300	Occurs every orbit
3-Bet / Fold to 3-Bet	~1,000	Sparse opportunities
C-Bet / Fold to C-Bet	~1,500	Position and pot-type splits
WTSD / W$SD	~3,000+	Showdown spots are rare

Below threshold, treat numbers as trend hints, not repair orders. A 60-hand session where VPIP spikes to 40% might be boredom, not structural looseness.

Positional split mandatory: UTG VPIP 18% + BTN VPIP 38% can average to 28% overall—healthy. Flat overall VPIP hides positional leaks.

The WTSD Triangle (Three Stats Together)

Showdown stats mislead in isolation. Read the triangle:

WTSD	W$SD	Likely profile	Exploit / fix
High (~32%+)	Low (~48%)	Calls too wide, loses at showdown	Value bet thinner; bluff less
Low (~22%)	High (~58%+)	Folds too much; strong when showing	Bluff more; avoid thin value
High	High	Station that wins—rare; often small sample	Confirm hand count
Low	Low	Nit—folds weak and strong	Pressure preflop, barrel folds

Smart Poker Study adds: cross WTSD with VPIP. WTSD 20% with VPIP 15% differs completely from WTSD 20% with VPIP 45%. Tighter entrants reach showdown with stronger ranges.

Positional BB/100 (Profit by Seat)

Sky Matsuhashi's leak framework uses BB/100 by position as profit breakdown:

Position	Healthy online band (large sample)	Leak signal
SB	Roughly −15 BB/100	Worse than −25 → blind defense / steal leaks
BB	Roughly −30 BB/100	Worse than −45 → over-defending
BTN	Strongly positive in most pools	Flat or negative → open/steal review
EP	Moderate negative acceptable	Catastrophic → too wide or too passive

These bands prove you are not paying full blind tax every orbit—not because −30 is magic, but because deviation flags review priority.

EV Loss and AI Flags (Triage Priority)

Modern reports show EV Loss or "mistake cost" per hand—chipEV difference vs a reference line. Treat like a priority flag:

1. Sort by EV loss, not by pot size—small-pot preflop errors repeat more than coolers
2. Cluster by street—five turn probes that each cost 2bb beat one 40bb cooler for fix ROI
3. Verify one hand manually before changing global strategy—AI and trackers mis-parse multiway pots

No report replaces arithmetic on the flagged hand. If the tool says −4bb EV on a river call, rebuild pot odds yourself once.

Five-Hand Verification Drill

When a stat crosses threshold:

1. Pick one suspicious stat (e.g., fold-to-c-bet 72% in BB)
2. Filter 10 hands matching that spot
3. Review 5—predict before expanding notes
4. Write one sentence rule ("vs BTN 33% cbet on dry boards, defend check-raise with top pair+")
5. Re-check stat after 2,000 more hands

nlh.poker's leak loop matches this: stat → filter → bookmark → review decisions not results → one adjustment → revisit.

Three Common Mistakes

1. Single-stat panic — Changing entire preflop chart because 3-bet% reads 4% at 200 hands.

2. Ignoring pot type — C-bet% mixes single-raised pots, 3-bet pots, and multiway. Split filters first.

3. Result-colored review — Won pot ≠ correct; lost pot ≠ leak. Reports show frequencies; you judge line quality.

Summary

Analysis reports are diagnostic data, not autopilot. Read VPIP with PFR, WTSD with W$SD, profit with position splits. Calibrate sample size before repairs. Sort EV loss by repeatability, verify one hand with pot math, fix one stat cluster per week.