Measured Sharpe dispersion across non-overlapping SPY windows, 2006 to 2025
Answered against 22 years of US equities and 12 years of US options data and published with the query that produced it. This result is stored as of 2026-08-14, from Bootstrapping Backtest Confidence Bands.
| horizon | block_count | sharpe_p05 | sharpe_p50 | sharpe_p95 | sharpe_band_width |
|---|---|---|---|---|---|
| 21 sessions | 238 | -3.37 | 1.47 | 6.97 | 10.34 |
| 63 sessions | 78 | -1.82 | 1.27 | 4.18 | 6 |
| 126 sessions | 38 | -1 | 1.16 | 2.87 | 3.86 |
| 252 sessions | 18 | -0.7 | 0.92 | 2.22 | 2.92 |
| 504 sessions | 8 | -0.09 | 0.54 | 1.58 | 1.67 |
- Rows × columns
- 5 × 6
- Computed
- Completeness
- No missing values
- Source
- US exchange, SIP and OPRA market data
- Licence
- Strasmore terms · free, no signup
What each column holds
| Column | Type | Range | Notes |
|---|---|---|---|
horizon |
text | 5 distinct values (126 sessions, 21 sessions, 252 sessions…) | |
block_count |
number | 8 to 238 | count |
sharpe_p05 |
number | -3.37 to -0.09 | |
sharpe_p50 |
number | 0.54 to 1.47 | |
sharpe_p95 |
number | 1.58 to 6.97 | |
sharpe_band_width |
number | 1.67 to 10.34 |
Computed from Strasmore's warehouse of US exchange, SIP and OPRA market data. Equity prices are delayed; options greeks and implied volatility are end-of-day. This result is stored, not recomputed on load — it is exactly the numbers that were returned on , and the query below is what returned them.
the exact SQL behind every number
WITH daily AS
(
SELECT
date,
toFloat64(close) / nullIf(lagInFrame(toFloat64(close), 1)
OVER (ORDER BY date ROWS BETWEEN 1 PRECEDING AND CURRENT ROW), 0) - 1 AS ret
FROM global_markets.stocks_daily_aggs
WHERE ticker = 'SPY'
AND date >= '2005-12-01'
AND date < '2026-01-01'
),
indexed AS
(
SELECT
ret,
row_number() OVER (ORDER BY date) AS i
FROM daily
WHERE date >= '2006-01-01'
AND ret IS NOT NULL
),
blocks AS
(
SELECT
w,
intDiv(i, w) AS blk,
count() AS n,
avg(ret) / nullIf(stddevSamp(ret), 0) * sqrt(252) AS sharpe
FROM indexed
CROSS JOIN (SELECT arrayJoin([21, 63, 126, 252, 504]) AS w) AS ws
GROUP BY w, blk
HAVING n = w
)
SELECT
concat(toString(w), ' sessions') AS horizon,
count() AS block_count,
round(quantileDeterministic(0.05)(sharpe, blk), 2) AS sharpe_p05,
round(quantileDeterministic(0.50)(sharpe, blk), 2) AS sharpe_p50,
round(quantileDeterministic(0.95)(sharpe, blk), 2) AS sharpe_p95,
round(quantileDeterministic(0.95)(sharpe, blk)
- quantileDeterministic(0.05)(sharpe, blk), 2) AS sharpe_band_width
FROM blocks
WHERE sharpe IS NOT NULL
GROUP BY w
ORDER BY w
Run your own version of this
The same 22 years of US equities and 12 years of options data are queryable in SQL or plain English. A free account runs 100 queries a day and takes no card.
More from this analysisBootstrapping Backtest Confidence Bands
One position, one year at a time: SPY annualized Sharpe by calendar year
table 14×5
→
Variance ratio by block length: does SPY variance scale like independent draws?
ranking 7×3
→
Lag-one autocorrelation: signed returns against absolute returns, 2016 to 2025
ranking 6×4
→
Weekly z score of the KO/PEP spread, hedge ratio fitted on 2023 only
series 105×2
→
Hedge ratio refitted each calendar year, two sector pairs
ranking 7×3
→
Where the KO/PEP spread sat twenty sessions later, by starting z score (2019-2025)
ranking 6×4
→
See all 2,170 queries →