Skip to content
HelmFi

Helm Index / GatedIndex

The strategy.
The evidence.

A regime-gated basket of major cryptocurrencies. The rule adjusts exposure as trend conditions change. Read the historical test and its limits together.

Explore the demo

$10,000 invested Jan 2021 — hypothetical, validated backtest

Helm Index and Bitcoin, buy & hold · through Apr 5, 2026

Helm Index$51,100
Bitcoin, buy & hold$23,300
Weekly close 275 / 275
$63,157$5,550
Helm Index versus Bitcoin, buy & holdBlue solid line: Helm Index. Gray dashed line: Bitcoin. Weekly backtest values, with an initial $10,000 normalization anchor.

Replay both strategies or drag the timeline. Values shown rounded to the nearest $100.

Backtest period & methodology
Read the Helm Index evidence ↗

275 weekly closes from Jan 3, 2021 to Apr 5, 2026. Both series use the existing January 1, 2021 normalization at $10,000; that starting point is an investment anchor, not an extra weekly observation.

Historical validation artifact · 11 major coins · fees modeled. Final values before display rounding: Helm Index $51,103; Bitcoin $23,278.

Hypothetical backtest results are not live-money results and do not predict future returns.

Live forward test

Live since April 2026.

The Helm Index runs live on the founder's own funds since April 2026, a forward test in progress and currently in drawdown. Everyone else runs it on paper. The backtest above is history; the live posture is on the account dashboard.

How the rule works

Exposure follows
the regime.

Equal-weighted holdings, volatility targeting, and a trend gate. The goal is to manage exposure to crypto markets.

Read the methodology
01

Observe

The gate checks how much of the universe is trading above its 50-day average.

02

Allocate

When the condition permits exposure, the basket targets equal weights with volatility adjustment.

03

Step aside

The gate moves to cash when the trend condition fails. Re-entry follows the same rule.

04

Rebalance

The existing strategy uses a 72-hour rebalance cadence. Between rebalances, individual assets still carry market risk.

The limits belong beside the chart.

It dampens bears, it does not erase them
2022 still drew down (-14% max drawdown within calendar 2022; -20% peak-to-trough across the 2021→2022 cycle, vs -38% ungated and -64% BTC over the same tested bear window — BTC's calendar-2022 max drawdown was -67%). Because of this it fails an absolute “no-negative-walk-forward-window” gate on the 2021–2026 history (the 2021→2022 window nets -7%). It is a risk-reducer, not a money-printer.
It sits in cash 47–57% of the time and whipsaws often
117 regime flips on 11-major 2021–2026, many 1–2 day blips. It will badly underperform buy-and-hold in strong bull runs (2020: +89% vs BTC +302%; 2023: +54% vs +156%). You are buying drawdown protection and paying for it in missed upside plus turnover.
No per-position stop — single-name bag-holding risk
The per-position intra-rebalance stop was removed 2026-06-13 (a validator A/B found it return-negative and drawdown-neutral — it cascade-liquidated the basket on routine dips). Between the 72h rebalances a single name can now draw down freely; drawdown control comes from the regime gate plus vol-targeting, not a stop. A single 1h candle can still drop a name -25% to -32% (e.g. DOT -31.6% on 2021-05-19).
PBO is borderline
PBO 0.457 passes <0.5 on the 11-major 2021–2026 run, but the CSCV noise floor is ~0.40, so the margin is thin. Bear evidence spans two independent bears — 2018 (4-name universe, downloaded 2017–2020 history) and 2022 (11-major) — but only 2022 exists on the full 11-major set. The specific 50d/0.5 parameters are not strongly learnable; what is sound is the binary in-cash mechanism, not the exact tuning.
Historical evidence is not a live account
The bear-market evidence is in-sample history. The saved record does not establish completed forward validation. These results are historical validation artifacts. The account dashboard is the source for current holdings and execution state. A UI replay does not rerun strategy validation.

Inspect the saved validation.

Historical artifact · 2026-04-05

Calendar-year maximum drawdown

YearHelm IndexBitcoin
2021-13%-53%
2022-14%-67%
2023-12%-20%
2024-26%-26%
2025-17%-32%
2026-11%-35%

Overfitting & walk-forward checks

11-major 2021–2026

PBO0.457
DSR at 100 trials0.995
Positive walk-forward windows91%
Median IS → OOS Sharpe1.16

DSR survives deflation for the entire sprint's search breadth (~100 configs), not just the 15-config gate grid. CSCV math independently verified against scipy + a López de Prado reference + a noise/dominant sanity test. PBO passes (<0.5) but the CSCV noise floor is ~0.40, so the margin is thinner than it looks.

Shipped config, 500 permutations — the observed result beat all 500 random relabelings (p < 0.002 at this resolution; the gate requires p < 0.10). Artifact re-run 2026-07-06.

Every figure here is a historical validation artifact, not a yield or return guarantee. Past performance does not predict future results.

Bear-window evidence

2018 · 4-name (BTC/ETH/LTC/BNB)

Gated index: -3%. Same index without the gate: -34%. Bitcoin: -73%. Mostly in cash through the 2018 bear.

2022 · 11-major

Gated index: -2.0% (window return). Same index without the gate: -35.0%. Bitcoin: -64.0%. ~100% cash on 69% of rebalances.

Window returns and calendar-year drawdowns measure different things. Every figure here is a historical validation artifact, not a yield or return guarantee. Past performance does not predict future results.

Historical cash windows
2021-11-18 → 2022-03-17Caught the Nov-2021 top, out through the start of the decline.
2022-04-11 → 2022-07-18Out through the May–Jul 2022 crash (LUNA / 3AC).
2022-11-08 → 2023-01-08Out through the FTX collapse.
2025-01-22 → 2025-04-22Out through early-2025 weakness.
2025-10-10 → 2026-01-02Out through late-2025 weakness.
Early 2026Cash through the spring drawdown, then re-entered the basket when the trend turned back up.

The cash windows above are from the 11-major 2021–2026 backtest; the 2018 bear is shown in the bear-protection cards.

Helm Index — evidence · HelmFi