Methodology

Run ID

452 · as of Oct 9, 2026

Weights hash

dc7b70fa956d4ded

Reference hash

cda608182dccd252

The Formula Score (FS) is the Dose score. Full label disclosure is the baseline, not a bonus — hiding amounts in a proprietary blend only ever costs points:

FS = Dose (D) − 0.3 × (100 − Transparency (T))

Dose (D) measures how much of the ideal pre a formula delivers. Every formula is scored against the same panel of proven pre-workout ingredients, built from the consensus of independent guides (Garage Gym Reviews, ATHLEAN-X, Renaissance Periodization). A panel ingredient that's missing — or hidden in a proprietary blend — scores 0, so a bare three-ingredient formula can't outscore a complete one just by leaving things out. An ingredient under 80% of its clinical dose also scores 0: at that dose it's not doing enough to matter.

actual = label amount × active yield
r = disclosed ? min(actual / clinical dose, 1) : 0      (missing = 0)
credit = r >= 80% ? r : 0                       (pixie dust = 0)
D = 100 × sum(weight × credit) / sum(panel weights)

Other ingredients (agmatine, alpha-GPC, nitrates…) are shown on the label math for the record but don't move the tier — no credit for extras, no penalty for pixie dust. Stim-free formulas are scored on the panel without caffeine and theanine, and ranked in their own list.

Creatine and beta-alanine aren't in the panel. Both work by building up in muscle with daily use over weeks (creatine stores, muscle carnosine), not from the dose you take before one workout, so they aren't required parts of a pre-workout and a formula without them loses nothing. When a label includes them we still show the dose on the label math, count them toward transparency (hiding them in a blend costs points like any active), and count beta-alanine toward the endurance sub-score. They never set real servings or cost per serving. The tingle beta-alanine causes is paresthesia, a harmless skin sensation: it tells you nothing about whether the formula works, so it earns nothing either way. Prefer none? Use the tingle-free filter.

Caffeine earns full credit at 200 mg; less earns proportional credit (it's exempt from the 80% pixie-dust rule, since a lighter dose is a stim choice) and there's no extra credit above 200 mg — more caffeine never improves a tier. Over 400 mg still triggers the safety cap. Caffeine never sets real servings, so we never suggest extra scoops just to hit a caffeine dose. For stim-free formulas caffeine doesn't affect the score at all: they're scored on the panel without caffeine (or theanine), so being stim-free is never a penalty.

Pump, focus, and endurance sub-scores average the scored ingredients in each category (weight floor 1, evidence weights strong 1 / moderate 0.6 / weak 0.3); they're shown for context and don't change the tier.

Transparency (T) is the share of total active milligrams that are individually disclosed, including non-caffeine stimulants but excluding caffeine itself (caffeine counts toward Dose, not Transparency). A formula with no active mass scores T = 0 and is flagged incomplete.

undisclosed mg = sum(max(0, blend total − all disclosed sub-amounts))
T = 100 × disclosed active mg / (disclosed active mg + undisclosed mg)

Disclosed caffeine is subtracted from its blend total but excluded from disclosed active mass. Micrograms are converted to milligrams; IU-unit lines are excluded from transparency.

Dyes and sweeteners aren't scored. They don't change what a formula does, and whether sucralose is a dealbreaker is your call — so they're shown on the label and you can filter them out (“No artificial dyes”, “No sucralose”) instead.

Some labels declare more than one serving size, such as separate 1-scoop and 2-scoop columns. We score only those declared servings and use the one with the highest Formula Score after safety and transparency caps. Ties favor the smaller serving.

The grade, label math, and caffeine amounts all use that same serving. Real servings and cost per effective serving account for it too: a tub with 40 base scoops has 20 two-scoop servings before any further effective-dose adjustment. The product page and rankings identify when a larger declared serving was graded.

Caffeine safety math uses caffeine at the graded serving too. If a two-scoop label has 200 mg caffeine in that graded serving, the cap starts from 200 mg per graded serving, not from a one-scoop column.

Salty reports how much sodium the graded serving delivers against a 300 mg reference: 150 mg scores 50, and 300 mg or more scores 100.

Salty = 100 × min(sodium / 300 mg, 1)

The 300 mg reference is roughly what a sports drink carries in a scoop's worth of water (about 0.5–0.7 g of sodium per litre, per the American College of Sports Medicine). It isn't a proven pre-workout dose: the sodium-loading studies that improved endurance used 1.3–4.4 g with a litre of fluid before long sessions in the heat.

Potassium, chloride, magnesium and calcium don't count. We show them on the label, but no human trial has found that taking them before exercise helps performance, hydration or cramping at any dose, and sweat carries far less potassium than sodium. Scoring them would also backfire: a product that adds a little potassium would score lower than one that leaves it out.

The score is hidden when a label has no sodium. Salty is separate from Dose (D) and never changes Formula Score, the tier, or awards.

Five conditions override the raw Formula Score:

• Any sketchy ingredient (e.g. DMHA) → automatic F tier, excluded from winner awards; still eligible for shame awards.
• Any ingredient (including caffeine) above its safety ceiling — caffeine > 400 mg per serving — → tier capped at C, ineligible for price-based or winner awards.
• Transparency below 50 (majority of actives hidden in blends) → tier capped at D.
• S tier means every key active is at its clinical dose. Any strong-evidence panel active below its reference dose → tier capped at A. Caffeine doesn't count here (how much you want is a stim-level choice), and neither do creatine or beta-alanine, which work by daily loading and aren't in the panel. No current panel ingredient besides caffeine is rated strong evidence, so this cap doesn't fire today; it applies if one joins the panel.
• A questionable stimulant (synephrine/bitter orange, yohimbine, hordenine) anywhere on the label → tier capped at B. They're legal, but every guide we follow says to avoid them.

Every capped product shows the exact rule that fired under “How we got this tier.”

A formula with no non-stim actives at all (a pure stimulant product) gets no tier — labelled ☕ Just A Caffeine Pill — and never appears in winner or shame awards. A stim-free label can never be a caffeine pill: if nothing on it is recognisable because it's all proprietary blend, it scores 🧚 Fairy Dust. A sketchy ingredient overrides this exception and gives an F.

Stim level is based on total caffeine, then bumped one level per additional non-caffeine stimulant (capped at Extreme). Ingredients sold for energy that we don't score, such as theobromine, never move the level, because there's no evidence they act as stimulants at label doses. Each product page spells out its own breakdown, e.g. “304 mg caffeine alone is 🚀 Heart Go Brrr; Rauwolscine bumps it up one level.”

Caffeine bands: 0 mg with no other stimulant is stim-free; 1–150 mg is low; over 150–250 mg is moderate; over 250–350 mg is high; over 350 mg is extreme. Caffeine is rounded to a whole milligram before tiering.

• 😴 Night Shift
• ☕ Coffee Tier
• ⚡ Standard Send
• 🚀 Heart Go Brrr
• ☢️ Talk To Your Cardiologist
• 🕵️ Mystery Stim

If a stimulant (caffeine or otherwise) sits inside a blend without its own listed amount, the formula is tagged 🕵️ Mystery Stim — excluded from caffeine-range filters by default and ineligible for winner awards, because its safety ceiling can't be verified.

A certification counts as active for 180 days from verification, provided its expiry date has not passed, and enters the re-verify queue 30 days before lapsing.

T3 · COA on File needs a published, per-batch certificate of analysis for the finished product from an ISO/IEC 17025-accredited lab, covering its actives. A COA from one flavor's batch counts for the whole formula, since flavors differ only in flavoring. COAs verify label accuracy; they don't screen for banned substances the way T1/T2 certifications do.

Value is judged on cost per effective serving. The serving driver uses disclosed, positive, scored non-stim ingredients with strong or moderate evidence, preferring ideal-pre panel ingredients when the label has any, and strong evidence within that pool. Stimulants, electrolytes, daily-loading actives (creatine, beta-alanine), unscored ingredients, hidden-in-blend amounts, and ingredients without a reference minimum are excluded. Amounts are measured in graded servings.

servings to clinical = max(1, eligible reference minimum / actual ratios)
base caffeine = caffeine per graded serving / graded-serving multiplier
caffeine limit = caffeine reference safetyMax, falling back to 400 mg
caffeine cap = floor((caffeine limit / base caffeine) × 100) / 100
base servings used = min(graded-serving multiplier × servings to clinical, caffeine cap)
real servings = servings per container / base servings used
effective servings multiplier = min(base servings used / graded-serving multiplier, 2)
cost per effective serving = price / servings per container × graded-serving multiplier × effective servings multiplier

Stim-free formulas have no caffeine cap. Caffeinated formulas never imply more than the caffeine reference safety maximum, or 400 mg when no reference ceiling is present, when showing real servings, explaining underdosed clinical servings, or calculating cost per effective serving. Cost also retains its maximum of 2 graded servings. When the caffeine cap binds, that serving can still be underdosed. When the clinical amount is not reachable within the caffeine limit, the product page says so and uses the capped serving count instead of printing an unsafe scoop count.

• 🍜 Broke Bro Approved: $0 through $1.50 per effective serving; value score 100.
• 👍 Fair Price: over $1.50 through $2.00 per effective serving; value score 75.
• 💸 Pricey: over $2.00 through $2.50 per effective serving; value score 50.
• 💎 Luxury Pixie Dust: over $2.50 per effective serving; value score 25.

Value bands are only for formulas that are clinically dosed: tier C or better. A tier D or F formula is marked Underdosed instead, whatever it costs, and gets the neutral value score of 50. Its “effective serving” doesn't deliver clinical doses, so a low price per serving would be meaningless, and we never call it Broke Bro Approved.

A product with ⚠️ a clinical-dose caffeine warning is excluded from every price-based award, no matter how cheap, and receives a neutral value score of 50. The warning appears when graded-serving caffeine multiplied by min(servings to clinical, 2) exceeds 400 mg, before applying the caffeine cap to servings and cost.

Automatically retrievable non-Amazon offers are polled daily, with each store's checks spread across the day. Manual-source prices must be re-entered by an administrator. Either kind is “fresh” for 14 days. The headline price never names a store, and no price is ever printed next to a store name — buy buttons are named by store and ordered cheapest-first, with a “Lowest price found” tag on whichever button(s) supply the headline number. The price-history chart plots the lowest available price per week across all tracked non-Amazon stores.

Formula Score is never adjusted for price or trust. A separate Overall score exists only for the default table sort and to break ties for “The GOAT”, which goes to the caffeinated formula with the highest tier score (a perfect 100 whenever one exists):

Overall = 0.65 × Formula Score (capped)
        + 0.2 × Trust score
        + 0.15 × Value score

An unpriced product scores a neutral 50 on value — enough to rank below any product priced at $2.00/serving or better, and above anything over $2.50.

Current published ingredient references, in their active/base units. Reference values can change after a scoring run; the reference hash above identifies the inputs used for the current grades.