# What 36,000 nootropic ratings show

Rating: Speculative — Little primary evidence yet: mechanism, anecdote or tradition
Type: Guide
Updated: 2026-10-14
Canonical: https://heretic.bio/articles/what-36000-nootropic-ratings-actually-tell-us.html

> The Nootroflix dataset holds about 36,000 ratings from 2,802 users of the Slate Star Codex nootropics surveys. Weightlifting, sleep and meditation rank alongside prescription stimulants at the top, but the data are unblinded, uncontrolled and self-selected, so high ratings describe reported experience, not causal efficacy, and the safety model is weaker still. Its real value is flagging which claims deserve controlled testing.

> A user-reported survey is a map of what people tried and how they felt. It is not a clinical efficacy trial.

## The bottom line

The Nootroflix project collected a large, voluntary dataset of nootropic ratings. The archived Troof analysis reports 2,802 people and 36,163 ratings, while the repository snapshot used here contains 36,073 rows from 2,802 users. The 90-rating difference is small relative to the dataset, but it is worth preserving rather than silently choosing one number.

The ranking is useful as a map of user experience, not as a treatment guide. Weightlifting, sleep-seeking, meditation, and exercise rank near the top alongside prescription stimulants. That does not mean a capsule is equivalent to training or sleep, and it does not show that a substance caused the reported effect. People selected what to try, took uncontrolled doses, and reported outcomes without blinding.[1][2]

The most defensible use of this dataset is to decide which claims deserve a closer look. It can generate hypotheses and expose where community experience diverges from clinical evidence. It cannot establish efficacy, safety, or the right dose.

## What was collected

Nootroflix is a recommender system built around the Slate Star Codex nootropics surveys. Users rated substances they had already tried, and the system compared people with similar rating patterns. The rating scale ran from 0 to 10:

  * **0:** useless, or intolerable because of side effects
  * **1–4:** subtle effects, possibly placebo, but useful
  * **5–9:** strong effects that felt unlikely to be placebo
  * **10:** life-changing



The original Nootroflix subset contains 29,387 ratings from 1,981 users. Combining it with the original Slate Star Codex data produced the larger table used in the analysis.[1][2][3]

The repository provides the raw ratings, fitted summaries, and the analysis notebooks. The snapshot reviewed here is commit `b79423a61801ab3197a2466839c120220e52c2a3`, dated 2022-04-06.[2][3][4][5][6][7]

## What the adjusted ratings show

The analysis used a random-intercept multilevel model with weakly informative priors. That approach adjusts for people who tend to use the whole scale differently, but it does not turn the survey into a randomized comparison. The numbers below are model estimates from the repository summary, not observed response rates.[4][6]

Intervention or substance | Ratings | Adjusted mean (95% interval) | Estimated positive effect (95% interval) | Estimated life-changing (95% interval)  
---|---|---|---|---  
Weightlifting | 567 | 6.67 (6.46–6.89) | 96.5% (94.6–97.8%) | 12.1% (9.3–15.3%)  
Trying to get more sleep | 993 | 5.86 (5.70–6.03) | 90.8% (88.3–92.8%) | 7.45% (5.95–9.28%)  
Meditation | 429 | 5.81 (5.55–6.06) | 94.7% (92.0–96.6%) | 6.58% (4.55–9.27%)  
Adderall | 930 | 7.25 (7.08–7.42) | 95.0% (93.1–96.5%) | 12.8% (10.6–15.4%)  
Dextroamphetamine (Dexedrine) | 181 | 7.31 (6.97–7.67) | 94.3% (90.1–96.9%) | 17.8% (12.4–24.5%)  
Methylphenidate (Ritalin) | 287 | 6.31 (6.02–6.63) | 91.1% (86.7–94.4%) | 7.16% (4.87–10.4%)  
Modafinil | 1,016 | 6.25 (6.09–6.41) | 91.7% (89.1–93.8%) | 3.13% (2.33–4.13%)  
Phenibut | 544 | 6.24 (6.02–6.46) | 91.4% (88.4–93.8%) | 8.91% (6.90–11.5%)  
Cerebrolysin | 33 | 5.85 (5.07–6.59) | 88.5% (76.5–95.4%) | 10.8% (4.72–22.0%)  
BPC-157 | 30 | 4.40 (3.63–5.14) | 85.6% (71.7–93.8%) | 5.36% (1.81–13.8%)  
Semax (or NA-Semax) | 190 | 5.33 (4.97–5.68) | 85.8% (79.4–91.0%) | 4.86% (2.85–7.74%)  
Piracetam | 602 | 4.10 (3.89–4.31) | 82.2% (77.5–86.1%) | 1.15% (0.68–1.85%)  
  
Three details are easy to miss.

First, the highest-rated entries are not necessarily the best nootropics. They are the entries with the most favorable self-reports among people who chose to rate them. The model estimates are conditional on that selection process.

Second, the sample sizes matter. Cerebrolysin and BPC-157 have wide intervals, so their high ranks are much less stable than the ranks for caffeine, weightlifting, or modafinil.

Third, the estimates for different outcomes answer different questions. A high mean rating is not the same as a high probability of a clinically meaningful benefit. The article's authors also warn that a rating of 10 was probably used by some people for a very good but not literally life-changing experience.[1]

## Why the rankings are biased

The main bias is self-selection. A person who tries modafinil, Adderall, or a new peptide is not a random sample of the population. They may be unusually motivated, unusually treatment-resistant, unusually susceptible to a positive response, or simply more interested in the topic than someone who tried nothing.

The ratings were also uncontrolled and unblinded. Expectations, regression to the mean, mood, sleep, co-administered substances, and the act of paying attention to an effect can all change a rating. The analysis itself says that its numbers are better for narrowing the search space than for making a decision.[1]

The project does not solve that problem by matching users. Collaborative filtering can make recommendations more personalized, but it cannot remove selection bias, standardize doses, or identify a causal effect. It is closer to a large observational survey with a useful recommender layer.

## The safety model is weaker than the rating model

The second part of the analysis tries to estimate addiction, tolerance, side effects, and persistent side effects. Those estimates come from a separate hierarchical model that includes missing response behavior. The authors explicitly warn that this section is more uncertain than the rating section and depends more heavily on modeling assumptions.[1][7]

A few model estimates illustrate the issue:

Substance or intervention | Model estimate: stopped because of side effects | Model estimate: persistent side effects | Model estimate: addiction | Model estimate: tolerance  
---|---|---|---|---  
Weightlifting | 3.4% | 2.3% | 1.5% | 3.4%  
Caffeine | 7.4% | 10.7% | 8.8% | 30.9%  
Modafinil | 10.5% | 11.4% | 1.9% | 10.5%  
Phenibut | 4.4% | 3.5% | 4.4% | 14.2%  
Bacopa | 16.5% | 2.1% | 0.04% | 1.0%  
  
These are not adverse-event rates from a clinical trial. They are model outputs built from voluntary reports, with the confidence intervals usually becoming very wide for compounds that few people rated. A zero estimate can mean that no events were reported, that few people answered the question, or that the model was pulled toward a baseline probability. The authors warn that the hierarchical model shrinks small samples toward the average, which can make a distinctive compound look ordinary.[1][5][7]

This is especially important for compounds with strong positive ratings. A high subjective benefit score does not cancel a risk of tolerance, dependence, or adverse effects. The right comparison is benefit and harm in the same population, with a defined dose and follow-up period. This dataset does not provide that comparison.

## What it adds to Heretic

This survey is worth keeping because it covers a gap that randomized trials do not: the everyday experience of people who try many different interventions and can compare them in one rough scale. It also gives a useful reality check to claims that a popular supplement is obviously effective or that a prescription stimulant is the only meaningful cognitive enhancer.

It is not a replacement for the pages on [Overview - Cognitive Enhancement](nootropics-and-smart-drugs-overview.html), [Modafinil and Eugeroics](modafinil-and-eugeroics.html), [Racetams](racetams.html), [Nootropic Stacks and Protocols](nootropic-combinations.html), [Semax](semax-acth4-7-pgp-neurotrophic-peptide.html), [Selank](selank-tuftsin-derived-anxiolytic-peptide.html), [BPC-157 (Body Protection Compound)](bpc-157-body-protection-compound.html), [Sleep Optimization](sleep-optimization.html), or [Exercise Science and Longevity](exercise-science-and-longevity.html). Those pages should still lead with mechanisms, clinical evidence, interactions, and known risks.

The survey also supports a useful editorial rule for Heretic: community reports can identify a pattern worth testing, but they should not silently become efficacy evidence. A claim should be labeled observational, unblinded, or anecdotal when that is all the source provides.

## A cautious way to use the numbers

If the goal is to decide what to investigate first, sort the compounds by the combination of:

  1. a sufficiently large number of independent ratings;
  2. a stable estimate for the outcome that matters;
  3. a dose and use pattern that can actually be traced;
  4. a safety record that survives comparison with the claimed benefit; and
  5. support from controlled human evidence, where that evidence exists.



For a personal experiment, the survey cannot supply a protocol. Keep one variable at a time, record dose and timing, and use a pre-defined outcome rather than remembering only the days when something felt noticeable. Do not combine a high-ranked stimulant, sedative, or experimental compound just because its rating is high. The relevant Heretic guidance is in [Nootropic Stacks and Protocols](nootropic-combinations.html), but the safer starting point is usually the foundation pages for [Sleep Optimization](sleep-optimization.html) and [Exercise Science and Longevity](exercise-science-and-longevity.html).

## Verdict

The Nootroflix ratings are a useful observational snapshot and a good source of hypotheses. They are weak evidence for efficacy and even weaker evidence for safety. Their most important contribution is negative: popularity and subjective intensity do not establish that a nootropic works, that its risk is acceptable, or that the experience transfers from one person to another.

For Heretic, this belongs in the Cognitive Enhancement section as a transparent secondary analysis. It should be cited, caveated, and kept separate from primary evidence.

## Sources