REG-10 · Regime Detection

HMM Regime Detection

Hidden Markov Model identifies latent market states (bull/bear). Uses Gaussian emissions and Baum-Welch for parameter estimation. Trades based on inferred regime.

RegimeMachine LearningStatistical

Open interactive charts Compare with others

At a glance

Best onMarkets with distinct regimes

Signal frequencyDaily

Typical holding periodWeeks

Asset classesIndex, Commodity, FX

Markets tested10

Data through2026-11-10

What the backtest shows

Beat buy-and-hold on 3 of 10 markets by annual return over 5 years.
Had a smaller maximum drawdown than buy-and-hold on 8 of 10 markets.
Best market by Sharpe ratio: SPY (Sharpe 0.78, CAGR 7.5%).
Weakest market: DBA (Sharpe -0.06, CAGR -1.0%).

Equity curve

HMM Regime Detection Buy & hold · SPY, 2010-01-04 → 2026-11-10, growth of 1

Score and risk profile

Arena Score: 74.8SPY, 5 years

Return quality · 30%85

Drawdown control · 25%77

Consistency · 20%63

Robustness · 15%52

Significance · 10%96

Each bar is a percentile against every strategy on the same market and window (100 = best).

Time in market49%

Ulcer index4.4%

Longest drawdown571 days

Daily VaR (95%)-0.88%

Probability Sharpe > 096%

Beta vs buy & hold0.32

Alpha (annual)+3.8%

Positive months48%

CAGR after 10 bp costs5.3%

Deflated Sharpe (luck-adjusted)62%

CAGR, 90% bootstrap range0.3% to 16.0%

Max drawdown, 90% bootstrap range-25.6% to -8.1%

How it held up in sell-offs (SPY)

Period	Strategy	Buy & hold
Q4 2018 sell-off	-3.7%	-19.2%
COVID crash (Feb–Mar 2020)	-7.6%	-33.7%
2022 bear market	-8.4%	-24.5%
Market above its 200-day average (annualized) · 81%	+5.2%	+12.5%
Market below its 200-day average (annualized) · 19%	+9.6%	+19.8%

Sell-off dates follow the US stock market. Above/below the 200-day average uses the previous day's position, so there is no look-ahead.

Out-of-sample check (SPY)

Check	Value
Chosen parameters on their 3-year training windows (average Sharpe)	1.06
Same parameters in the following 6 months, when traded (Sharpe)	0.81
Default parameters over the same months (Sharpe)	0.93
Parameter re-selections / changes	28 / 10
Parameter settings with a positive Sharpe (full history)	9 of 9
Sharpe across settings: worst / median / best	0.56 / 0.76 / 0.83

Out-of-sample periods start on 2013-01-04. Training-window figures are flattered by picking the best of 9 settings; the out-of-sample figure is what the strategy actually earned. A large drop, or a result that depends on one setting, points to overfitting. How walk-forward testing works.

Results by market (5 years)

Market	CAGR	Buy & hold CAGR	Sharpe	Max drawdown	Buy & hold max DD	Volatility	Arena Score
SPYSPDR S&P 500	7.5%	11.6%	0.78	-14.8%	-24.5%	9.8%	75
USOUnited States Oil Fund	15.5%	20.2%	0.67	-28.8%	-36.2%	27.5%	79
GLDSPDR Gold Shares	5.7%	16.4%	0.47	-19.2%	-28.3%	14.4%	40
UNGUnited States Natural Gas	-1.1%	-30.7%	0.21	-66.8%	-93.3%	45.6%	80
ETH-USDEthereum	0.9%	-15.1%	0.20	-58.2%	-79.0%	38.2%	65
SLViShares Silver Trust	-2.6%	15.2%	0.07	-52.5%	-57.8%	29.1%	27
BTC-USDBitcoin	-2.9%	2.6%	0.06	-65.8%	-75.9%	28.2%	34
EURUSD=XEUR/USD	0.0%	-0.7%	0.03	-12.4%	-17.2%	5.4%	55
DBCInvesco DB Commodity Index	-0.6%	10.8%	0.03	-43.1%	-27.3%	14.1%	24
DBAInvesco DB Agriculture	-1.0%	8.8%	-0.06	-22.4%	-15.9%	9.4%	22

Strategy metrics use daily returns without trading costs. Buy-and-hold figures are for the same market and window. Rows are sorted by Sharpe ratio.

SPY across time windows

Window	Total return	CAGR	Sharpe	Sortino	Max DD	Win days	Trades
1 year	+1.6%	1.6%	0.28	0.20	-7.6%	17%	23
3 years	+42.2%	12.5%	1.53	1.42	-7.6%	29%	80
5 years	+43.8%	7.5%	0.78	0.74	-14.8%	27%	111
Full history	+167.3%	6.0%	0.68	0.60	-18.2%	28%	426

Rules

Entry

Buy when in bullish regime (higher mean state)

Exit

Sell when in bearish regime

Indicators

Returns as observations
HMM with Gaussian emissions
Infer most likely state sequence

Inputs

N_States = 2
Lookback = 252

Position sizing

Fixed 100% allocation

Risk controls

Regime probability threshold

Strategy evidence

Only available, validated bundles appear below. Missing evidence is not a passing validation result.

Evidence coverage

10 / 10 declared strategy–market pairs validated

Notes

Simplified HMM without external libraries
Uses rolling window for regime detection
Bull regime = positive mean returns, Bear = negative

Related strategies

MACRO-02 · SPY 5Y Sharpe 0.40Risk-On/Risk-Off Switch MOM-06 · GLD 5Y Sharpe 0.57Residual Momentum MR-03 · SPY 5Y Sharpe 0.25Kalman Filter Mean Reversion MR-04 · SPY 5Y Sharpe 0.24Ornstein-Uhlenbeck Mean Reversion MR-06 · SPY 5Y Sharpe 0.86Z-Score Mean Reversion MR-07 · SPY 5Y Sharpe 0.37Hurst-Filtered Mean Reversion

Embed this score

Add the live badge to a blog post, README or forum signature. It updates with the daily data.

<a href="https://quantarenas.com/strategy/reg-10-hmm-regime-detection/"><img src="https://quantarenas.com/badge/reg-10-hmm-regime-detection.svg" alt="HMM Regime Detection Arena Score on Quant Arena" height="20"></a>

[![HMM Regime Detection Arena Score](https://quantarenas.com/badge/reg-10-hmm-regime-detection.svg)](https://quantarenas.com/strategy/reg-10-hmm-regime-detection/)

Backtests are historical simulations on daily data and do not include commissions or slippage. They do not guarantee future results and are not investment advice. See the methodology for how every strategy is tested.