Scoring model v0.3
A transparent scoreboard for messy nutrition evidence.
The index summarizes where current research leans, weighted by study quality, relevance, recency, and disagreement. It is deterministic and auditable; it is not medical advice.
This project started from the very real question: "Is this even good for you today?"
The Short Version
Each food gets a -100 to +100 evidence-sentiment score showing whether current research leans favorable, unfavorable, or mixed, weighted by study quality and disagreement.
The score is deterministic and auditable: same evidence in, same score out. It does not decide what you personally should eat.
Score + Controversy
Score
-100 to +100
Direction of evidence sentiment. Positive means the ledger leans favorable; negative means it leans unfavorable; values near zero are mixed.
It tracks how one-sidedly the research leans, not absolute danger: two foods can sit close together because their evidence points the same way, not because they carry the same real-world risk.
Market analogy: two companies can both have a bearish analyst consensus without having the same bankruptcy risk. The rating captures direction, not total magnitude of harm.
Controversy
0 to 100
Disagreement in the ledger. A low number reads as settled; a high number reads as contested, even when the headline score is positive or negative.
Controversy blends weighted conflict, study-count conflict, and uncertainty share. It helps separate "positive and calm" from "positive but still litigated."
How The Ledger Model Works
The model keeps three ledgers:
- Benefit evidence
- Harm evidence
- Uncertainty drag
Favorable studies add to the benefit ledger. Unfavorable studies add to the harm ledger. Mixed or indirect studies add drag, which keeps the model from acting too certain when the evidence desk is still arguing.
What Goes Into Evidence Weight
A single viral headline study cannot move the score much. A large, directly relevant meta-analysis can. Each study is weighted before it enters the ledger.
Direction
Does the study point favorable, unfavorable, or mixed for the food?
Quality
What kind of evidence is it? Guidelines and meta-analyses carry more weight than animal studies or expert commentary.
Relevance
Is the study actually about this food, or only loosely related through a nutrient, dietary pattern, or subgroup?
Population fit
Does the study population generalize to ordinary consumers, or is it limited to a specific group?
Effect strength
Is the signal large and clinically meaningful, or small and endpoint-specific?
Independence
Does the source appear independent, or is there industry or advocacy risk?
Recency
Newer evidence matters more, but older evidence does not disappear.
Uncertainty
Mixed, indirect, or endpoint-specific evidence adds drag instead of pushing the score strongly up or down.
Why Scores Change
Nutrition consensus is not static. As new studies enter the evidence ledger, scores may rise, fall, or become more controversial.
Some foods remain stable because the evidence points in a consistent direction. Others shift as stronger studies accumulate, older work fades, or the ledger becomes more mixed. The goal is to track where the evidence currently points, not freeze nutrition science forever.
Confidence And Controversy
Confidence reflects how much usable evidence the model has, how strong the evidence is, and whether the direction is coherent. A high-confidence score can still be wrong; it simply means the model has more signal to work with.
Controversy rises when favorable and unfavorable evidence both have meaningful weight, or when uncertainty drag is high.
What This Is Not
This is not medical advice, a dietary recommendation, or an individualized risk assessment. It is a research-sentiment model built from linked studies and original summaries.
Think the nutrition desk missed something? Send a study for review.
Did we miss a study?Opens an email draft to [email protected].
About the author
Why this exists
The author is a finance professional with backgrounds in engineering, quantitative finance, and capital markets. Throughout his career, he has built models, decision frameworks, and analytical systems designed to help make sense of complex and often conflicting information.
The idea behind this methodology began as a simple question: nutrition advice seems to change constantly, but does the underlying evidence really change that quickly? After years of reading nutrition research and watching public opinion swing from one headline to the next, the author became interested in whether nutrition consensus could be tracked the same way markets track changing expectations over time.
AI tools assist with literature review, summarization, and data organization. Methodology, scoring design, and editorial decisions remain human-directed.
The goal is not to provide medical advice or definitive answers. The goal is to make it easier to understand where nutrition consensus currently stands, how confident that consensus is, and how it changes over time.
Support the Project
Nutrition Consensus is independently maintained.
If you find the site useful and would like to support future improvements, you can buy me a coffee.
Support helps cover:
- AI-assisted literature review
- Hosting and infrastructure
- Domain registration
- Data collection and maintenance
- New food additions and ongoing research updates