GOD PLAYS DICE™
Evolution & mutation
In one minute
Author’s hypothesisSoftware and workflow release; no new lab test reported.
Status reviewed 1 October 2026. This page describes proposed research by the site’s author. It is not established science, and a public repository record is not peer review.
On this page
Could mutation patterns be more predictable than they look?
What the idea proposes
Author’s hypothesis. When many populations of the same microbe face the same challenge, such as an antibiotic, they often evolve similar fixes. Ricardo’s Theory of Mutation by Natural Dominance (TMD) asks if the “winning” mutations follow patterns we could predict. It also asks how much the setting matters.
TMD pictures the possible genetic changes, or routes, as runners in a race. A route may win because its change arises often, sits in an open or exposed part of the DNA, or is likely to spread and last. Newer versions also list the setting. Early versions multiplied the first three into one rule (see Corrections).
As of 1 October 2026, the newest software version is 2.6.5, dated 5 September 2026 and posted on Zenodo on 30 September 2026. It builds on version 2.6.4. Its tools, for analysis only, include a check that a lab runs first, checks on results sent in from outside, and a log of every step that others can replay. Version 2.6.4 compares the proposed patterns with “null” explanations based on known mechanisms. A separate report is described below.
The goal is an open biology test that outside labs could repeat and check.
What it claims, and what it does not
What it claims: a reproducible workflow for testing proposed mutation-route patterns.
Ricardo’s report: the first version (30 September 2026). It is labeled “research extension version 0.1.0,” separate from the software. It says its aim is “to make predictions that can fail fairly on new data.” It checks where older evidence came from, and it corrects how earlier work used some counts (see Corrections). It proposes a new check: within the same conditions, do early and late winning changes use the same mix of routes? It also takes a new look at data that other scientists published in 2021. The data cover 111 samples of a bacterium that became resistant to the antibiotic rifampin, from three sets of experiments. That study concluded that the growing conditions shape which of these mutations show up and how well the mutants compete. The report says predictions made separately for each setting did better than one shared prediction, when checked on blocks of data held back for testing. It calls this work “exploratory, not preregistered.” It says it makes no claim that the result is statistically significant or that it shows a cause. Because it reuses data that were already published, it is not a test fixed in advance. The Test Ledger’s rule applies: “A fit to existing data is not a successful prediction.” The report says it has no new lab experiment and no independent biological check. It also says AI helped prepare its analysis, code and documents.
The report’s version 0.5.0. As of 1 October 2026, it is the newest version of the report. Zenodo dates it 30 September 2026. It was posted at 05:54 UTC on 1 October 2026. It calls itself a “methods report.” It says it is separate from the software and from version 0.1.0, which is still available. It adds a way to compare routes across several conditions chosen in advance. It uses control cells, added on purpose, to check how many cells an experiment gets back. It corrects the comparison of routes with those numbers and keeps a range of possible values, not one number. If the ranges for different conditions share no value, it says this challenges “the proposed common contrast,” under its stated assumptions. In plain words, that is the idea that the routes compare the same way in every condition. If the ranges overlap, “the data have not settled the question.” It says getting back all 100 added control cells is “useful evidence that recovery is high,” but does not show it is perfect. Its checks used software tests and simulated data, which it calls “limited software diagnostics.” It says an internal review checked the software and the math. It says AI helped with the development, the analysis and the review. It says none of the studies it reviewed supplies a matched set of TMD route and control data. It says real lab records would still have to back up its assumptions. It lists other problems that still need more work, such as samples put in the wrong group. It claims no new lab experiment and no analysis of other scientists’ raw data. It also describes a correction (see Corrections).
What it does not claim:
- Version 2.6.4 says it claims no new support from experiments beyond its own test and example files. Versions 2.3 to 2.5.4 say the same. Version 2.6.5 and report versions 0.1.0 and 0.5.0 describe no new lab experiment.
- Passing software checks does not show the pattern happens in nature. Version 2.6.5’s checks used examples, simulated results and deliberately broken files. It says passing them “means that a tested software workflow behaved as expected.” It also says its checks on returned results do not show that an experiment really took place.
- It does not establish a new law of evolution or replace natural selection.
- Some records compare the count of winners with a baseline in which every route has the same chance. But ordinary mutation bias and selection also beat that baseline. So beating it is not evidence for TMD over them. Version 2.5.4 also allows baselines built from known causes.
- No public TMD record claims to show how life began. Version 1.5 said it was “Not a claim that life started in one specific way.”
- No public TMD record reports evidence that quantum effects steer which mutations win.
- Ricardo’s book, God Plays Dice, Volume One, does not discuss TMD. It is about the origin of the universe.
The strongest challenge
Mainstream biology already studies how mutation bias (some mutations simply happen more often) shapes which fixes win. A 2026 study in Nature Communications found this effect can matter even in large groups of bacteria. How much depends on how the bias lines up with selection. The same paper notes the idea is still debated. Mainstream scientists also try to predict which routes evolution will take before an experiment. A 2019 study did this for the “wrinkly spreader,” a form that the bacterium Pseudomonas fluorescens evolves again and again in the lab. TMD uses the same form as an example. So TMD must beat these existing models at picking the winners. It must do so on new data, not the data it was built from.
TMD’s rule also overlaps standard models. “Origin-fixation” models date from 1969. They already multiply how often a new mutation appears by its chance of spreading to the whole population. TMD splits that first part into two factors. Mainstream studies already count the local DNA setting as one cause of mutation bias. And the idea that the mutations that arrive first can steer evolution is not new with TMD. A 2001 model explored it. The 2026 paper puts this view in a phrase: survival not just of the fittest, but “of the likeliest among the fittest.”
Repeated winners can have other causes, too. Natural selection may simply favor those changes most. Some routes may have been missed. In a 2015 study of the wrinkly spreader, blocking its three usual pathways revealed 13 new ones. The way an experiment collects its mutants can also tilt the count. And TMD’s rule pictures a race won by the first change to take hold. That fits best when helpful changes arise one at a time. In large groups of microbes, many changes compete at once. There, the 2026 study found, the effect of mutation bias can weaken or even grow.
Version 0.1.0 of Ricardo’s report (30 September 2026) adds a limit of its own. It shows that different causes can produce exactly the same pattern of which route wins first, and when. So these data alone cannot show which cause is at work. The report says separate measurements of mutation rates and of how well each change performs are needed. It also says this math is standard probability, and it does not claim the math is new. Version 0.5.0 says its method uses “established” statistical tools. It says the method does not identify a mechanism and is not a new statistical theorem.
What would rule it out
- An outside lab runs a test written down in advance. TMD’s picks do no better than a standard model of mutation bias plus selection.
- As of 1 October 2026, no such dated, preregistered prediction appears in the public record.
- A caution for any such test. Early records say that when the numbers do not match, a missing factor should be added or worked out from the results (see Corrections). A model that adds a factor after each miss can match old results without predicting new ones. So a fair test fixes the model in writing before the new data come in.
The first point is the Research Notebook’s own test. Ricardo’s records state the bar more generally. Version 2.6.5 says whether the models predict better is still a question for “independent, preregistered comparison against appropriate alternatives.” It says “the next scientific step” is to set out predictions and rival models before looking at new data. Earlier records describe predictions only in general terms. None names a specific experiment, the result expected from it or a date (see Corrections). Version 0.1.0 of the report (30 September 2026) says its files include a plan for a future test. Its public description does not say what the plan predicts. This page did not open the files. Version 0.5.0 says what result would challenge “the proposed common contrast” (see What it claims, and what it does not). It names no lab experiment, real data set or date for such a test. As of 1 October 2026, the Test Ledger lists TMD as “No public precommitted test.”
Corrections
- 1 October 2026 · What changed since earlier versions: Until this date, this page described version 0.1.0 of Ricardo’s report as the newest TMD record. As of 1 October 2026, the newest is the report’s version 0.5.0 (Zenodo 23075312). Zenodo dates it 30 September 2026, and it was posted at 05:54 UTC on 1 October 2026. This page now describes it and keeps version 0.1.0 as the earlier version. On 1 October 2026, Zenodo listed nine software versions and two report versions, 0.1.0 and 0.5.0. The 0.5.0 record says it keeps the files of versions 0.2.0 to 0.4.0. Lines on this page about the newest records now say when they were checked. The entries below list, by date, what earlier versions said that later ones do not. They also list two corrections by the author. As of 1 October 2026, some earlier versions use stronger words than the newest records. Examples are a press-summary heading in version 1.5 (10 November 2025) and file names in versions 1.5, 2.1 and 2.3. This page does not adopt those words.
- 30 September 2026 · What changed since earlier versions: Until this date, this page described version 2.6.4 (14 May 2026) as the current record. On 30 September 2026, version 2.6.5 of the software and version 0.1.0 of a separate report were posted on Zenodo. This page now describes both.
- 2025-10-12 to 2026-02-05 · The rule, versions 0.8 to 2.5.4 (Zenodo 17335715 to 18501085): Versions 0.8 to 2.4 state a simple rule. Multiply three things to score each route: how often the change arises, how open that part of the DNA is, and how likely the change is to spread and last. A route’s chance of winning first is its share of all the scores. Version 2.5.4 (5 February 2026) does not state the rule. Versions 2.6.4 and 2.6.5 list these factors, along with the setting. They do not spell out the rule or say whether it still holds.
- 2025-10-12 to 2026-09-30 · The name, from version 0.8 to report version 0.5.0: Versions 0.8 to 2.4 called TMD a “testable framework (pre‑theory).” They said the full name, “The Theory of Mutation by Natural Dominance,” would fit only after tests written down in advance had passed. Versions 2.3 and 2.4 said those tests should pass “across real systems and independent labs.” From version 2.5.4 (5 February 2026), the software’s title begins “Theory of Mutation by Natural Dominance.” Its description gives no reason for the change. No public TMD record reports that such tests have passed. Versions 2.6.4 and 2.6.5 still call TMD a “proposed framework.” The titles of report versions 0.1.0 and 0.5.0 leave out the word “Theory.” On this page, “Theory” is part of a formal title, not a sign that the idea has passed tests.
- 2025-10-12 to 2026-02-05 · The origin of life, versions 0.8 to 2.5.4: Version 0.8’s title named “evolution and the origin of life.” Versions 1.5 to 2.4 applied the same kind of rule to the chemistry of life’s beginnings. Versions 2.5.4 to 2.6.5 and report versions 0.1.0 and 0.5.0 do not mention the origin of life. They do not say why.
- 2025-10-12 to 2026-02-05 · Quantum effects, versions 0.8 to 2.5.4: Versions 0.8 to 2.4 listed “quantum tunneling” among their keywords. Version 1.5 said tiny movements of protons, parts of atoms, inside DNA can slightly change how often mutations happen. It said the toolkit only takes in measured numbers for such effects. Version 2.1 proposed a check involving “quantum‑assisted proton tunnelling.” Versions 2.5.4 to 2.6.5 and report versions 0.1.0 and 0.5.0 do not mention quantum effects. They do not say whether this part was dropped.
- 2025-11-26 and 2026-01-02 · Adding a factor after a miss, versions 2.1 and 2.3 (Zenodo 17718716 and 18133653): Version 2.1 covers the case where a measured bias in mutation rates does not show up in which route wins. It says mutation rates alone are then not enough, and hidden structure in the spread-and-last factor or the environment “must be added.” Version 2.3 has a worked example on the “wrinkly spreader,” built from published counts and mutation rates. Where mutation rates and winners did not match, it worked out each route’s chance to spread and last from the results (see What would rule it out).
- 2025-11-26 to 2026-02-05 · How predictions are described, versions 2.1, 2.2 and 2.5.4: Versions 2.1 (26 November 2025) and 2.2 (14 December 2025) list what they call “crisp, falsifiable predictions,” meaning predictions that could be shown wrong. One is that a measured bias in mutation rates should carry over into which route wins. Version 2.5.4 (5 February 2026) lists “sharp predictions” about the timing of the first win, for experiments that record those times. It mentions “falsifier sheets” and a ready-made text for preregistering a test, which means writing it down, with a date, before running it. None of these descriptions names a specific experiment or the result expected from it. None says a test was preregistered. Version 0.1.0 of the report (30 September 2026) shows that different causes can produce the same timing pattern (see The strongest challenge).
- 2026-05-14 · Version 2.6.4 (Zenodo 20186340): Its plain-language summary says “if the framework continues to pass independent tests.” No public TMD record reports an independent test of the idea. Version 2.6.5 does not repeat the line.
- 2026-09-30 · Report, research extension 0.1.0 (Zenodo 23068056): a correction by the author: The report says it “corrects the use of pathway-isolated WS sample totals as competing-route winner frequencies.” In plain words: wrinkly-spreader counts from tests that looked at one pathway at a time had been used as if they showed how often each route wins when routes compete. The report also tells repeated data and simulated timings apart from independent observations of living things. Its words are “duplicate panels and synthetic timings.” Its description does not say which earlier versions this affects. Earlier records include wrinkly-spreader work: the worked example in version 2.3 and a “WS branch” in version 2.6.4.
- 2026-09-30; posted on Zenodo 2026-10-01, 05:54 UTC · Report, version 0.5.0 (Zenodo 23075312): a correction by the author: The report says an audit of its archive “recovered a historical incorrect interval inversion.” In plain words: a check of the project’s saved files found an old mistake in how a range of possible values was worked out. It says the original files are kept. Its description does not say which file or version had the mistake. It also says the version before it, 0.4.0, could not handle some control results: those in which every control cell, or none, was found. Version 0.5.0 uses “established” methods to keep them.
- 30 September 2026 · What the idea proposes: The page said the tools “check that a lab’s data are ready.” The records do not say that. They describe a check that a lab runs first, and checks on results sent in from outside, so the page now says that.
Related questions in the Library
Sources
- Mutation by Natural Dominance: finite-sample bounds for recovery-corrected route comparisons (report, version 0.5.0; DOI 10.5281/zenodo.23075312) (2026-09-30; posted on Zenodo 2026-10-01, 05:54 UTC) — As of 1 October 2026, the newest TMD record and the newest version of the report: what it adds, when it says its result would challenge “the proposed common contrast,” its stated limits, the author’s correction it describes, the files it keeps from versions 0.2.0 to 0.4.0, and a title without “Theory.”
- Mutation by Natural Dominance: source-audited evidence, competing-risk diagnostics, and limits of mechanism identification (report, research extension version 0.1.0; DOI 10.5281/zenodo.23068056) (2026-09-30) — The report’s first version (0.1.0): its stated aim, the author’s correction of how wrinkly-spreader counts were used, the limit on what timing data can show (standard probability, no new math claimed), the exploratory re-analysis of published rpoB data and its caveats, the test plan in its files, and a title without “Theory.”
- Theory of Mutation by Natural Dominance (TMD) — v2.6.5: portable campaign execution and stricter result acceptance (DOI 10.5281/zenodo.22398093) (2026-09-05; posted on Zenodo 2026-09-30) — As of 1 October 2026, the newest software version: what it adds, its scope, the factors it lists, “proposed framework,” and its general wording on testing.
- Theory of Mutation by Natural Dominance (TMD) — v2.6.4: closed-loop external-validation runtime + DAG campaign executor + audit ledger (DOI 10.5281/zenodo.20186340) (2026-05-14) — What the software release does and its scope note; its comparison with “null” explanations based on known mechanisms; the wrinkly-spreader branch, the factors it lists, “proposed framework,” and the “continues to pass” line.
- Theory of Mutation by Natural Dominance (TMD) — v2.5.4: ETLR + gamma‑race time + information‑geometry decomposition (DOI 10.5281/zenodo.18501085) (2026-02-05) — The title change; the rule no longer stated; the “sharp predictions” about timing, “falsifier sheets” and a preregistration text; uniform and mechanism-based baselines.
- Mutation by Natural Dominance (TMD): analysis-only toolkit v2.4 — clean evidence & entropy atlas, exact Bayesian evidence, and time-to-first-winner extensions (DOI 10.5281/zenodo.18166188) (2026-01-06) — The three-factor rule and the race picture; “pre‑theory” and the bar for the full name; the origin-of-life rule; the equal-odds baseline; the “quantum tunneling” keyword.
- Mutation by Natural Dominance (TMD): analysis‑only toolkit v2.3 — real‑system readiness, uncertainty, and multi‑route coherence (with WS pilot case study) (DOI 10.5281/zenodo.18133653) (2026-01-02) — The wrinkly-spreader worked example, which worked out a missing factor from the results; the bar for the full name.
- Mutation by Natural Dominance (TMD): analysis‑only toolkit v2.2 — dominance geometry, path entropy & theory‑status criteria (with G3 information geometry) (DOI 10.5281/zenodo.17934374) (2025-12-14) — The “crisp, falsifiable predictions” wording; the origin-of-life rule.
- Mutation by Natural Dominance (TMD): analysis‑only toolkit v2.1 — dominance geometry, path entropy & theory‑status criteria (DOI 10.5281/zenodo.17718716) (2025-11-26) — The “crisp, falsifiable predictions” wording; adding a factor after a mismatch; the quantum check.
- Mutation by Natural Dominance (TMD): analysis‑only toolkit v1.5 — μ·A·F first‑fixation race, dominance ridge, and origins bootstrap (DOI 10.5281/zenodo.17574851) (2025-11-10) — The rule written as a race; the note on proton movements; “Not a claim that life started in one specific way”; the press-summary heading.
- Mutation by Natural Dominance (TMD): a testable framework for evolution and the origin of life — software & data (v0.8) (DOI 10.5281/zenodo.17335715) (2025-10-12) — The first title, the first statement of the rule and the “pre‑theory” status.
- The Research Desk, officialgodplaysdice.com (research snapshot 20 September 2026) (2026-09-20) — The field card, its question and the site’s own calibrated wording.
- God Plays Dice, Volume One (Ricardo Maldonado), title page, Introduction and Appendix (2026) — The book is about the origin of the universe; its list of related side projects does not include TMD, and the book does not discuss TMD; the word theory as “a high honor.”
- J. N. Barber and A. Couce, Mutation-biased adaptation is consequential even in large bacterial populations, Nature Communications 17, 7357 (2026-06-10) — Challenge: mutation bias can shape which repeated routes win, even in large populations; the idea is still debated; local DNA context as one cause of mutation bias; many changes compete at once in large populations; “the likeliest among the fittest.”
- P. A. Lind, E. Libby, J. Herzog and P. B. Rainey, Predicting mutational routes to new adaptive phenotypes, eLife 8, e38822 (2019-01-08) — Mainstream prediction of evolutionary routes for the wrinkly spreader; mutation counts differ with and without selection.
- P. A. Lind, A. D. Farr and P. B. Rainey, Experimental evolution reveals hidden diversity in evolutionary pathways, eLife 4, e07074 (2015-03-25) — Hidden routes: blocking the wrinkly spreader’s three usual pathways revealed 13 new ones.
- D. M. McCandlish and A. Stoltzfus, Modeling evolution using the probability of fixation: history and implications, The Quarterly Review of Biology 89, 225–252 (free abstract on PubMed) (2014-09) — “Origin-fixation” models, first introduced in 1969, multiply the rate at which new mutations arise by their chance of fixation; they describe mutation-limited evolution.
- L. Y. Yampolsky and A. Stoltzfus, Bias in the introduction of variation as an orienting factor in evolution, Evolution & Development 3, 73–83 (free abstract on PubMed) (2001) — An early model in which bias in which mutations arise can strongly influence the course of evolution.
- J. D. Leehan and W. L. Nicholson, The Spectrum of Spontaneous Rifampin Resistance Mutations in the Bacillus subtilis rpoB Gene Depends on the Growth Environment, Applied and Environmental Microbiology 87, e01237-21 (free abstract on PubMed) (2021-10-28) — The published data that report version 0.1.0 (30 September 2026) re-analyzes, and the original authors’ conclusion that the growth setting shapes these mutations and the mutants’ fitness.
See every committed prediction in one place: the Test Ledger. Back to the Research Desk. See a mistake? Report a problem; corrections are made openly.