Are Dawn and Dusk Really the Best Bite? A Within-Trip Test on 1,185 Taiwan Fishing Trips

"Dawn and dusk are the best bite" may be the least questioned belief in angling. It is also the hardest to test with catch records, because anglers already fish at those hours: the timing of catches is just the timing of anglers. Getting past that needs a denominator that does not rely on anglers reporting effort. We found one: the inside of each trip. Between the first catch and the last, the angler was certainly fishing.

Finding: While actually fishing, twilight bite rates are no higher than midday. The dawn-dusk peak in raw records is almost entirely when people fish. A 1.5x twilight edge or more would have been detected; the two-to-threefold edge of lore does not exist.

Analysis date 2026-09-05 · Data 2025-01-01 to 2026-09-04 · 中文版

• H1 Crepuscular peak: while fishing, bite rate with the sun within ±6° of the horizon exceeds midday (>30°). → rejected: ratio 0.80, constant-rate expectation 0.83, p = 0.69
• H2 Night: deep night (< −18°) exceeds daytime. → appears only with manually entered times and vanishes on EXIF; judged a logging artefact
• H3 Effort explains the raw peak: the raw catch timing is explained by exposure timing. → holds: the two nearly coincide
• H0: while fishing, bite rate does not depend on sun altitude. → consistent with the data (the only excursion is a 1.2x at 6–15° on the EXIF subset, with a wide interval)

Three or more catches by one angler on one day, first and last at least an hour apart, make a trip. Every 5 minutes between first and last catch we compute sun altitude at the trip's coordinates and accumulate exposure minutes in eight altitude bins: the denominator. The first and last catches define the window and are excluded; interior catches are the numerator. The quotient is bites per hour while fishing, with effort built in.

But because the window ends are catches and interior catches are removed, the per-bin ratio would not be exactly 1 even under a perfectly constant bite rate. So we redistribute each trip's interior catches uniformly within its own window 1,000 times to see what constant rate produces, and divide the observed value by it. On the charts, 1.0 means "same as any moment while fishing"; the grey band is the 95% range a constant rate also reaches.

All trips (including manually entered times): 1,185 trips, 3,355 interior catches, 5,658 exposure hours

EXIF photo timestamps only (reliable clock): 422 trips, 1,113 catches, 1,534 hours

Teal = twilight bins (sun from −6° to +6°). Vertical lines are trip-level bootstrap 95% intervals; grey bands are the constant-rate permutation 95% range.

In both charts the twilight bins sit near 1.0 and inside the grey band. Twilight over midday: all trips 0.80 (constant-rate 0.83, p = 0.69); EXIF subset 0.84 (0.71, p = 0.09). The only bin that escapes the band on the EXIF subset is low sun at 6–15°, the stretch just after sunrise and before sunset, at about 1.2x with an interval from 0.9 to 1.6: not enough to conclude anything.

The night bin deserves its own paragraph: all trips show 1.34x, significantly high, and the EXIF subset shows nothing. The likeliest explanation is people, not fish: catches made by day and logged at home in the evening, with the time field left at "now", pile into the night. We flag it as a data-quality warning, not a finding.

The two numbers in every row are nearly identical. That is the other half of this study: the timing of catches is the timing of fishing. Anglers catch a lot at dawn and dusk because they fish a lot at dawn and dusk. It is the same structure as our four earlier studies: put effort back in the denominator and the claim flattens into a line.

Freshwater, marine, lure and non-lure all agree: twilight does not exceed the constant-rate expectation. The power column answers "is this just blindness?": a true 1.5x twilight advantage would be detected 100% of the time across all trips and 78% on the EXIF subset; 2x or more is caught in every subset. The two-to-threefold edge implied by lore is clearly excluded.

Measured

• Twilight/midday bite ratio 0.80 (constant-rate 0.83, p 0.69); EXIF subset 0.84 (0.71, p 0.09).
• Power: 1.5x detected 100% (EXIF 78%); 2x 100%.
• Fishing-time share and raw-catch share match bin by bin (see table).
• Night bite rate 1.34x appears only with manually entered times.
• Sun-altitude formula agrees with the stored field to a median 0.0° (r = 0.9896) over 3,000 records.

Inferred, to be read separately

• Constant fishing intensity within the window is an assumption. Anglers rest, move spots and eat; if those pauses cluster in a period, it reads as a low bite rate.
• Time fished before the first catch is unobserved. If anglers start before dawn and land the first fish after, twilight exposure is understated and its bite rate overstated: a bias that favours twilight, which still does not exceed expectation.
• The sample is days with three or more catches, skewed toward productive days and skilled anglers; blank and one-fish days are absent.
• The "late logging" explanation of the night bin is inference, consistent with its disappearance on EXIF but not directly measured.
• Whether twilight matters for particular species is not tested here; species-level samples are too small.

Crepuscular activity peaks are real in fish ecology, but the evidence comes mainly from underwater observation of particular species (reef predators, stream salmonids), and the pattern varies by species. How recreational catch per unit effort varies across the day, a NOAA technical report states plainly, "has not been well studied". The often-quoted claim that twilight bite rates are three times midday traces to angling-club material rather than a study we could locate, and is marked unverified. What this study adds is a measurement in free-ranging behaviour at scale, with a denominator that never asks anglers for their hours. The answer is flat.

• Fishing in the Dark: the science and management of night fishing. NOAA Institutional Repository (technical report). sourceUS government technical report noting that catch per unit effort with recreational gear across the day-night cycle has not been well studied.
• Temporal links in daily activity patterns between coral reef predators and their prey. PLOS ONE (2014), PMC4213059. sourceDaily activity of reef predators and prey: some predators peak at dusk, but patterns are species-specific; a universal crepuscular peak is not supported.

1. Trips. Same angler, same day, three or more catches, first and last at least one hour apart; location is the median of the trip's catch coordinates. 1,185 trips overall, 422 on the EXIF subset.
2. Exposure. Sun altitude every 5 minutes between first and last catch by the NOAA approximation (validated against the stored field), accumulated into eight bins: <−18°, −18 to −6°, −6 to 0°, 0 to 6°, 6 to 15°, 15 to 30°, 30 to 45°, >45°.
3. Numerator. Interior catches, binned by the sun altitude at their logged time.
4. Null. Each trip's interior catches redrawn uniformly within its own window, 1,000 times, giving the null for per-bin relative rates and for the twilight/midday ratio; p is the share of null draws at or above the observed value.
5. Intervals. Trip-level bootstrap, 1,000 replicates.
6. Power. Null redraws with twilight weighted by R; the share of draws whose ratio exceeds the null 95th percentile.

Read-only analysis; no production data modified. Binned aggregates for replication are available on request.

The aggregates, figures and methods here are free to read and cite; please retain the data period, sample size and analysis date. Public pages are free to read and cite page by page. Raw records are not available for download, and bulk reuse licensing for aggregated data is still under review.

We are looking for collaborators. The within-trip design applies to any timestamped catch dataset and needs no effort reporting. If you hold catch records from another region or species, we will share the analysis code and bin definitions for a cross-regional replication. Write to support@dudefishing.net.