I Tracked 350 Brawl RNG Pulls to Map the Actual Drop Distribution

The Brawl RNG community drop rate estimates come from Reddit posts and Discord threads, most of which cite each other in a circle. I wanted to know what the actual pull distribution looks like from a clean data set.So I ran 350 pulls, logged every outcome, and here's what I found.## MethodI ran pulls in 50-pull batches across seven sessions, recording:- Pull number within session- Brawler rarity tier- Whether any boost was active- Time of day (to check for any time-based variance)I excluded boosted pulls from rate analysis and kept only baseline sessions (233 qualifying pulls after exclusions).## Tier distribution: what I got vs. what's commonly posted*Common tier (Tier 1/2):* 71.2% of my pulls. Community estimates range from 65-75%. My data is within that range.Rare tier (Tier 3): 19.8%. Community estimates typically say 15-25%. Also consistent.Epic tier (Tier 4): 7.3%. This is where it gets interesting. Most community posts put Epic at 5-8%, so my number is on the high end but within the range.Legendary tier (Tier 5): 1.7%. Community estimates range from 0.5-2%. My data skews toward the high end.The distribution isn't wildly different from what people say — but the within-tier variance is more interesting than the tier percentages.## Pull clustering: the thing nobody documentsHere's what I noticed that I haven't seen discussed anywhere: Legendary pulls appear to cluster.In my 233 qualifying pulls, I got 4 Legendary outcomes. Three of them occurred within a 40-pull window in session 5. The fourth was in session 2.This is a small sample — 4 events don't establish a statistically significant pattern. But it's consistent with what some players anecdotally report ("went 80 pulls with nothing then hit twice in ten pulls").The question is whether this is real clustering or just randomness with memorable high points. I don't have a definitive answer. But if you're modeling expected pulls for a specific Legendary brawler, the assumption of independent uniform probability may underestimate variance.## Pity mechanics: are they real?Some players claim there's a soft pity system in Brawl RNG — that after a long dry streak, rates increase. I tested this by tracking pull count since last rare+ outcome versus next outcome timing.My data doesn't show a clear pity signal. The intervals between rare+ pulls in my sessions were roughly Poisson-distributed (exponentially distributed inter-event times), which is what you'd expect from a simple random process with no pity modification.This doesn't prove there's no pity. My sample is too small to detect a weak effect. But there's no evidence of strong pity in my data.## Specific brawler distribution within tiersThis is where I had to give up due to sample size. To get meaningful rate estimates for specific brawlers within each tier (not just the tier frequency), you'd need thousands of pulls per brawler rarity.I observed 4 distinct Legendary brawlers across my 4 Legendary pulls. With those numbers, I can't say anything about which Legendary is more or less likely than others.If the within-tier distribution is uniform (each brawler equally likely within tier), then your per-brawler rate is just the tier rate divided by the number of brawlers in that tier. If it's non-uniform, you'd need a much larger sample to detect it.## What this actually tells youFor practical farming purposes:1. The community tier estimates are roughly right — you're not going to get Legendaries at dramatically different rates than the 0.5-2% range.2. The variance is higher than a simple probability model suggests. If you're trying to get a specific Legendary, plan for the possibility that you'll hit your "expected" pull count and still miss — the tail probability on a geometric distribution is significant.3. Boosts appear to matter. My boosted sessions showed higher rare+ rates than baseline, which is expected, but the magnitude was larger than I anticipated. I'll do a more systematic boost analysis in a follow-up.4. Time of day didn't show clear variance in my data. I can't rule it out with this sample, but I'd be skeptical of "best time to pull" claims.---If you're tracking your own Brawl RNG data, comparing notes would be useful — the main limitation here is sample size, and pooled community data would help a lot.