Skateboarding Dryness Engine (SDE) ©

SDE unified formula ©

The entire Skateboarding Dryness Engine expressed as a single composed equation. Each term is explained in detail in the sections below.

Three generations of scoring

The skateability score has evolved through three distinct engine versions, each building on the last. Every version produces a 0 to 100 score representing how dry and skateable a spot is right now.

Engine v1.0 Legacy

The original scoring engine shipped with Skateable 1.0 in March 2025. It used a simple weather-only model: fetch current conditions and 48 hours of precipitation history, then apply penalties and bonuses.

How it worked

The score started at 100 and was reduced by precipitation and humidity penalties. A wind bonus partially offset those penalties. The rain decay was a linear ramp: immediately after rain the penalty was 70 points, tapering to zero over 6 hours.

Limitations

• No awareness of surface type, sun exposure, or drainage quality.
• A smooth concrete park and a rough asphalt lot scored identically in the same weather.
• Community session reports were not used.
• The linear 6-hour drying curve was the same for all conditions, whether 35 degrees and sunny or 5 degrees and overcast.

Engine v1.1 Current

Version 1.1 introduced park attributes and community session data, making the score location-specific for the first time.

How it worked

The base model was the same additive approach as v1.0, but with four new modifier terms for surface type, sun exposure, drainage quality, and temperature. Community session reports also contributed: each recent dry session added +5 (up to +15), while each wet session subtracted 10 (up to -30).

Improvements over v1.0

• A modern smooth concrete park in direct sun with good drainage now scored significantly higher than a shaded, cracked asphalt lot.
• Community session reports provided real ground truth to adjust the score.
• Temperature awareness meant cold mornings correctly showed slower drying.

Remaining limitations

• The modifiers were fixed point values, not physics-derived. A "+8" for modern concrete was a hand-tuned guess, not calculated from the material's actual thermal properties.
• The 6-hour linear drying curve still ignored real evaporation conditions. A hot, dry, windy day dried a park just as slowly as a cold, humid, still night.
• Session adjustment was a simple +/- arithmetic, not statistically weighted. A single unreliable report could shift the score as much as a trustworthy one.
• No condensation or dew point awareness. A metal ramp could show "dry" at midnight even when condensation was forming on the surface.

Skateboarding Dryness Engine (SDE) v1.2

The SDE is a ground-up rewrite that replaces the additive points model with real evaporation physics. Instead of guessing how long a surface takes to dry, it calculates the actual rate of moisture leaving the surface based on thermodynamic principles used in meteorology, hydrology, and materials science.

Algorithm specification

The complete SDE pipeline expressed as formal equations. Every score calculation passes through these stages in sequence.

Stage 1: Thermodynamic Core

The core treats each skatepark surface as a boundary layer where two forces compete to evaporate moisture:

• Net Radiation (R_n) provides the energy for the latent heat of vaporization, the energy required to break the molecular bonds of liquid water and turn it into gas. Darker surfaces (low albedo) absorb more solar energy. At night, R_n drops to zero and drying relies entirely on wind.
• Vapor Pressure Deficit (VPD) measures the difference between how much moisture the air could hold (saturation pressure) and how much it actually holds. High VPD means the air is "thirsty" and can absorb moisture quickly. On a humid day, even strong sun will not dry a park fast because the air has no room for more water vapor.

The engine also detects condensation risk. When the estimated surface temperature drops below the dew point (common on metal ramps at night), a condensation penalty of 40% to 80% is applied depending on the material's thermal conductivity. This is why a metal ramp can show as slick even on a rainless night.

Materials science

Different surfaces interact with solar radiation and water in fundamentally different ways. The SDE uses measured physical properties for each material type:

Albedo determines how much solar energy a surface absorbs. Asphalt (albedo 0.10) absorbs 90% of incoming radiation, while metal (0.40) reflects nearly half of it.

Porosity determines how much water a surface traps in its pores. Aged concrete and wood hold moisture long after the surface appears dry to the eye.

Grip multiplier only matters when the surface is wet. Dry surfaces always score 100% regardless of material. When wet, grip amplifies the wetness penalty: a metal ramp (0.6) in light rain scores much lower than concrete (1.0) in the same conditions, because wet metal is genuinely dangerous. Once both dry out, both read 100%.

Stage 2: Bayesian Data Fusion

Weather models are imperfect. Microclimates, nearby buildings, tree cover, and drainage quirks can all make a specific spot wetter or drier than the forecast suggests. Community session reports provide real ground truth to correct the physics model.

Instead of simply adding or subtracting points (as v1.1 did), the SDE uses Bayesian statistical inference. The physics model is treated as the prior probability (our best guess before seeing any evidence), and each community session report is treated as new evidence.

Each report is weighted by two factors:

• Trust factor (0.1 to 1.0): How reliable this reporter has been historically.
• Time decay: Reports lose influence exponentially with a 30-minute half-life. A session logged 5 minutes ago carries far more weight than one from 2 hours ago.

The physics model always retains at least 50% of the weight, so even multiple reports cannot fully override the thermodynamic prediction. This prevents a single unreliable report from dramatically distorting the score.

Stage 3: Graduated Rain Scoring

Not all rain is equal. The SDE uses graduated rain scoring instead of a binary wet/dry switch:

• Heavy rain (1+ mm/h): score drops to 0%. The surface is genuinely soaked.
• Moderate rain (0.3-1 mm/h): score drops to a low floor (up to 10%) based on how quickly the surface will dry once rain stops.
• Light rain / spitting (<0.3 mm/h): score can stay as high as 25% if conditions are warm and windy enough to dry the surface within 30 minutes. The dashboard shows "est. dry in ~Xm" so you know exactly when to head out.

Once rain stops, the score recovers along an exponential curve whose rate depends on VPD, surface porosity, drainage quality, and whether it is day or night. The recovery estimate is shown throughout the app: on the dashboard, in widgets, and when sharing.

For spots with poor drainage, the score is hard-capped at 60% until a VPD-based "puddle timer" expires. This reflects reality: even when most of the surface is dry, standing puddles in low spots remain. Good drainage has no cap and recovers faster.

The engine also estimates a recovery time: how many minutes until the spot reaches 90% dryness. This powers the "EST. DRY IN" indicator on the score forecast screen.

Stage 4: Display Filter

The final score applies two adjustments before display:

• Grip multiplier (G): Only affects wet surfaces. When dry, every material scores 100%. Grip amplifies the wetness penalty so slippery surfaces (metal, wood) show lower scores during and after rain. A dry metal ramp and dry concrete both read 100%.
• Tolerance offset (P): Your personal preference from -15 (strict, wants fully dry) to +15 (relaxed, comfortable skating slightly damp).

What changed in practice

Here is a concrete example showing how the three engines score the same scenario: a modern concrete park, direct sun, good drainage, 2 hours after light rain, 22 degrees, 55% humidity, 12 km/h wind.

The SDE score of 78 is lower than v1.1's 92 because the physics model accounts for the fact that 55% humidity and moderate wind at 22 degrees produce a specific evaporation rate that requires more than 2 hours for full recovery. The v1.1 linear decay was too optimistic.

Score forecast timeline

Because the SDE can calculate a score for any point in time (given forecast weather data), it now produces a full score trajectory: 48 hours in the past and 48 hours into the future.

Tap the score ring on the Dashboard to see the forecast. The chart shows how the score has changed and how it is predicted to change, with colour-coded fills (green for dry, amber for drying, red for wet) and optional weather overlays for temperature, wind, rain, and humidity.

A mini preview of the forecast curve also appears inside the score ring itself, giving you an at-a-glance sense of the trend without opening the full forecast view.

Session window

If rain is forecast within 2 hours, the score drops to reflect the shrinking window for a session. The closer the rain, the larger the penalty (up to -15 points). The dashboard hint shows exactly when rain arrives: "Rain expected in ~45m. Session window closing."

This means you never start a session only to get rained out 20 minutes in. If the window is too tight, the score tells you to wait.

Latitude-aware solar model

Instead of a hardcoded 6am to 8pm day, the SDE calculates actual sunrise and sunset from each spot's latitude and the date. In a UK summer, that means drying starts at 4:30am and continues until 9:30pm - nearly 5 extra hours of solar-driven evaporation compared to winter. In Scandinavia, the difference is even larger.

Graduated rain scoring

Previous engines used a binary switch: raining = 0%, not raining = recovery curve. The SDE recognises that a 5-minute sprinkle in sunshine is not the same as a downpour:

• Heavy rain (1+ mm/h): score 0%. Surface is soaked.
• Moderate rain (0.3-1 mm/h): floor up to 10% based on how fast the surface will dry once it stops.
• Light rain / spitting (<0.3 mm/h): floor up to 25% if drying conditions are good. The dashboard shows "est. dry in ~Xm" so you know when to roll.

Microclimate learning

Weather models describe conditions over a broad area, but a specific spot might be sheltered by buildings, sit in a wind tunnel, or retain heat from surrounding asphalt. The SDE learns these quirks by comparing historical session reports with what the physics model predicted at the time.

If a spot is consistently wetter than predicted (for instance, a sheltered bowl that takes longer to dry), the engine applies a per-spot correction. This is aggregate data about the location only - no individual user data is tracked or stored.

Wind chill awareness

When wind makes it feel significantly colder than the air temperature (3C+ difference), the dashboard hint shows the feels-like temperature using the North American Wind Chill Index formula. The surface might be dry, but you will want an extra layer.

Ice, frost, and snow detection

Freezing conditions make surfaces dangerous regardless of wetness. The SDE detects freeze risk using air temperature adjusted for wind chill:

• Hard freeze (below -3C): -15 point penalty. Surface likely iced over.
• Light frost (0C to -3C): -5 to -8 point penalty. Frost likely in shaded areas. Burns off once the sun hits.
• Near freezing (0C to 2C): hint only, minimal score impact. Cold air is often very dry, so the surface can be perfectly skateable.
• Above 2C: no freeze risk.

The freeze penalties are intentionally gentle. Cold air is frequently very dry, and a light overnight frost burns off quickly in morning sun. The hints are more important than the score impact: they warn you to watch for slick patches without crying wolf. Only sustained hard freezes warrant a significant penalty, as actual ice formation makes surfaces genuinely dangerous. Recovery from a freeze requires sustained above-zero temperatures to melt the ice first, then normal evaporation to dry the meltwater. This makes freeze recovery significantly slower than rain recovery.

Environment attributes

Five optional per-spot attributes let the engine account for local conditions that weather data alone cannot capture:

All environment attributes are optional. Parks without them use sensible defaults (no cover, normal wind, level, no trees, no water) and score identically to before. Setting them fine-tunes the prediction for spots where weather data alone misses local quirks.

Scientific foundations

The SDE is grounded in peer-reviewed research spanning decades of evaporation science and surface hydrology.