How the fishing forecast is calculated

No secret sauce: for every day the model computes 24 numbers — the chance of a bite, hour by hour. Below is what those numbers are made of, drawn on charts that come from the very model behind your forecast.

Open the forecast →

What the model actually computes

The output is 24 numbers from 0 to 100 for every day × species pair on a given water body. Peak windows (the hours close to the day's maximum) and the day's overall score are derived from them. The calculation is deterministic: the same conditions always give the same answer, and you can check it.

A language model writes the wording — never the numbers. It receives values that are already computed and explains them in words; it cannot pick a different number. That is why the text and the chart in a forecast card never disagree.

1. The species' daily rhythm

The curve starts with when a species hunts. Crepuscular predators (pike, zander, perch) get two peaks at dawn and dusk, daytime cyprinids a broad midday hump, nocturnal species (catfish, burbot, eel) a hump at solar midnight. The peaks are pinned to that date's sunrise and sunset rather than to clock hours, so they sit at different times in summer and winter.

  • Crepuscular
  • Diurnal
  • Nocturnal
Daily activity rhythm: crepuscular, diurnal and nocturnal species, bite chance by hour
01234567891011121314151617181920212223
Crepuscular20%20%22%29%48%74%88%74%48%29%22%20%20%20%22%29%48%74%88%74%48%29%22%20%
Diurnal21%22%23%25%29%35%42%51%61%71%80%86%88%86%80%71%61%51%42%35%29%25%23%22%
Nocturnal88%86%80%71%61%51%42%35%29%25%23%22%21%22%23%25%29%35%42%51%61%71%80%86%
The same calm day with sunrise at 6:00 and sunset at 18:00 — only the species' daily pattern differs. The bite never drops to zero: even in the deadest hour the model keeps a base level.

2. Weather moves the level

Weather is layered on top of the daily rhythm. Pressure falling ahead of a front lifts activity, rising pressure presses it down; wind, rain and thunderstorms cut it the harder the more weather-sensitive the species is; a sharp drop in temperature over the day hits the whole day at once.

  • Pressure falling+15%
  • Strong wind10%
  • Light rain10%
  • Sharp cold front during the day10%
  • Pressure rising12%
  • Thunderstorm or downpour30%
  • Water above comfort range40%
The shift in the daily average against a calm, clear day. These are not "weights copied out of a table": each value comes from running the model twice — a calm day against a day with a single condition changed.

3. Water temperature, not air

A species' comfort range is a water range. Air swings 10–15 °C over a day; water barely moves — it is slow and lags by weeks. So the model derives the expected water temperature from the region's climate normals and adds a damped anomaly for the specific day, instead of plugging in the air temperature outside.

  • Water
  • Air
Yearly air and water temperature at mid-latitude
JFMAMJJASOND
Water0.5°0.5°0.5°3.3°9°14.1°17.5°18°15.7°11°5.4°0.5°
Air-4.8°-3.2°1.2°7.2°13.2°17.5°19.2°17.5°13.2°7.2°1.2°-3.2°
Yearly normals for a mid-latitude region. Water warms later than air and cools more slowly — in May the air is already in the pike's comfort range while the water is not. The band is that comfort range: the further the water sits from it, the more the model cuts the level.

4. Every species has its own year

Seasonal activity is computed per species, not "by the weather" in general. It is what makes the model lift burbot and drop carp in November. When that shape arrives in a given region is decided by climate: warming comes earlier in the south, later in the north, and the curve shifts with it.

  • Pike
  • Burbot
  • Carp
  • Roach
Yearly activity curve of four species by month
JFMAMJJASOND
Pike50%45%55%50%45%60%60%70%95%100%85%60%
Burbot55%50%40%30%15%10%10%15%35%90%100%70%
Carp10%10%20%45%75%95%100%95%70%40%20%10%
Roach70%80%100%85%80%50%45%45%50%55%75%70%
The home-market curves are calibrated against 25,554 real catches in Russia and Belarus from 2011 to 2026. That check overturned part of the received wisdom: peaceful cyprinids peak in late spring rather than autumn, and burbot's autumn pre-spawn feeding run is seven times stronger than under-the-ice January.

5. A river holds up better than a lake

Flowing water reacts less to pressure swings, the current mixes the water itself, and in a heatwave a river does not lose oxygen the way a still pond does. In good weather both curves coincide — the water-body correction only switches on under stress.

  • River
  • Lake
River against lake on a hot, windy day: bite chance by hour
01234567891011121314151617181920212223
River13%13%14%18%30%47%55%47%30%18%14%13%13%13%14%18%30%47%55%47%30%18%14%13%
Lake10%10%11%14%23%36%43%36%23%14%11%10%10%10%11%14%23%36%43%36%23%14%11%10%
The same hard day for pike: water above its comfort range, strong wind, rising pressure. The river loses noticeably less — on the lake, oxygen shortage is added to the heat.

Separately — the solunar day score

Alongside the species forecast, the page shows a separate solunar score from 0 to 100. It combines the number of major and minor feeding periods, proximity to a new or full moon, and whether a period overlaps dawn or dusk. This score does not change the main species bite curve: it is an additional planning signal, not another coefficient in the model. The solunar calendar explains how the periods themselves are calculated.

Build your own day

Switch the conditions and watch the curve change. Every combination is computed by the same model as the live forecast, at a water temperature comfortable for the species, so what you see is the contribution of the factors you picked.

Fish
Water
Moon
Pressure
Weather

Hourly activity

0:006:0012:0018:0023:00
Bite activity:highmediumlow

Peak bite per the model: 05:00–08:00 and 17:00–20:00

What the model does not know

  • Tides. The model has no tide data yet. The separate solunar score follows the moon's position and is not a tide forecast.
  • The real water temperature. It is derived from climate normals and weather, not measured in your water body.
  • What happens on the bank. Angling pressure, water releases, algal blooms, forage and your own skill are not part of the calculation.
  • Which species your region does not have. The forecast only shows fish that live in your region and in that type of water — what lives where is collected in the range table.
  • Fishing rules. The model rates the bite, not the legality of the trip: closed seasons, size and bag limits are set by your state or country, and none of them are part of the calculation. Check your local regulations before you go.
  • Weather beyond a week. Past the forecast horizon only the moon and the season remain — a planning aid, not a precise calculation.

The forecast is open in full, with no subscription — you can test the model on your own water right now.

Frequently asked questions

How is the fishing forecast calculated?

For each day the model builds 24 numbers — the chance of a bite by hour, from 0 to 100. It starts from the chosen species' daily rhythm (dawn and dusk, day or night), then layers on the moon phase, the atmospheric pressure trend, wind and precipitation, water temperature and the water body type. It is an ordinary formula-based calculation, not a neural network: the same conditions always produce the same number.

Does AI write the forecast numbers?

No. Every number — the hourly bite chance, the peak windows, the day's score — comes from a deterministic model. A language model only writes the prose around numbers that already exist: it explains the forecast in words but cannot change it.

Why does the model use water temperature instead of air?

Fish live in water, and a species' comfort range is a water range. Air swings 10–15 °C over a day; water hardly moves, because it is slow and lags behind the air. So water temperature is derived from the region's climate normals with a correction for the specific day's anomaly, and hourly air stays only where it genuinely is a signal — a sharp cold front.

Where do the seasonal species curves come from?

Each species has its own yearly activity curve: pike's autumn feeding run, burbot before spawning, carp in the warm months, roach in spring. The home-market curves are calibrated against 25,554 real catches in Russia and Belarus from 2011 to 2026 — that check, for instance, flipped the season for peaceful cyprinids: they peak in late spring, not autumn.

How accurate is the bite forecast?

The model rates conditions; it does not promise a bite. It knows nothing about angling pressure on the water, forage or your own skill. The closer the day, the more accurate the calculation — beyond a week the weather itself becomes a rough guide. The honest way to read it is "which hours of this day beat the others", not "how much will I catch".

Ready to find out when they're biting?

Open the bite forecast for your waterbody right now.

View the bite forecast