How Anonymous Are Ignition’s Tables? We Measured It

Ignition, Bovada and Bodog run anonymous tables: no screen names, and the seat labels in the hand history rotate with the button every hand. The pitch is that nobody can build a database on you. The reality is more interesting, in both directions.

The part that is not anonymous at all: your cards

The exported hand history reveals every dealt seat’s hole cards, folded hands included. We checked 2,208 hands from our own exports: 13,127 of 13,129 dealt seats showed their cards. The two exceptions were sit-outs.

This is unusual. On a normal site, the only hands you ever see are the ones that reach showdown, so any “population range” built from histories is biased toward the hands people are willing to show. On Ignition, the limping range is the hands people limped. The 3-betting range is the hands people 3-bet. Every stat on this blog uses this, and it is the reason our pool data has no showdown bias.

It also means every opponent who downloads their histories sees your folds. Not your name, your folds.

The part that is only partly anonymous: who is in the seat

The history labels seats by position (“UTG”, “Dealer”, “Small Blind”), and those rotate every hand. There is no player id. But Seat N is a physical chair, and a chair’s stack at the start of the next hand must equal its stack at the end of this one, so a chair is the same person across consecutive hands when

stack(next) = stack(prev) + net(prev)

We built that chain and scored it against ground truth: traded hand-history files that carry a per-table hashed player id for each seat (these exist; the hashes are stable within a table and nowhere else).

Two details decided whether it worked:

1. A seat that is not dealt a hand is sitting out, not gone. Dropping it splits one player into several ids.
2. “Table deposit $X” is an automatic top-up by the seated player, not a new buy-in. Folding top-ups into the net moved recall from 51% to 71% and fragmentation from 2.33 to 1.36 ids per player. Top-ups were 86% of all continuity breaks.

Scored on 14,000 traded hands: precision 98.6%, recall 71.4%, F1 82.8%. It almost never merges two people into one id, and when it fails it splits one person into about 1.36 ids, which is the safe direction to be wrong. In our ring exports it gives at least 20 observations on 61.8% of seat-hands, against a ceiling of 76.8% with perfect identity.

Why the seat changes hands so often

The ground-truth hashes also told us something about the tables themselves: 40.9% of chairs change occupant during a table’s life, and one chair had seven different people in it. Ring tables on this network are a revolving door. Even a perfect per-table id gives you a stack of short samples, not a long read on anyone.

That shapes what is worth doing with the information. A read like “this seat 3-bet twice in 40 hands” is noise. What the data supports is a population prior, refined a little by whatever the current occupant has shown.

Zone is actually anonymous

Zone Poker re-seats you at a new table every hand, so there is no table to chain within. On 33,449 Zone seat-observations the method produced 21,270 ids, about 1.57 observations each. That is not a limitation of the method; the information is not there. On Zone, every opponent is the population.

What this means in practice

For HUDs. A conventional HUD needs a persistent id. On Ignition ring it can at best reconstruct a within-table id with the method above, and the 41% churn means the samples are small. On Zone it has nothing. This is why the HUD market for these rooms is dead and why anyone playing them well plays against a model of the pool rather than against reads. We publish that model: how often the pool limps, what the leak is worth.

For your own exposure. Nobody can follow you from session to session. Within a session, a player with the history can see every hand you folded and, on ring, can tell it was you. If you play a very distinctive style, it is readable for as long as you sit.

For the “they can see my cards” worry. Hand histories are generated after the hand. Nothing in the export is available during play. The disclosure is about what people can study afterwards, which, for the population, is exactly what makes this data good.

Data & method

Dataset

Ignition ring and Zone hand histories: own exports (NL5–NL500 ring and Zone, June–July 2026) and traded NL100/NL200 files (May 2026) that include a `Player Info` line with per-table hashed ids.

Sample

14,000 traded hands as ground truth for the re-identification score; 2,208 own hands for the hole-card visibility check; 33,449 Zone seat-observations.

Period

May – July 2026

What the numbers cannot tell you

• Precision and recall are measured on ring tables of 5–6 players; shorter tables have fewer stack collisions and should score higher.
• The traded files’ hashes are per table. We have no cross-table ground truth, so we cannot test cross-table linking (and did not try).
• Churn (40.9%) is from the traded May sample, which skews toward long-running tables.

Why every seat's cards are visible

Ignition, Bovada and Bodog export hand histories that reveal every dealt seat's hole cards, including players who folded preflop. Population stats built from them have no showdown bias: a "limping range" here is the hands people actually limped, not only the ones that reached showdown.