Educational worksheet. A practical data-quality checklist for verifying Baccarat shoe records before analysing results, including missing ties and chart resets.
First establish what a record represents
A baccarat scorecard may contain a row for every dealt round, just the non-tied rounds, or a graphical summary whose cells do not correspond one-to-one with hands. Those formats should not be merged as if they were identical. The Bead Plate commonly records every completed result chronologically; Big Road ties are generally annotations to the most recent Banker or Player win. Before calculating anything, label each field and decide whether the unit is “completed round,” “decided round” or “displayed circle.” Otherwise a correct count can be attached to the wrong denominator.
The minimum data dictionary
For an educational log, record a unique shoe identifier, a sequential hand number, an outcome from the set {Banker, Player, Tie}, the relevant rules/version and whether the sequence is complete. A start and end marker can also clarify where the record begins and stops. Avoid recording personal data or identifiable players. If the display shows additional markers such as natural or pair outcomes, keep those in separate fields rather than replacing the primary outcome; a Banker win remains a Banker win regardless of whether it was a natural.
Worked example: the hidden tie problem
Suppose an illustrative log shows B, B, T, P, T, B. There are six completed hands: three Banker wins, one Player win and two Ties. If someone removes ties and publishes the shorter series B, B, P, B, they have described four decided hands, not six completed hands. The proportion of Banker wins across all completed hands is 3/6 = 50%. Across decided hands it is 3/4 = 75%. Both percentages describe the invented log accurately, but only if the denominator is labelled. Neither estimates future win probability from such a tiny sample.
Shoe boundaries are a separate source of errors
A fresh shuffle or new shoe resets the physical card composition and frequently resets the display. If a screen capture begins halfway through a shoe, its first mark is not necessarily the first actual hand. Two screenshots separated by a reset cannot safely be concatenated into an apparent continuous streak. Record each boundary and mark an unknown number of missing hands rather than filling the gap with guessed results. The cut card, shuffle rules and dealing procedures determine when a shoe ends; they do not create an obligation for the next result to match an old pattern.
Five validation checks
Check 1: all hand IDs are unique and ordered. Check 2: outcome categories are defined and any illegible mark is entered as “unknown,” not silently recoded. Check 3: Banker + Player + Tie + unknown counts equal the number of logged rounds. Check 4: calculated percentages state whether ties are included. Check 5: every known shoe boundary is explicit. If uncertain, report that uncertainty before comparing two logs or producing a graph.
What clean data still cannot do
A perfectly transcribed history can prove that a sequence occurred and help test whether a theoretical model describes a large dataset. It does not reveal a guaranteed profitable bet or make a derived road predictive. Small samples and repeatedly selected “interesting” streaks are especially prone to misleading stories. Keep scorecard audit separate from interpreting the underlying chance process.
Related lessons
Bead Plate and Big Road fundamentals · Why streaks are not predictions
Sources and methodology
Wizard of Odds — scoreboards; Wizard of Odds — Baccarat rules. Worked examples are hypothetical, not recorded casino results or predictive systems. Figures are educational estimates for stated rules, not live odds.
For adults in jurisdictions where gambling is legal. Learning the rules does not reduce the financial risks of gambling; stop if it causes harm.