Choose an index
Filter the score
Highest scores
More than 15 score 100: they are ordered by the score before the 100 cap, then by dwellings.
Compose it
Each ingredient is one number per building: the homes, the grid squares holding a kind of place within Report 4's walk (full weight to 10 minutes, about 830 m; half at 15; none from 20), or a map layer of its grid square. Set its weight; flip the arrow to count it as a need (less of it → higher score).
How this map is made
Both modes
- Buildings. Every Tallinn building in ETAK/EHR (; planned, ruined, underground and foundation-only buildings left out), walked from a point inside its footprint. Home buildings are those with dwellings: dwellings in buildings, from the EHR building register (open-data extract of 27 Jan 2025, counted as Raul's address data counts them). are estimated from Report 4's square counts, for large apartment buildings and dormitories that the register lists with two dwellings or fewer. A few buildings registered with non-residential parts (Västriku tn 8, Tornimäe tn 3) hold fewer dwellings in the register than in Raul's address data.
- Walking. OSM streets and paths (extract of ), as in OSRM's walking profile. Roads tagged foot=no (not trunk roads) stay walkable where a sidewalk is tagged on the road, or where the separately mapped sidewalk is missing for 100 m or more (Pärnu mnt by Rahumäe). Private ways serve only a building's or a place's own walk to the first public street, and two points in the same yard walk inside it; nobody else passes through. Aegna keeps its own paths. A building or a place joins the nearest path at the foot of the perpendicular, plus that short walk.
- Parcels, blocks and grid squares. Plain mean of the buildings whose inside point lies in them (all buildings, or only those with homes). Parcels: Tallinn cadastre, October 2026. Blocks: Raul's street blocks (Kvartali täisehitus, in Tallinn). Grid: Report 4's squares of 125 m to 2 km ( drawn: those in Tallinn or holding a Tallinn building); “Building mean” shows the buildings' mean, “Report 4 scores” or “Raul's values” the square's own number.
Index
- Report 4's arithmetic, per building. Each ingredient's level is its value divided by the 95th percentile of the Report 4 grid squares, capped at 1; a need counts the missing part. The weighted sum is stretched to 0–100 with the grid's 1st and 99th percentiles, so a building that scores 72 sits where a grid square scoring 72 would. The published scores exist only for the grid; the building, parcel and block scores are new.
- Report 4's walk. For each kind of place, the grid squares that hold one, weighted by the walk from the building to the square's point: full weight up to 10 minutes (about 830 m), half at 15, nothing from 20. That is close to Raul's W10, not his W15. Nature places are the POI list's outdoor places, three quarters of them outdoor gyms, skate parks, courts and fire pits; the list has no parks or forests, so tree cover is a map layer of its own. As in Report 4, basic schools and sport places are not counted: the POI file gives them no grid square (upper-secondary schools, universities and adult education are counted).
- Homes. Report 4 counts the dwellings in each grid square. "In its grid square" gives every building the count of the square that holds most of its footprint, Report 4's own term. "Around the building" counts the dwellings within 70.5 m, the area of a 125 m square; in squares of 250 m to 2 km Report 4 counts far more.
- Map layers (housing density, built density, jobs by public transport, sealed surface, tree cover) come from the same grid square. buildings lie outside Report 4's grid ( on Aegna, 6 at Paljassaare, Järvevana tee 3) and have no value where an ingredient comes from the grid square.
Check. Report 4's grid replayed on this network: r for the places; Fair-Share, Team 1 and Team 2 differ from the published scores by , and points on average. Building means per square against the published scores: r , , with homes counted in the grid square; , , with homes around the building.
Weight by people served
- What it does. Off by default. Switched on, each score is multiplied by the people a park there would serve, so that a place with the same need but more people nearby comes first. People served are the residents and workers within a 5-minute walk of the building, or of the grid square's point: full weight up to 400 m along the walking network, half at 600 m, none from 800 m (Raul Kalvo's W5 curve). The building's own people count too.
- The arithmetic. The score without the homes term is multiplied by a people weight: the people served (P) divided by the 95th percentile of P over Report 4's grid squares, capped at 1, then raised to the power k, the weight strength. The 5 % of squares with the most people served keep their score. The result is stretched to 0–100 again with the 1st and 99th percentiles of the grid squares' weighted scores, so the five bands, the filter, Top 20 % and Top 10 % work as before, and a building sits on the grid's scale. This second stretch raises the scores by a factor of 100 ÷ that 99th percentile (with “Both” at 0.5, about 1.11 for Fair-Share at strength 0.5 and 1.15 at strength 1; 1.0 for Team 1), so a place at full weight can score higher than without the weight. Parcels, blocks and “Building mean” take the mean of their buildings' weighted scores.
- Weight strength. k is 0.5 unless you change it (0 to 1, in steps of 0.1). At 1 the weight is in proportion to the people: a place with a quarter of the people of the busiest squares keeps a quarter of its score. At 0.5 it is the square root: a quarter of the people keeps half the score, so four times the people doubles the weight. At 0 the people do not count, and the score is the index without the homes term, stretched as usual. Above 0, a place with nobody within the 5-minute walk scores 0. Like the worker weight, 0.5 is a judgement: the need counts for more than in proportion, and places with more people still come first.
- What it does to the map. With Fair-Share, “Both” (a worker = 0.5 of a resident) and weight strength 0.5, the median grid square gets a weight of 0.39 (with workers alone, 0.28), six squares in ten fall in the lowest band (one in ten at strength 0), and the order follows the people served a little more than the need: the rank correlation over the grid squares is 0.86 with the people served and 0.68 with the score without homes. At weight strength 1 the median weight is 0.15, seven squares in ten fall in the lowest band, and the order mostly follows the people served (0.93 and 0.57). The need decides among places with many people.
- Homes are set aside. While the weight is on, the index's own Homes term is left out, and so are Housing density and Built density if they are in the mix: they measure people too, and people count once. The other weights are scaled up to 100 %, so Fair-Share becomes children's places 27 %, missing nature places 47 % and everyday services 27 %. On the grid the presets are then computed from their ingredients, not taken from the published Report 4 scores.
- Residents, workers or both. With “Both”, P = residents + a × workers, where a is how much a worker counts against a resident: 0.5 unless you change it. That 0.5 is a judgement, not a measurement: workers are near a park on weekday breaks, residents also in the evenings, at weekends and with children.
- Sources. Residents: Statistikaamet, 1 km population grid, 1 Jan 2026, CC BY-SA 4.0 (Eesti rahvastiku 1 x 1 km ruutkaart 2026), shared out inside each cell by the EHR dwellings of its buildings (for the grid squares, by Raul Kalvo's dwelling counts per square, so a square's residents can differ from the sum of its buildings'). Prisoners are left out, as they cannot use a public park; care homes stay. The three 1 km cells holding a prison (Tallinn prison at Soodevahe, Viru prison in Jõhvi, Tartu prison) keep only their dwellings × their municipality's residents per dwelling, which leaves out 1 250 residents in Estonia. Here that touches one square of Report 4's grid, at Soodevahe in Rae, outside the city and not drawn on this map (701 residents); no building on the map is within a 5-minute walk of it. Workers: Raul Kalvo, workers by place of work (terms being confirmed), shared out by workplace floor area from the building register (EHR use rows of Jan 2025, with jobs per m² by use from the HCA Employment Density Guide, UK 2015; the figures for education, health, hotels, defence and farm buildings are our assumptions). The densities only split Raul Kalvo's block totals; they add no workers. Walking curve: Raul Kalvo. Buildings are walked on this map's OSM network and also count the people in the buildings just outside the city; the grid squares use the site finder's numbers, walked on Raul Kalvo's network between square points. At the same place a building's count comes out about 5 % lower for residents and about 10 % lower for workers (medians), so building scores sit a little low on the grid's scale. The resident counts in the people layer (data/pw/) are adapted from the Statistikaamet grid; they and everything this map computes from them (the people served and the people-weighted scores) are shared under CC BY-SA 4.0, as the grid they come from; the licence does not reach the worker counts or the rest of the map. With “Workers” alone the people served are the worker counts, and the weight is computed from them, not from the residents. The worker counts are Raul Kalvo's; their terms, and those of his dwelling counts per square and walking distances, which the grid squares' resident figures also use, are being confirmed with him. The years are mixed: residents 1 Jan 2026; the dwellings that place them January 2025 (EHR), 2024–25 for the grid squares (Raul Kalvo); workers probably 2020 or 2021 (being confirmed); floor area 2025–26; walking network October 2026 for buildings, 2024 for the grid squares (Raul Kalvo).
Places in reach
- Raul's method. The points of his POI list v12 in Tallinn + 3 km ( of , kinds; lie more than 500 m from the network: 4 geocoded into the bay, 2 on a cut-off piece at Lagedi), each counted once, weighted by his curve: full weight up to 400 / 800 / 1 250 m (W5 / W10 / W15), half at 600 / 1 200 / 1 875 m, nothing from 800 / 1 600 / 2 500 m. A value reads as the number of places in reach. Four of its eight Report 4 baskets add kinds that Report 4's index does not have: basic schools (children's places, education) and sport places (nature places, places to move).
- Service targets. Our reading of his service list as the June 2025 deck shows it (walking, urban), not yet confirmed by Raul: targets such as a kindergarten or childcare within W10, 9 kinds of shop within W10, 4 kinds of culture venue within W15. A kind of place counts when its value reaches 1 (rounded to 2 decimals); a target is met when enough kinds count (the deck looks at coverage, not totals). This reading puts 15.0 % of Estonia's dwellings at his urban walking level with his own tables, the deck's 15 %. In Tallinn it is of dwellings ( of home buildings, as small houses meet fewer targets); his grid gives . Care homes are left out as in the deck; private and special-needs schools and non-stationary learning have no points in v12; the park-path target (3 km within W15), checked on his grid table, would change Tallinn's share by 0.03 points.
Check against his grid. His grid replayed on this network: r at W15 (median ratio ), at W10, at W5 (median ratio ). At W5 the network alone moves the values by about a fifth (the summed absolute difference is of the total) and exact positions by another fifth (); at W15 by and .
Differences from his tables. In v12, indoor play rooms (Mängutuba) have about a third of the points his tables used; deposit-return points (Pandipakend) and regional centres (Keskus, Piirkond) are missing.
POI list, grid, decay and service list: Raul Kalvo, EKA. Report 4 indices and the site finder's composer: EDITUA. Streets © OpenStreetMap contributors (ODbL). Buildings, cadastre and districts: Maa- ja Ruumiamet. Dwellings: EHR building register. Residents (people weight): Statistikaamet, 1 km population grid 2026, prisoners left out; the resident counts and what is computed from them (people served, people-weighted scores) are shared under CC BY-SA 4.0, which does not reach the worker counts or the rest of the map. Workers (people weight), and the dwelling counts and walking distances behind the grid squares' residents: Raul Kalvo, terms being confirmed. Computed 16–17 Oct 2026 for EDITUA; people weight 19 Oct 2026.