A 55% reading is "High" at Gold but only "Neutral" at Challenger, because Should I Dodge's five verdict bands are cut from each tier's own standard deviation, not a fixed scale — Neutral spans half a standard deviation either side of 50%, with Low/High and Really Low/Really High stepping out from there. The label, not the raw percentage, is what's comparable across tiers.
The same number, a different verdict
Should I Dodge's gauge shows a percentage, but the label under it — Neutral, High, Really High, and so on — isn't read off a fixed scale. It's read off a scale that stretches or compresses depending on which tier the lobby was scored at, because "ordinary" is defined relative to how spread out results typically are at that rank, not against some universal 50%-is-the-middle yardstick.
The homepage gauge shows this in compressed form for whichever lobby you've just scored. This article publishes the full table behind it — all eleven tiers, all four boundaries — so you can look up any reading at any rank without having to score a lobby first.
Where the boundaries come from
Should I Dodge's calibration measured a standard deviation for the win-probability output at each tier — see the noise floor behind every tier's number for the full methodology. The five verdict bands are built directly from that standard deviation: Neutral spans half a standard deviation either side of 50%, Low and High each span from there out to one and a half standard deviations, and Really Low and Really High cover everything beyond that. Converted into win-probability percentages, here are all four boundaries for every tier:
| Tier | Really low below | Low below | High above | Really high above |
|---|---|---|---|---|
| All tiers combined | 42.83 | 47.61 | 52.39 | 57.17 |
| Iron | 38.62 | 46.20 | 53.80 | 61.38 |
| Bronze | 40.89 | 46.97 | 53.03 | 59.11 |
| Silver | 41.23 | 47.08 | 52.92 | 58.77 |
| Gold | 41.83 | 47.27 | 52.73 | 58.17 |
| Platinum | 42.50 | 47.50 | 52.50 | 57.50 |
| Emerald | 43.20 | 47.73 | 52.27 | 56.80 |
| Diamond | 43.40 | 47.80 | 52.20 | 56.60 |
| Master | 42.84 | 47.62 | 52.38 | 57.16 |
| Grandmaster | 38.75 | 46.25 | 53.75 | 61.25 |
| Challenger | 31.66 | 43.88 | 56.12 | 68.34 |
What the bands actually carve up
Those boundaries aren't picked to look clean — they carve the same fixed set of probabilities out of every tier's distribution, regardless of how wide or narrow that distribution is. Neutral always covers the middle 38.3% of lobbies at a given tier, Low and High each cover the next 24.2% on their side, and Really Low and Really High each cover the outer 6.7%. A "Really High" verdict, at any tier, describes roughly a 1-in-15 lobby — genuinely rare, not just "a bit above average" — because the boundary is defined by that share of the distribution first and then translated into whatever raw percentage happens to correspond to it at that tier's own spread.
The headline contrast
Put Diamond and Challenger side by side and the gap is stark. At Diamond, the tightest tier on the board, a lobby earns a "really" label the moment it drifts past 56.60% or under 43.40% — not a huge move off the middle. At Challenger, the widest tier, a lobby has to clear 68.34% or fall under 31.66% to earn that same label — a swing more than twice as large in either direction. The two tiers are using the same five-band vocabulary to describe wildly different amounts of actual spread.
Worked example: 55%
Take a lobby that reads 55% team win probability. At Gold, the high cut sits at 52.73 and the really-high cut at 58.17 — 55% clears the first and misses the second, so it lands squarely in the High band. In standard-deviation terms that's 0.92 sigma above the Gold mean (5 points off 50%, divided by Gold's 5.45 standard deviation), which works out to roughly the 82nd percentile of Gold lobbies — a clearly one-sided reading, though not a rare one.
Run the same 55% through Challenger and it doesn't clear the high cut at all: Challenger's boundary sits at 56.12%, so 55% falls inside the Neutral band. That's only 0.41 sigma off the Challenger mean — around the 66th percentile — comfortably inside the range Challenger lobbies land in as a matter of routine. Identical number, different tier, different verdict: High at Gold, Neutral at Challenger.
You can also run the comparison the other way. To sit at the same relative rarity Challenger would need a much bigger raw number: matching Gold's 82nd-percentile reading at Challenger takes a lobby around 61%, not 55% — because Challenger's wider spread means the same number of standard deviations out corresponds to a much larger swing in raw percentage points.
Those percentile figures aren't a separate estimate layered on top of the bands — they come from the same z-score the bands themselves are cut from. Once a tier's distribution is known to be normal (see the noise floor article for why that holds at every tier), a reading's distance from 50% in standard deviations converts directly into a percentile through the standard normal cumulative distribution, the same conversion this app runs internally to decide which of the five bands a reading falls into.
Worked example: 40%
The same asymmetry runs the other direction too, and it's sharper at the bottom of the ladder than at the top. Iron's low cut sits at 46.20%; a 40% reading falls below that but still above Iron's really-low cut of 38.62%, so it lands in the plain Low band: a below-average lobby, but not an alarming one, at roughly the 9th percentile of Iron lobbies.
Run 40% through Diamond and it falls below Diamond's really-low cut of 43.40% by a wide margin, landing in the Really Low band — one of the most lopsided readings Diamond typically produces, sitting around the 1st percentile. The same raw number is an unremarkable bad game at Iron and a headline result at Diamond.
Why it works this way
None of this is the app being inconsistent — it's the app refusing to apply a single fixed scale to tiers that don't share a single fixed amount of variance. A flat cutoff (say, "below 45% is Low, everywhere") would flag routine Challenger noise as alarming and undersell how genuinely lopsided a 40% Diamond lobby actually is. See why tier affects win rates for the broader picture of why the spread differs by tier at all, and why the highest-elo data is the least trustworthy for the specific mechanism that makes Challenger's column so wide.
Every boundary above assumes a full, ten-champion lobby. Score a lobby before champion select has finished and the app widens these same cuts further still, since fewer known picks is thinner evidence, not stronger evidence — see how many picks it takes before a reading means anything for exactly how much wider, and a worked example of the same raw percentage earning a smaller label once fewer picks are locked in.
Reading a band correctly
The practical rule: treat the label, not the raw percentage, as the thing that's directly comparable across tiers. A "High" lobby at Iron and a "High" lobby at Challenger occupy the same relative position in their own tier's distribution, even though the raw percentages that earn that label are worlds apart. A raw percentage on its own tells you almost nothing about how unusual a lobby is until you know which tier's scale it's being measured against — see how the win estimate and its bands are built for the calculation underneath both, and why the app reports a band instead of a raw percentage-point delta for the reasoning that makes the label, not the number, the thing worth trusting.