Dash Gas Mash Probability Field

A spatial probability field for good-order likelihood under current modeled conditions, built from public OpenStreetMap data. Not affiliated with DoorDash.

Adjusts late-night weighting (public heuristic). Note: restaurants with parseable OSM opening_hours tags are filtered by the client’s current local time before they can affect scores.
Smaller = only very close merchants matter. Larger = broader influence.
Heatmap detail vs speed. Larger = faster, less detail.
DGM's core output is fixed: pGood means the probability of a good order within the next 10 minutes at this location under current modeled conditions.
Penalty using nearby parking density as a rough “other drivers” proxy. Set to 0 to disable. Note: in strip-mall areas, parking density correlates with merchant density — this can suppress spots that are actually strong. Treat non-zero values as experimental.
Blends in nearby apartment and residential anchors as structural demand support. Set to 0 to revert to merchant-only demand.
Static Census tract anchors for the default Rancho/Ontario region will be blended into residential demand when they overlap the current view.
This uses a small preprocessed public U.S. Census tract dataset. Outside the default region, DGM falls back to OSM residential anchors only.
Live weather will be fetched from Open-Meteo on refresh. If it fails, the manual rain lift slider remains the fallback.
Bounded weather proxy. With live weather enabled, DGM derives this lift from public Open-Meteo precipitation data for the current map center; otherwise the slider stays manual.
0 = prioritize short pickup. 1 = prioritize higher-ticket “ticket proxy”.
Changes how intensity maps to % (still public proxy; not trained on DoorDash data).
Higher \u03b2 boosts top clusters more aggressively. Values above 3 can create sharp visual cliffs that are artifacts of the slider, not real-world signals. Use with caution.
How many parking suggestions to output.

Quick read

Friendly summaries for the current view. These stay view-relative and descriptive.

Best parking spots to wait at
    Help: what everything means

    What the percentages mean

    • Probability range (for example, 25%–45%) — DGM shows the modeled probability of receiving a good order within the next 10 minutes. The range comes from perturbing arrival intensity by ±30% so uncertainty stays explicit.
    • Merchant share — How much of the 10-minute probability is being supported by nearby merchants.
    • Residential share — How much of the 10-minute probability is being supported by surrounding homes and apartments. This is structural context, not a live feed.
    • Relative intensity — How strong this point is compared with the current view's reference intensity.
    • Rain lift — The current modeled demand lift from the rain control, if any.

    Sliders

    • Local hour: What time of day to model. Late night boosts fast food weight.
    • Probability horizon: Fixed at 10 minutes so every spot on the map stays directly comparable.
    • Distance decay: How far restaurants count. Smaller = only very close restaurants matter.
    • Competition strength: Penalizes spots near many parking lots (rough proxy for other drivers). Set to 0 to ignore.
    • Residential demand blend: Adds nearby housing and apartment anchors as a structural demand field. Set to 0 for the pre-promotion merchant-only model.
    • Use nearby Census tract anchors: Blends a small preprocessed public Census tract dataset into the same residential demand path when the current view overlaps the default Rancho/Ontario region.
    • Rain demand lift: Adds a bounded weather uplift to the demand field. With live weather enabled, DGM derives this from public Open-Meteo precipitation data for the current map center; if the request fails, the manual slider remains the fallback.
    • Tip emphasis: Slide right to favor higher-ticket restaurants. Slide left to favor short pickup distance.
    • Grid step: Heatmap detail. Smaller = prettier but slower.
    • K spots: How many parking suggestions to show. DGM uses the coverage-diversity selector to keep suggestions spread out without a separate hard-separation control.
    Model diagram
    Notes & limitations
    • This uses publicly available OSM POIs as a proxy for merchant density and late-night demand.
    • It does not use any proprietary DoorDash dispatch logic, merchant volume, customer demand, or courier supply data.
    • Use this for strategy exploration, not guarantees.

    Basemap © OpenStreetMap contributors. Vector map via MapTiler when configured, otherwise MapLibre demo tiles.

    Search