# BTC volatility: 2025 vs 2026 (YTD)

Symbol: `BTCUSDT` perpetual (Binance USDⓈ-M). Source: normalized 1m bars,
aggregated to UTC days. 2025 is complete (365 days); 2026 data ends
**30 Sep 2026**, so all cross-year statements use the common
**1 Jan – 30 Sep** window unless labelled "full year".

Reproduce with `python scripts/btc_vol_compare.py`.
Chart: `reports/figures/btc_vol_2025_vs_2026.png`.

## Headline

| Metric (annualized) | 2025 full year | 2025 Jan–Sep | 2026 Jan–Sep |
|---|---:|---:|---:|
| Close-to-close vol | 41.8% | 41.7% | **46.1%** |
| Intraday realized vol (1m RV) | 45.3% | 44.2% | 44.8% |
| Parkinson (high/low) vol | 46.2% | 44.4% | 46.2% |
| Average daily range | 3.41% | 3.28% | 3.41% |
| Mean \|daily return\| | 1.57% | 1.52% | 1.63% |
| Max daily gain | +9.1% | +9.1% | **+11.5%** |
| Max daily loss | −8.9% | −8.9% | **−15.1%** |
| Days \|return\| > 3% | 54 | 42 | 37 |
| Days \|return\| > 5% | 15 | 11 | 13 |
| Excess kurtosis | 2.3 | 2.8 | **7.5** |
| Max drawdown | −32.1% | −28.1% | **−39.5%** |
| Price change over window | −7.4% | +20.5% | −4.6% |

**Same calendar window: 2026 realized vol is ~10.6% higher than 2025 on a
close-to-close basis (46.1% vs 41.7%).** But the increase is not broad-based:
intraday 1-minute realized variance is essentially unchanged (44.8% vs 44.2%),
and average daily range is identical at 3.41%. The whole difference sits in the
tails — 2026 produced fewer mid-sized (3%+) days but much larger outlier days,
a −39.5% drawdown, and 2.5–3× the excess kurtosis.

## Volatility regime by month

Annualized close-to-close vol:

| Month | 2025 | 2026 |
|---|---:|---:|
| Jan | 43.3% | 43.2% |
| Feb | 38.7% | **82.3%** |
| Mar | **66.7%** | 49.5% |
| Apr | 54.4% | 36.4% |
| May | 36.0% | 26.1% |
| Jun | 31.9% | 44.1% |
| Jul | 26.5% | 31.3% |
| Aug | 35.9% | 41.5% |
| Sep | 24.5% | 41.2% |
| Oct | 44.0% | — |
| Nov | 44.6% | — |
| Dec | 37.3% | — |

Monthly return (%), for context:

| Month | 2025 | 2026 |
|---|---:|---:|
| Jan | +7.9% | −10.7% |
| Feb | −19.4% | −16.2% |
| Mar | −2.1% | +1.9% |
| Apr | +13.2% | +11.2% |
| May | +10.5% | −3.5% |
| Jun | +2.4% | −22.9% |
| Jul | +7.7% | +7.0% |
| Aug | −6.7% | +22.3% |
| Sep | +5.2% | +6.2% |

Quarterly vol: 2025 Q1 51.5 → Q2 41.5 → Q3 29.4 → Q4 41.9;
2026 Q1 59.5 → Q2 36.9 → Q3 38.1.

The two years are close to **mirror images within the year**:

- **2025 front-loaded its volatility.** Q1 was the hot quarter (66.7% March,
  the Feb–Mar selloff), then vol ground down to a 24–26% floor in Jul–Sep, and
  re-spiked only in Oct (the 10 Oct liquidation, −7.6% day, biggest 1-minute
  drop of the sample at −5.1%) and Nov.
- **2026 carried its volatility in bursts and never fully calmed down.** The
  5 Feb crash (−15.1% in a day, low 62.2k from a 73.1k open, +11.5% the next
  day) took the 30-day vol to **84%**, the highest of either year. June was
  another −22.9% month. And unlike 2025, Q3 stayed elevated (Sep 41.2% vs
  24.5%) despite a +22.3% August rally.

## Distribution of 30-day vol

Rolling 30-day annualized vol:

| Stat | 2025 | 2026 (to Sep 30) |
|---|---:|---:|
| Mean | 40.6% | 42.4% |
| Median | 37.8% | 40.8% |
| Std (vol-of-vol) | 11.7% | **15.7%** |
| Min | 21.4% | 21.9% |
| Max | 71.8% | **84.0%** |
| Share of days > 60% | 7% | 11% |
| Share of days < 30% | 19% | 21% |

Medians are close (37.8% vs 40.8%, 30d vol sits around the 50th percentile of
the pooled 2025–26 distribution in both years), but the dispersion is very
different: vol-of-vol is up ~34%, and 2026 spends noticeably more time above
60% while also giving more calm sub-30% days. That is a **two-regime year**
rather than a uniformly more volatile one.

## Interpretation

1. **Vol level:** 2026 is modestly higher (+4.3 vol points, +10.6% relative)
   over the comparable window. Not a different volatility regime by level, but
   a different *shape*.
2. **Where the extra variance comes from:** not from continuous intraday
   churn — 1m RV and average range are flat across years. It comes from a
   handful of directional shock days. In 2025 the biggest daily loss was
   −8.9%; in 2026 it was −15.1%, and the loss distribution has visibly fatter
   tails (kurtosis 7.5 vs 2.3).
3. **Path risk:** 2026's max drawdown (−39.5%) is ~40% deeper than 2025's
   (−32.1%), with a similar number of >5% days (13 vs 15). Fewer medium shocks,
   more large ones.
4. **Practical takeaway:** vol-targeting or fixed-vol sizing calibrated on 2025
   would have under-risked the median 2026 day slightly but badly under-priced
   the tail. Stress/scenario sizing based on the 2025 max daily loss would have
   been off by ~6 vol points on the worst 2026 day. If anything, 2026 argues
   for sizing on tail quantiles rather than average realized vol.

## Caveats

- 2026 is YTD through 30 Sep; Q4 is missing and Q4 2025 was itself elevated
  (Oct 44.0%, Nov 44.6%), so the full-year 2026 number could move either way.
- Vol is measured on the perp only; no funding, basis, or options-implied vol is
  included. Perp trades 24/7, so "overnight gap" variance is ~0 and
  close-to-close ≈ open-to-close.
- 1m RV excludes the 00:00 UTC bar (overnight jump) by construction; it is
  therefore an intraday variance measure and is expected to sit below
  close-to-close vol when moves trend across days.