The Lab

Development

Every release, logged — newest at the top.

v1.2

The Rebuilt Forecast

Current September 2026

The forecast page has been rebuilt on ground that holds: per species, per day, as a percentage above or below a normal day that month. Wind, moon phase and air pressure are gone. Not because they did not fit, but because they did not earn their place. Plus a new big-fish atlas, language in the URL, and a noticeably faster site.

Every species against its own annual average. Where a card says there is no day-to-day difference, we could not measure a reliable one, and we say so.
Every species against its own annual average. Where a card says there is no day-to-day difference, we could not measure a reliable one, and we say so.
Every species for the week in one table, with the weather above it. Most rows running flat is the point: for those species we measure no day-to-day difference of their own.
Every species for the week in one table, with the weather above it. Most rows running flat is the point: for those species we measure no day-to-day difference of their own.
Big fish per water rather than per grid cell. What counts as big is derived per species from the national length distribution, following the Gabelhouse quality-length convention from fisheries science, not a round number we picked. Specialist
Big fish per water rather than per grid cell. What counts as big is derived per species from the national length distribution, following the Gabelhouse quality-length convention from fisheries science, not a round number we picked.
The fishing year per species, corrected for how much fishing actually happens each month. A quiet month means quiet fish, not quiet anglers.
The fishing year per species, corrected for how much fishing actually happens each month. A quiet month means quiet fish, not quiet anglers.

The forecast, rebuilt

  • Per species, per day: how many percent above or below a normal day in that month. The 0-to-100 score is gone, because it implied a precision that was never measured.
  • The series is monotonic: a day the model rates higher never shows a lower percentage than one it rates lower.
  • Only species with demonstrable day-to-day skill get a daily strip. For the rest we show what does hold up: the season.

What we took out

  • Wind was by far the largest factor in the old model. It turned out to be the shape of the coastline rather than the weather. We wrote up how we found it and what it cost us.
  • Moon phase as timing, air pressure as a feeding switch, water level as height above datum: all three failed against our own data.
  • What survived carries the model: when in the year you fish, and then water temperature. In that order.

Season per species

  • A monthly index per species, corrected for how much fishing happened that month at all. Without that correction every species is a summer species, simply because August reports over four times as much as January.
  • The seasonal charts were computing from survey data rather than angler catches. Fixed.
  • Closed season and naming for eel reviewed and corrected.

Big fish, measured again

  • Big-fish analysis now works per water rather than per grid cell. A canal and a city pond in the same cell are not the same place.
  • What counts as big is derived per species from the national length distribution rather than picked as a round number. That follows the Gabelhouse quality-length convention from fisheries science, anchored against the published values where they exist.
  • Every species is re-tested each night: against shuffled data to expose patterns that are only clustering, and against itself split in half to see whether the result reproduces. Only species that pass get a claim.
  • Hundreds of thousands of length measurements from survey work added, which had been lost at import.

Language in the URL

  • Dutch sits on the plain URL, English and German under /en/ and /de/. Every page now has its own shareable link per language, instead of a language choice that existed only inside your own browser.

Under the hood

  • Three independent data checks run every night: is the data fresh, did each source actually write anything, and is any month far below its seasonal normal. A silent two-month outage is what put us onto this.
  • Critical CSS is now inlined up front, so the first paint no longer waits on a stylesheet.
  • An interrupted sign-in now lands on a page in your own language instead of an error.
  • Every public page can also be requested as plain markdown, for anyone reading the site through a language model.
v1.1

Lakes & Depth Maps

Released July 2026

Every named Dutch lake gets its own page — with all organisms, rarity, a depth map and the sharpest aerial imagery. Plus a 48-hour hourly forecast that shows, per species, when the fish bite, and a spot analysis that honestly shows how over-represented each organism is at a location.

Every named lake its own page: species, water quality and the food web of that specific water.
Every named lake its own page: species, water quality and the food web of that specific water.
Aerial imagery as a map layer — zoom right in on any bank, sharper than the standard map. Explorer
Aerial imagery as a map layer — zoom right in on any bank, sharper than the standard map.
48-hour catch outlook per species — colour shows the hourly chance, weather combined with each species' active window (day, dusk, night). Pikeperch and eel light up at night. Explorer
48-hour catch outlook per species — colour shows the hourly chance, weather combined with each species' active window (day, dusk, night). Pikeperch and eel light up at night.
In 'why this is a good spot', every organism factor shows how over-represented it is here — as a percentage of a typical water (100% = normal), corrected for how much each spot was surveyed.
In 'why this is a good spot', every organism factor shows how over-represented it is here — as a percentage of a typical water (100% = normal), corrected for how much each spot was surveyed.
Expanded, you see the strength, the trend over time and the model weight side by side — spatial and temporal signal kept cleanly apart.
Expanded, you see the strength, the trend over time and the model weight side by side — spatial and temporal signal kept cleanly apart.

Lake pages

  • A dedicated page per named lake, with map, fish species, water quality and the food web of that specific water.
  • Every recorded organism per lake — fish, aquatic plants, insects, molluscs — not just the familiar target species.
  • Rarity classification on every species and organism card, based on how often something is recorded nationally.

Depth maps & structure

  • Per-lake depth maps wherever reliable bathymetry exists, labelled 'surveyed' or 'modelled'.
  • Structure layers that reveal drop-offs, channel edges and shallows — the places fish hold.

Satellite imagery

  • The sharpest aerial imagery available as a map layer, served same-origin so it loads everywhere.

Hourly forecast

  • 48-hour hourly forecast per species (Explorer and up): a heatmap showing how promising each fish is by the hour — weather combined with the species' natural active window (day, dusk or night).
  • So you can watch pikeperch and eel peak at different hours — the temporal patterns from our moon research, now readable hour by hour.

Spot analysis

  • In 'why this is a good spot', every organism factor now shows how strongly it is over-represented here — as a percentage of a typical water (100% = normal), corrected for how much each spot was surveyed.
  • Strength, the trend over time and the model weight are shown separately, so spatial and temporal signal don't blur together. Worded neutrally: co-occurrence points to suitable habitat — more is not automatically better.

Under the hood

  • A food-web story per lake, generated from that water's own co-occurrence signal rather than boilerplate text.
  • Expanded structured data and hreflang (NL/EN/DE); all lake pages added to the sitemap.
v1.0

First Cast

Released March 2026

VisScanner went live: a data-driven fishing map for the whole of the Netherlands, built on open data and models validated on unseen data.

The best spots per species, ranked by where you catch the most — nationwide, from canal to city pond. Explorer
The best spots per species, ranked by where you catch the most — nationwide, from canal to city pond.
Why a spot is suitable: the per-species habitat score, plus the organisms that predict the habitat. Explorer
Why a spot is suitable: the per-species habitat score, plus the organisms that predict the habitat.
Trophy spots: where a species grows larger than average. Specialist
Trophy spots: where a species grows larger than average.
7-day forecast: which day — and which species — gives you the best odds. Explorer
7-day forecast: which day — and which species — gives you the best odds.
The models
Catch model gradient boosting

Predicts how promising a day is from weather, water level, wind and season. Trained across all species and the engine behind the 7-day forecast. The data itself showed water level and wind dominate — not the moon.

Habitat model gradient boosting · trained per species

Estimates per grid cell how suitable the habitat is for a species, from water quality, prey and surroundings. Each model is validated on cells it has never seen (ROC-AUC).

Spot model precomputed

Ranks the best spots per species by combining catch data and habitat scores. Computed nightly so the map stays instant.

Size model random forest · trained per species

Predicts where a species grows larger than average — the trophy spots — from length measurements and the local conditions that drive growth.

The map

  • Interactive fishing map with the best spots based on catch data and habitat models.
  • Trophy spots and catch hotspots per species.

Species

  • Species pages with season, regulations, distribution and statistics for over 250 species.

Prediction

  • 7-day fishing forecast based on weather forecasts, fed by the general catch model.

Ecosystem & water

  • Ecosystem explorer with Shannon diversity, sampling coverage and cyanobacteria risk.
  • Water quality (Rijkswaterstaat), weather (KNMI/Open-Meteo) and biodiversity (GBIF).

Platform

  • Three languages (Dutch, English, German) and subscription tiers (Free, Explorer, Specialist).