Home / How it works
How it worksLureScore does two separate jobs: it reads the water in front of you, and it scores the day you are fishing it in. Here is what happens inside each one — and where the honest limits are.
Water flow analysis is not a filter applied to a photo. It is a measurement of motion over time, which is why the app asks for a short video rather than a still frame.
You record five to thirty seconds of the water surface, from the bank or from anything elevated. The clip does not need to be cinematic — a steady hand and the horizon in frame matter more than resolution.
The pipeline computes dense optical flow between consecutive frames: for each small region of the image, which direction it moved and how fast. The result is a motion field over the whole surface rather than a single average "current speed".
From that motion field the app looks for the signatures of hydrodynamic features — converging flows that indicate a shear line, reverse-direction patches that indicate a backflow, rotational patterns that indicate a vortex, and consistently slow regions that indicate a slow-flow pocket. Each candidate is scored on how clearly it stands out from the surrounding water.
Features are drawn back onto your own frame, and the spots are ranked so you have an order to fish in. The output is deliberately a list to work through, not a map pin — surface flow is one input among many, and the app stays inside what it can actually measure.
The index answers a different question: not where, but when. It is a 0 to 100 score for one day at one location, computed from air temperature, barometric pressure and its trend, wind speed and direction, humidity and light level.
The weighting is not arbitrary. Temperature and flow dominate on moving water, pressure trend comes next because it governs whether fish are willing to feed at all, and wind, rain and light act as modifiers that can lift or ruin an otherwise good window. Scores are recomputed as the forecast is revised through the day.
The index is deliberately comparable rather than absolute: two windows on your home river can be ranked against each other even when neither number means anything on its own.
Want the long version of how each input is weighted and why? That explainer, with worked examples, lives on our content site: reading river current and conditions, and the daily scores themselves are published at fishing conditions by state and city.
System data can tell you the day is good. Only your own record can tell you the day is good for you — that on your stretch of river, a falling barometer in October produces, and a rising one in July does not.
Each logged trip adds a row: date, water, species, conditions, bait, result. LureScore builds a per-species model from that history and shifts its recommendations toward the patterns that appear in your own data. Early on it leans on the system index; after a season it leans on you.
Most fishing logs die in the second week because they are tedious. LureScore's is built to be a sub-minute job: conditions are filled from the day's data, species identification is automatic from the photo, and a trip you did not catch on is still worth logging — blanks are the rows that tell the model what not to recommend.
Being clear about this matters more than a feature list:
LureScore