Tide Is the Gate on Flemish Banks

We built a matching model that treated tide as a detail.

Wave height carried the most weight. Period and swell direction came next. Wind had a seat. Tide sat at the bottom, at five points in a hundred, because that is how a "balanced" surf model is supposed to look. Height is the sea. Tide is the clock.

On the Flemish sandbanks that ranking was wrong.

We put the model against twenty-three rated sessions on the Belgian coast. Same breaks the app already knew. The match between those ratings and the model's similarity score was noise. A junk low-tide day could still look like a close cousin of a classic high-tide day, because the swell and wind were in the window and tide was a rounding error.

What the ratings actually said

On these banks, identical wave and wind flip with the tide.

Low water: 0 or 1 star. High water: 4 or 5. The swell had not changed. The wind had not changed. The bank had. At low tide the outer crests take the energy before it stands up as a rideable face. At high tide the same pulse has water to work with.

That is not a subtle modifier. It is a gate. If the gate is shut, the rest of the forecast is a description of a sea you cannot use.

Wind direction was the second surprise. Textbook surfing wants offshore. The rated days did not. Onshore at the right tide still came back as 4 and 5 star sessions. Once the bank is working, a bit of onshore is a texture problem, not a veto. We had been giving wind direction enough credit to float a clean-direction, flat, wrong-tide day. The ratings would not.

What we changed

Tide is now the heaviest weight in the match, at 28. Wave height stays close behind. Wind direction dropped. The weights still add to 100. Nothing was deleted. The order changed.

Against the same twenty-three sessions, the match rose from noise to a fit we could stand on. That is not a claim that every North Sea beach is a high-tide break. It is a claim that a model trained on "balanced" surf logic will misread this coast.

The match is what the app uses when it compares a custom pin to an official break, and when it asks whether tomorrow looks like a day you already rated. If tide is under-weighted, those comparisons lie in a specific way: they recommend the wrong hour of a day that looks similar on paper.

What this is not

It is not a rule that high tide is always better. Mid-tide banks exist. Low-tide reefs exist. Jæren barely has a tide. The Flemish finding does not travel as a slogan.

It is also not a tide table. The table still only tells you when the water moves. The gate is how that water sits on a named bank. Two beaches a few kilometres apart can reverse the pattern because their crests are not at the same depth.

If you surf the western Belgian beaches, start with the tide window for that bank, then read swell and wind inside it. The app now matches days in that order, because the sessions we already had would not let us do it the other way.