The $1,231/sqft cousins · what 1465 23rd Ave is worth

Last-month sales at 1820 Kirkham's price-per-foot, judged down to true block-level comps · $2.8M cousins → · when to buy → · bid odds → · browser →

The four price-cousins — sold 2026-05 … 2026-06

All sold last month within −5% / +8% of $1,231/sqft. Sorted small → large.

Each cousin, against its real block-level comps

The same $1,231/sqft means a different thing for each — and only the homes that truly match on location, finish, and size are kept (with the LLM's one-line reason).

All the surviving comps, one picture

Every kept comp across the cousins, $/sqft over four years, colored by cousin. Far fewer than a blanket size filter would admit — but each is a genuine block-level twin.

● 1683 21st   ● 1623 34th   ● 1820 Kirkham

So where does 1465 land? — 14 judged comps

1465's own comps: the renovated 4-beds an LLM judged genuinely similar to its two closest recent twins, 1820 Kirkham and 1501 28th. As a group they sit at $1,155/sqft today (median). One adjustment matters: within this set, $/sqft falls as homes get bigger (corr -0.65), and 1465 is ~5% bigger than the twin — that shaves about $27/sqft off the twin's rate, folded into the twin anchor below.

Judged comps' $/sqft over four years — recent prints are firming up:

But 1465 isn't a median home — it's the nicest one

So we scored each comp's finish quality 0–100 from its description alone (renovation depth, finishes, features — no location or size). 1465 scores 95 — the highest in the whole set (the best comp reaches 94; the twin 1820 Kirkham only 83, yet it still sold for $1,231/sqft). Finish does move price here, modestly, so 1465 belongs near the top of its comps' range, not the middle:

Each dot a block-level comp (x = finish, y = today's $/sqft); the line is the finish→price fit. ◆ 1465 is placed at our estimate (finish 95) — it hasn't sold, so it's a marker, not an observation.

Location — scored like finish, and just as independent

People pay for the spot, not just the finish. So we also score each home's location 0–100: how much buyers pay per foot on its block once renovation and size are stripped out, ranked across the district. 1465 scores 34 — mid-Sunset, well below the inner avenues (8th–17th run 85–100) but above its own comps' median of 16, which skew toward the beach.

Finish and location turn out to be independent here — they barely correlate, so they're two separate levers. Fit both at once and $/sqft ≈ +$4 per finish point and +$2 per location point. That lets us place every comp on a finish × location plane, color it by $/sqft, and drop 1465 onto it:

Brighter dots sold in the last year. ★ = 1465 (finish 95, location 34); the surface there reads $1,160–$1,117/sqft → $2.72M. This is the honest tension: 1465's top-tier finish argues for the top of its cohort, but its mid-Sunset location pulls a two-axis estimate back toward the floor. It still out-scores its comps on both axes — which is why our centre sits above the cohort median, just not at the inner-avenue ceiling.

A fair price today for 1465

Against the homes that truly share its blocks and its calibre of finish, 1465 reads $2.88M–$2.96M, center about $2.88M (~$1,207/sqft). If it sold like a median comp it would fetch $2.76M — but it won't: it's the single nicest home in the set. The twin 1820 Kirkham (finish 83, two blocks away) sold at $1,231/sqft; adjusted to 1465's larger size that's $1,204/sqft → $2.88M, and 1465 out-finishes the twin, which opens the top of the range near $2.96M. It doesn't go higher because 1465 sits on central 23rd Ave, not the pricey inner avenues (8th–17th, $1,300–1,600/sqft). The $2.30M list is a strategic under-price; expect a sale well over ask. The $2.8M-cousins view → reaches the same range from total prices.

Method & caveats

• Seeds: Compass closings with sold_date ≥ 2026-05-01 (the last ~6 weeks of closings) and $/sqft ∈ [$1,169, $1,329] (1820 Kirkham's $1,231, −5%/+8%). The band is asymmetric on purpose — the +8% upper reach lets in comps a notch nicer than the twin, since 1465 out-finishes it — though at today's data a symmetric ±5% admits the identical seeds.
• Sanity vs Zillow: the public Zestimate (~$1.87M) sits far below every comp-based read here. Zestimates anchor on the bait list price / county assessment and badly misprice down-to-studs renovations — we don't use it.
• The top of the range is a judgment, not a measurement: the "designer premium" is a hand-set +3% over the twin's actual $/sqft (sensitivity 0–6% → $2.88M–$3.05M), and the centre is the (top-decile + twin)/2 midpoint. Treat them as a reasoned upper case; the measured part is the comp floor ($2.76M) and the twin's print.
• 1465's finish (95) sits one point above the best comp (94); the verdict leans on the cohort's top decile and the twin's actual sale, not the exact score, so it's unchanged for 1465 ∈ {90–95}.
• Candidates: each seed's nearest 50 single-family sales of the same renovation family (±25% sqft, ±1 bed, trailing 4 yrs). An LLM scored each 0–100 on micro-location + renovation/finish (from the listing description) + size; comps scoring ≥ 65 are kept. Scores in data/compass-sold/cousins-scores.json.
• Finish: a second LLM pass scored each home's renovation/finish quality 0–100 from its description alone (location/size-independent), on a fixed rubric. 1465 scores 95, the top of its comp set. 1465 is placed near the cohort's top decile of $/sqft accordingly (data/compass-sold/cousins-finish.json).
• "Today" = re-priced by the hedonic price index (how these homes actually moved — not national CPI); comp percentiles are of the kept comps' today's-$ $/sqft. Small, hand-judged samples — a centre-of-range, not an appraisal. 1465 is the on-market reference, not a comp.
• Generated by scripts/build-1231-sqft-cousins.mjs (scoring by scripts/score-cousins.mjs).