PrimeBite Fishing Walleye Research & Methodology

When Do Walleye Bite? What the Research Shows

Last updated: 2026-09-14

This page explains the Walleye model in plain English. We show the drivers, the open-access sources behind them, what is research-backed, and what is a PrimeBite modeling choice. We do not publish the private numeric recipe.

The Walleye rating is an hourly activity curve. Walleye are a low-light fish, so the one thing that changes hour to hour — the light — drives the score. Dawn, dusk, and the twilight-adjacent night hours score high; bright midday scores low. The number you see for a day is the best 2-hour stretch in that curve, and the Today screen shows the full low-to-high range plus the hourly chart so you can pick your window.

How it works

PrimeBite scores Walleye hour by hour. Every hour of the day gets its own 1.0-5.0 rating, and those 24 hourly ratings form an activity curve. The headline number for a day is the best contiguous 2-hour window in that curve — the best opportunity the day offers, not a whole-day average.

Each hour's rating is a simple, readable add-up: a fixed starting point plus five drivers — low-light time, cloud cover, season, water temperature and moon phase. Low-light time and cloud cover change within the day; season, water temperature and moon phase are the same at every hour and shift the whole curve up or down. Each driver's influence is sized to match how strong the evidence behind it is: low-light timing leads, season is a strong second, and cloud cover, water temperature and moon phase are each small. There is no percentile stretching, no rounding row, no calibration row and no hidden multiplier — the rows you see in "Why this rating" are each driver's real contribution and they add up to the number exactly.

What changed in the September 2026 update. Cloud cover used to be folded inside the light driver, and hot summer water inside the season driver, so neither had a line of its own. Both are now separate rows. At the same time the drivers were re-sized so the best possible day lands exactly at the top of the scale instead of overshooting it — previously the extra had to be trimmed away, and the trimming was absorbed into the light row, which made that row read higher than the light actually deserved. Every row now shows the real amount, to two decimal places.

The low-light term reads each hour's sun position from sunrise and sunset timing. Cloud cover is scored separately and only during daylight: we read how much of the clear-sky sunlight is actually getting through at that same sun angle, so a low evening sun is never mistaken for cloud. Note this is light blocked, not the familiar "cloud cover %": thin clouds can cover much of the sky while blocking little light. Cloud never improves a night hour, and if the sky reading is missing — a far-out forecast, for instance — cloud adds nothing at all rather than guessing; the dawn and dusk peaks still render.

No sunrise and sunset, no rating. Low-light timing is the primary input, so if the app cannot work out sunrise and sunset for the day it shows no Walleye rating rather than a confident-looking low number.

A high rating means conditions look better relative to the Walleye benchmark set. It does not guarantee a catch.

Drivers we use

Research-backed means the source supports the direction of the effect. PrimeBite modeling choice means we picked the exact curve shape and how far that driver may move the rating, inside what the evidence supports. One source means exactly that — a single study, which is why that driver is held small.

1. Low-light time — the lead driver

Research-backed: walleye are more catchable in low light. Trip success is highest at dusk, then dawn, then daytime, and the same ordering shows up in harvest rate by hour and in an at-depth light curve. The evening and pre-dawn hours stay high too, because the bite peaks at the twilight crossings. The one honest exception is the deepest part of the night (around solar midnight): a lab feeding study found capture success in full darkness drops back to about the same low level as bright midday, so the curve softens there instead of staying at its peak.

Why "catchable," not "feeding": most of this evidence measures how often anglers succeed, not how much the fish eat. We say walleye are more catchable and forage efficiently in low light — not that they "feed more." The only true feeding source here is a lab study (Michels). And because Escanaba Lake — the source of the main catch study — banned night fishing, there is no angler night-catch data; the night part of the curve leans partly on that lab feeding work plus telemetry showing walleye move and forage through the night.

PrimeBite modeling choice: the exact curve shape — how fast the score falls from twilight to midday, and how far the deep-dark hours soften — and how far this driver may move the rating. Three independent adult-catch studies support the direction, which is why it is the largest driver.

Sources:

• Shaw et al. 2021 — PLOS ONE (Escanaba Lake, WI)
• Michels et al. 2025 — Environmental Biology of Fishes (lab foraging)
• Lester et al. 2004 — PDF (angling optimum)
• Vasquez 2024 — UWSP thesis (telemetry)

“The predicted odds of a successful trip was highest at dusk (0.47, 95% CI 0.33–0.63) followed by dawn (0.34, 95% CI 0.23–0.47), and then day (0.26, 95% CI 0.18–0.37). Anglers were observed to be about 10–15% more successful when fishing at dawn and dusk relative to the daytime period.”

(Shaw et al. 2021, p.8)

“Prey capture success of age-0 walleye was greatest under low light conditions, averaging 20.5% from nautical twilight to civil twilight at − 3° solar altitude (0.05–1 lx …) and declined to 10–12% as light intensity increased and in full darkness.”

(Michels et al. 2025, p.1)

“…as illuminance dropped from 32,000 to 100 lux (due to the setting of the sun), angling catch rate initially increased, reached a peak when surface illuminance was approximately 300 lux, and then fell dramatically.”

(Lester et al. 2004, p.589)

2. Cloud cover — small, daylight only, one source

Research-backed (one source): in the Escanaba Lake study, trips were more often successful when less solar radiation reached the water — a duller daytime sky went with a better bite.

Why it stays small: that is a single study, and it did not hold the sun's angle steady, so it cannot separate "cloudy" from "the sun is low." No other checked adult-catch study isolates cloud while holding sun angle and underwater visibility constant. We therefore read cloud as the angle-free share of clear-sky sunlight getting through, score it only during daylight, and cap it well below the low-light and season drivers. Cloud never improves a night hour, and an unknown sky adds exactly nothing.

PrimeBite modeling choice: the exact shape and how far cloud may move the rating, plus the smaller cap used when only a whole-day sunlight average is available instead of an hour-by-hour reading.

Source: Shaw et al. 2021 — PLOS ONE

“An angler's odds of success were 0.44 (95% CI 0.28–0.58) at the lowest observed solar radiation 3 W/m2 declining to a predicted 0.26 (95% CI 0.14–0.41) probability of success at 570 W/m2 the maximum solar radiation value.”

(Shaw et al. 2021, p.11)

3. Season — a strong second (same at every hour)

Research-backed: walleye have a strong spring post-spawn peak (roughly April–May), a smaller fall bump (September–October), a summer slump (July–August), and a moderate, region-dependent winter. This is the best-replicated pattern of any Walleye driver — it shows up across multiple lakes and multiple regions, using harvest-rate data that already corrects for how many anglers are out. Water temperature is a separate row, described next.

PrimeBite modeling choice: the exact month-by-month curve and how far season may move the rating. The app knows the local date only — not ice cover, not whether the spawn has finished locally, and not how many anglers are out — so this is a calendar pattern, not a measurement of your lake.

Sources:

• Deroba et al. 2007 — North American Journal of Fisheries Management 27:717–727 (multi-lake Wisconsin creel; harvest rate). Print/journal source — no verified open-access URL.
• Pope & Willis 1996 — Digital Commons (multi-region CPUE review)

“On weekdays, the harvest rate decreased from May to July, increased to October, and then remained stable for the rest of the season (F8, 2,298 = 14.97, P < 0.001…). … The average harvest per acre … declined from May to February (weekdays: F8, 2,294 = 23.65, P < 0.001; weekends: F8, 2,382 = 31.78, P < 0.001…).”

(Deroba et al. 2007, p.721)

“For many freshwater fishes and common sampling gears, CPUE, size and age structure, and condition are highest in the spring and fall, while growth commonly is fastest during the summer growing season.”

(Pope & Willis 1996, p.57)

4. Water temperature — small, and it only ever subtracts

Research-backed (supporting, not direct catch): walleye prefer water around the low twenties Celsius and avoid the warmest water; and in the Escanaba study, successful-trip catch rate fell as the air warmed.

Why it stays small, and why it never adds: no checked study measured lake temperature against angler catch. One measured air, the other measured where fish chose to be. Fish can also simply move deeper or into cover, so warm surface water does not prove a poor day. The season driver already carries a summer slump, so this row exists only to deepen a genuinely hot summer a little. There is no cool-water bonus: below the avoidance threshold, or with no reading at all, this row is exactly zero. The reading itself is estimated from air temperature, not measured in your lake, and the app says so wherever it is shown.

PrimeBite modeling choice: the ramp shape and the small floor this row can never go past.

Sources:

• Vasquez 2024 — UWSP thesis (thermal and optical habitat occupancy)
• Shaw et al. 2021 — PLOS ONE (successful-trip catch rate against air temperature)

5. Moon phase — the smallest nudge, one source

Research-backed: in the one study that measured it, gibbous phases were slightly better than quarter phases. That is the whole effect — and it is genuinely tiny.

Why it stays small: the effect rests on a single lake, the authors call every one of these variables "very small," and the nearest independent test found lunar phase was not a significant predictor. So moon phase can only ever be a small nudge here — never a headline. Note this is not about moonlight: the study's own authors suspect a geomagnetic or gravitational mechanism they call "unknown," and the lake banned night fishing, so brightness cannot be the cause. We never claim "a brighter moon means more feeding."

PrimeBite modeling choice: the shallow curve shape and how little this driver may move the rating. Only the gibbous phases add anything at all — new, crescent and quarter phases add exactly nothing, and an unreadable phase adds nothing rather than a guess.

Source: Shaw et al. 2021 — PLOS ONE

“The odds of a successful trip were highest during the gibbous phases (waxing gibbous 0.47, 95% CI 0.33–0.61; waning gibbous 0.45…) and lowest during the quarter phases (last quarter phase 0.28…; first quarter phase 0.34…). The observed proportion of trips successful did not vary greatly among lunar phases…”

(Shaw et al. 2021, p.13)

“…all of these variables had very small effect sizes (< 0.20; Table 1).”

(Shaw et al. 2021, p.8)

What we do not use

• Moon position / "Major" windows (dropped July 2026): the earlier Walleye model scored only the two daily windows when the moon is overhead or underfoot. A full re-read of the evidence retired them. The effect is "very small," it reverses direction between the two things the study measured (trip success went up when the moon was overhead or underfoot, but catch rate was actually highest when it was neither), the mechanism is unknown, and an independent test found it non-significant. Something that weak and unreplicated should not decide a day's rating, so moon position is gone entirely — only the tiny moon-phase nudge remains.
• Solunar minor windows: no Walleye-specific evidence supports them.
• Wind, barometric pressure, and precipitation: not used. Wind's only measured effect was "very small," pressure and precipitation have no quantitative catch evidence in our source library, and wind's real mechanism (chop dimming the light) is already captured by the light driver.
• Water clarity (honest limit): clarity genuinely shifts when walleye bite — gin-clear lakes push the bite harder toward night and twilight, stained water spreads it into the day. It matters, but there is no freshwater water-clarity feed we can read, so we cannot model it. We assume clear-to-moderate water (the walleye optimum) and disclose it rather than guess. This is our single biggest honest limit.
• A cool-water bonus: water temperature can only subtract, never add. No checked study measured lake temperature against angler catch, and the season driver already carries the seasonal signal, so a bonus would be double-counting something we never measured.

Key definitions and checks

Moon phase position vs moon illumination

The phase model uses phase position, not just percent illuminated. That matters because first quarter and last quarter can look similar in brightness but are different phases.

Source: SunCalc docs

“Moon phase value should be interpreted like this: … 0 | New Moon … 0.25 | First Quarter … 0.5 | Full Moon … 0.75 | Last Quarter …”

What PrimeBite is and is not claiming

• We are claiming these drivers have evidence behind them.
• We are not claiming any rating guarantees a catch.
• We are not publishing the private numeric tuning used to turn those drivers into a stable 1-5 score.

Limitations / Caveats

• This page explains the live Walleye model at a public-safe level. It does not publish the private numeric recipe.
• The evidence supports the direction and rough size of these drivers, but PrimeBite still calibrates the exact curve shapes and the overall scale.
• Water clarity flips when walleye bite and we have no feed for it — the biggest honest limit (see "What we do not use").
• Most of the evidence measures catchability, not feeding. A high rating means walleye are more catchable, not that they are guaranteed to be feeding.
• The deep-night part of the curve leans partly on lab feeding and telemetry, because the main catch study's lake banned night fishing — so there is no angler night-catch data.
• Because a walleye day almost always includes a dawn and a dusk, the day headline honestly clusters on the higher side. The full 1-to-5 spread lives in the hourly chart (bright midday can sit near the bottom), not across days — and it is never percentile-stretched to look more dramatic.
• Ratings are relative condition scores for the benchmark set, not a promise that fish will bite on a given trip.

Changelog

• 2026-09-14 (public in app version 1.6.6, September 23): Split cloud cover out of the light driver and hot summer water out of the season driver, so each is its own visible row. Re-sized every driver so the best possible conditions land exactly at the top of the scale instead of overshooting it — previously the overshoot was trimmed away and the trimming showed up inside the light row, making that row read higher than the light deserved. Every row is now the driver's real contribution, shown to two decimal places, and the rows add to the headline exactly. Moon phase lost its old always-on floor: only the gibbous phases add anything now. A day with no sunrise and sunset times shows no rating instead of a confident-looking low one. Saved Logbook ratings are frozen and keep the number they were saved with.
• 2026-07-03: Walleye v3 (hourly light model). Rebuilt the engine from a full re-read of the evidence. Replaced the old Major-windows-only, moon-led model with an hourly activity curve driven mainly by light, then season, then a very small moon-phase nudge. Dropped moon position / Major windows entirely (very small, sign-reversing, unreplicated). Season promoted to a strong second driver on multi-lake, multi-region harvest-rate evidence; water temperature folded into it. The day number is now the best 2-hour stretch of the curve.
• 2026-03-31: Added a working #variables anchor plus the missing Limitations / Caveats and Changelog sections.
• 2026-03-18: Published the Walleye v2 transparency page for the rebuilt live model.