IEGM Institute for Emerging Gaming Markets

Committee records

Committee records

These are the records of the committees that draft the Institute's standards. Each is published in full and unedited: every proposal, every objection, every vote, and the reason given for each vote. Nothing is summarized, nothing is removed, and no record is regenerated after publication.

New to this? How the standards are made explains who sits on these committees and how a proposal becomes an amendment. It is one short section and it is worth reading before the vote tables.

What these records are not

1 of these 6 records is hash-chained and publishes its head hash; the rest are not and carry no tamper evidence. A head hash proves that no section was altered after sealing. It does not prove that a deliberation happened as described, and no record here is reproducible or replayable.

Check a chain yourself

verify-minutes.py is the verifier written by the session that produces these records. It was copied here without modification and the build refuses unless it matches their file byte for byte, which is a check the Institute runs on itself and not one you can repeat from this page. It uses only the Python standard library and takes the path of a raw record. Run it against the raw record of a chained meeting to recompute the chain and compare the head hash yourself. Against an unchained record it reports that there is no chain, which is the honest answer rather than a pass.

sha256 ddaa0e0665ccde607c628aa3948828638c001a221e723f4a5b37b2a3255972e2

That digest confirms your download matches what we served. It is not evidence that the tool is sound, because we computed it and we publish it.

How this page is built

The description above each record is fixed before that record exists, so that no framing can be written to fit a result after seeing one. Where a description could not be fixed before the record existed, none is published.

A set that was pre-registered and is not here. Before the skill-based gaming committee ran a batch of controlled sessions in late September, a description of each arm was written and fixed, including which of them was a demonstration rather than a measurement. That batch did not produce a publishable set. As of November 2, 2026, the run intended to demonstrate the full process produced no record at all, and the three records that would support a comparison on one document family are held for a rendering defect. The baseline covers three document families at three replicates each; the treatment arm covers one family. A comparison on that family would rest on three pairs. There is none here. The descriptions remain fixed and unused rather than rewritten to suit what came out, which is the whole point of fixing them in advance.

The build reads each record and refuses to publish a description the record does not support. It re-counts the vote tables, the vote rows and the disclosure, and it parses every seat and the model it reasoned on out of the record itself. If any one of them disagrees with what the page says, the build stops rather than publishes. The same check catches the subtler case: the disclosure in one record is accurate only because it refers to members and the chair is not a member, so the build refuses any wording that widens that word without the record to support it.

The Institute's style rules do not apply to these files. A record is published exactly as the committee produced it, and the check that keeps typographic dashes out of the Institute's own prose is deliberately scoped away from this directory, because a record edited to satisfy a style rule is no longer the record.

Nothing is smoothed. One of these meetings stopped partway through on a network timeout and was resumed and finished later the same morning; that is stated on the record rather than left for someone to find in a log.

A check you can run on us, and it needs nothing from us. Take the published standard and the published record, both in full on this site. Find any passage of twelve consecutive words that appears in both. Then discard any passage that also appears in the committee's mission or in a seat's instructions, which are published with the record. What is left is text the committee reproduced from the document itself. A seat cannot quote twelve consecutive words of a section it was never given.

On the September 9 record this leaves dozens of passages, all but one of them from the second of the two documents, which is the one that gets truncated first when a committee is given more than it can read. One to try: "enforce every limit, exclusion, and restriction required by these Standards at the point of order entry". We are not publishing a count. Two implementations of this check agreed on the finding and disagreed on the number, and the reasons are instructive: how you split text into words, whether you merge overlapping passages, and what you do with characters like the non-breaking hyphen all change the total. The same corpus under the same twelve-word rule yields anywhere from 87 to 101 matches depending on those choices, and never anything close to zero. A figure a careful reader could fail to reproduce while following our instructions exactly would be worse than no figure. Count them yourself under whatever rule you prefer.

The limit of that check, which is the only source left. The committee could in principle have reproduced the text from having memorized a document published weeks earlier rather than from being given it. That is not excluded here, and at twelve-word fidelity on text this specialized it is unlikely. We also checked the research results that were returned to the seats during the meeting, in case the standard had reached them that way instead, and the passages survive that exclusion. Those research results are not published, so that particular exclusion is the one part of this check you cannot reproduce, and you have only our word for it. To say how much that is: for this meeting the six seats were given thirty research results between them, five each, and not one of the addresses in them appears anywhere in the record.

One more thing about characters, because it has changed an answer three times. The records are published exactly as produced, which means they contain the punctuation the committee software emitted, including non-breaking hyphens inside document names. Whatever tool you use has to decide whether to treat those as ordinary hyphens before it searches. That decision is a rule like any other: it broke our own first attempt at counting provisions, it is why the Institute's typographic rules are switched off for this directory, and it accounts for a one-window difference between two careful implementations of the check above.

A weaker second check, published because its weakness is instructive. Instead of matching text you can count provisions: take each provision cited in the record and ask whether it is named anywhere the committee was pointed at. On the September 9 record, four of the fifteen provisions cited with a document prefix appear in neither the mission nor any seat's instructions. Four is a low number, and it reflects this committee's mission rather than this record: that mission names Parts 300 through 500, which makes eleven of the fifteen citations unfalsifiable by this method. A committee with a terser mission would score far better on an identical record, which is why the text check above is the one to rely on.

Three things will trip you up on the provision count, and they tripped us up first. A reference to a Part covers every section inside it, so Part 500 reaches 501 through 505. The document name appears with an ordinary hyphen and with a non-breaking hyphen, because records are published exactly as produced and nothing is tidied, so accept both. And a search for a compound reference like section 207(a)(2) will fail even though the provision is there, because these documents nest a paragraph and then a subparagraph and never write the joined form: walk the section, then the paragraph, then the subparagraph. If a reference does not resolve, the honest conclusion is that it was not found as written, never that it was invented.

What these checks cannot do. They verify this page against the record. They cannot verify the record against what actually happened in the meeting, and nothing on this page should be read as claiming otherwise. A record can be internally consistent, pass every check here, and still describe the meeting inaccurately. Catching that requires measuring the run itself, which is the committee's work rather than this page's, and it is the reason records are published only when the committee that produced them is willing to stand behind what the record says.