The Expected Value of a Dodge, With the Actual Numbers

The break-even win probability for a first dodge, computed from the app's own DodgeAdvisor model, is 47.5% — not 50% — because League's LP is asymmetric (the app assumes 18 LP gained on a win against 22 lost on a loss). A second dodge only pays off below 30%, a bar ordinary variance rarely clears outside Challenger.

A lobby you're favoured to win can still be worth leaving

Should I Dodge's result panel renders a literal verdict — "Dodging is justified" or "Not worth dodging" — underneath every analysis, with a toggle for whether this would be your first or second dodge in the last 24 hours. That verdict comes from DodgeAdvisor.Evaluate, and the model behind it produces a genuinely counterintuitive result: at exactly 50% win probability, expected LP from playing is −2, not zero. At 52% — a lobby you're favoured to win — expected LP is still −1.2. A team you're more likely to win with is still, on paper, a bad bet for your LP. That's the headline this article is built around, and everything below follows from one deliberate design choice: League's LP isn't symmetric, and this model refuses to pretend it is.

The model Should I Dodge actually runs

DodgeAdvisor (src/ShouldIDodge.Core/Services/DodgeAdvisor.cs) weighs expected LP from playing against the flat cost of dodging: E = p·W − (1−p)·L, where W is LP gained on a win, L is LP lost on a loss (as a positive number), and p is the lobby's win probability. Dodging is worth it exactly when E is worse than the dodge itself costs — p·W − (1−p)·L < −c — which solves to a break-even win probability of

p* = (L − c) / (W + L)

The app's assumed values are W = 18 and L = 22. The class's own remarks are unusually direct about what these numbers are and aren't: "stated ASSUMPTIONS, not fitted or measured values" — a deliberate contrast with WinBandCalibration.SigmaByTier, the standard deviation figures behind the noise floor and the band boundaries, which come from 400,000 simulated lobbies per tier against real data from a third-party League of Legends statistics provider. Nobody measured W and L the same way. They're a plausible, stated product judgment, not a calibrated fact, and the article below treats them exactly that way rather than dressing them up as more solid than they are.

Why LP-neutral isn't 50%

W and L aren't equal because League's LP is MMR-driven, and the asymmetry is deliberate: a player whose MMR sits below their displayed rank gains less LP on a win and loses more on a loss than a symmetric model would assume — and that's exactly the situation a player is usually in when they're eyeing the dodge button in the first place, not an edge case. With W = 18 and L = 22, the point where playing is expected to neither gain nor lose LP sits at L / (W + L) = 55%, not 50%. Anything below that, including a lobby you're favoured to win, is expected to bleed LP over time — which is exactly the 52%-lobby-still-loses-LP result from the opening section, not a rounding artefact of it.

The two break-even bars

Applying p* = (L − c) / (W + L) with the app's own dodge costs — a first dodge costs 3 LP, a second costs 10 — gives the two thresholds the result panel's toggle actually switches between:

• First dodge (c = 3): 47.5%
• Second dodge (c = 10): 30.0%

A first dodge, on this model, is justified any time a lobby's win probability reads below 47.5% — a much lower bar than the naive "only dodge a coin-flip-or-worse lobby" instinct might suggest, precisely because the 55% neutral point already sits above 50%.

How often a lobby actually crosses 47.5%

Should I Dodge's per-tier standard deviations turn that 47.5% bar into a share of lobbies at each tier, the same way the noise floor article turns any win-rate gap into a share of the distribution:

Roughly one lobby in three clears the first-dodge bar at every tier shown — nowhere near a rare event. That matches what DodgeAdvisor's own remarks say plainly: with the default W = 18 / L = 22, "the first-dodge verdict now fires on roughly 31% of Platinum lobbies." This is not a tool that rarely speaks up. It's one that, on its own stated assumptions, tells roughly a third of all players in a third of their lobbies that dodging is the better play.

The second-dodge bar, 30%, is a different story entirely. Its share of lobbies is 0.012% at Gold, effectively 0% at Diamond, and 5.1% at Challenger — the one tier where this bar is even worth talking about, for the same reason Challenger's spread is nearly three times Diamond's. Below Challenger, a second dodge is essentially never LP-justified by this model: the bar is set low enough that ordinary variance almost never reaches it.

Reading the gauge against these bars

The band-boundary table puts each tier's Low/Neutral cutoff — the point below which a reading stops being called "Neutral" — at 46.20% (Iron), 47.27% (Gold), 47.62% (Master), 47.80% (Diamond), and 43.88% (Challenger). Set the 47.5% first-dodge bar next to those and something worth knowing falls out: at Iron, Gold, and Challenger, 47.5% sits above the Low cutoff, meaning a lobby the gauge still calls "Neutral" can already be past the point where dodging is the better expected-value play. At Master and Diamond it's the reverse by a hair — 47.5% falls just inside the "Low" band, so the verdict and the label agree there. Either way, the label and the LP verdict are answering different questions and can disagree by design: the band tells you how unusual a reading is for its tier, and the dodge verdict tells you whether a specific, stated LP trade-off has tipped — they were never guaranteed to move together, and near 47.5% they routinely don't. The second-dodge bar has no such ambiguity: 30% sits comfortably inside "Really Low" at every tier in the table above, so a second dodge only ever gets LP-justified alongside a verdict the gauge is already flagging as one of the most lopsided readings it produces.

What actually moves the threshold

Three things follow from the numbers above, and none of them is a restatement of the others.

The asymmetry, not the dodge fee, is what moves the bar off 50%. Holding the first-dodge cost fixed at c = 3 and W = 18, and sweeping L: L = 18 (the symmetric case, gain equals loss) gives a 41.7% break-even; L = 20 gives 44.7%; the app's actual L = 22 gives 47.5%; L = 25 gives 51.2% — past 50%, where the model would recommend dodging a coin-flip lobby. The class's own remarks call that "a real consequence of taking the MMR asymmetry seriously," stated plainly rather than smoothed over, and this article holds the same line: the 3 LP dodge fee barely matters next to how large the W/L split is assumed to be. Move the assumption a few points and the entire verdict's character changes.

A second dodge is essentially never LP-justified below Challenger. 0.012% of Gold lobbies and effectively none at Diamond clear the 30% bar — on LP alone, the escalating penalty schedule is doing real work by making the second dodge a threshold this model almost never reaches.

Challenger is the punchline. It has the highest share of lobbies clearing the first-dodge bar (41.9%, nearly 1 in 2) and is the only tier where the second-dodge bar is reachable at all (5.1%). It's also apex tier, where a dodge costs −20 LP and an MMR penalty Riot treats as equivalent to a loss — a materially larger real cost than the c = 3 / c = 10 this model assumes, and one DodgeAdvisor's own remarks already concede the model doesn't price in at all (alongside the queue lock and any tilt effect). So the tier where this verdict fires most often is also the tier whose actual dodge cost this LP-only model most understates. That isn't a contradiction in the numbers — it's the model being honest about the one tier it was never built to fully cover.

One number this article won't try to resolve

When to Dodge states the in-game penalty schedule as −5 LP for a first dodge and −15 for a second. DodgeAdvisor computes with c = 3 and c = 10. Those two figures don't match, and this article isn't going to quietly pick one and pretend the disagreement isn't there. What's presented above are the constants DodgeAdvisor actually computes with today, labelled for what they are — stated assumptions the app's verdict runs on, not a claim about what the client currently charges you for a dodge. For the in-game penalty schedule itself, use the numbers in When to Dodge, not the ones on this page.