Best timeframes for automated Hyperliquid perps: what the optimizer found

The question

Every automated strategy runs on a chart timeframe, and the choice changes everything: how often it trades, how much fees matter, how noisy the signals are and how long positions are held. TensorTrader's optimizer tunes the Lorentzian Classification strategy separately for each token and timeframe, so its catalog is a large natural experiment on this question.

What the catalog shows

Snapshot from the public TT-Autotune catalog on 25 September 2026. The live numbers per token are on the TT-Autotune pages.

trained tokens

63

token × timeframe cells

288, across 5m, 15m, 30m, 1h, 4h and 1d

best timeframe per token

1d for 35 tokens, 4h for 25, 1h for 3 (by tuned ROI averaged across regimes)

regime cells beating default settings

1,576 of 1,728 (91%)

• Browse every tuned cell →

Why higher timeframes win on paper

Two effects push the optimizer toward 4-hour and daily charts. Signal quality: lower timeframes carry more noise relative to the move, so a nearest-neighbour vote has less to work with. Costs: a 5-minute strategy trades many times more often than a daily one, and every trade pays fees and slippage. With a realistic cost model of 5 basis points fee and 2 basis points slippage per side, the fast cells have to clear a much higher bar.

Hyperliquid's fees are competitive, but the arithmetic still favors fewer, larger moves. That lines up with the fee drag we saw in our own testnet books.

The caveats, stated plainly

1. 1ROI is summed over each cell’s own history window, and daily cells cover a longer calendar span than 5-minute cells, so the comparison is not perfectly like for like.
2. 2These are optimizer results scored on walk-forward windows, not live trading. Beating the default in-sample is necessary, not sufficient.
3. 3Some tokens have short daily histories; results on thin data are fragile.
4. 4Our forward testnet books have not shown an edge after fees. Treat every number here as a starting hypothesis to test on paper.

A practical approach

Start with 4-hour and daily cells for your chosen tokens, because they trade less and suffer less from costs. If you want more activity, add a lower timeframe as a DCA leg rather than a separate strategy, so it adjusts a position instead of opening its own. TensorTrader's DCA Across Timeframes does exactly that, weighting margin toward the base timeframe.

Then measure: run the setup on the Hyperliquid testnet for a few weeks, compare net ROI after fees per timeframe, and cut the timeframes that do not pay for themselves.

Per-token results

Each trained token has its own tuned-versus-default numbers per timeframe and regime on the TT-Autotune catalog, refreshed as new optimizer runs finish.

Timeframe and regime interact

The best timeframe for a token also depends on the market regime. A trend-following setting that shines on the 4-hour chart in a strong bull regime can bleed on the same chart in chop. TT-Autotune handles this by tuning a separate champion per regime for every token and timeframe, then switching champions live as the classifier changes its label. When you pick timeframes, you are also picking how often regime switches will hit your positions: faster timeframes see more of them.

• Market regimes explained →

Mapping timeframes to holding times

5m-15m

Minutes to a few hours; fee-sensitive; needs strong filters

30m-1h

Hours; the usual intraday compromise

4h

Half a day to several days; funding starts to matter

1d

Days to weeks; fewest trades; funding and gaps matter most

Frequently asked questions

Is the daily chart always best?

No. It won most often in our catalog, but results vary by token and regime, and the comparison has window-length caveats.

Should I avoid 5-minute strategies?

Be careful with them: they need a large edge to cover fees and slippage. Test them on paper first.

Keep reading

• Hyperliquid trading bot: how to automate Hyperliquid perps safely

  The complete guide to automating Hyperliquid: agent wallets, testnet, TradingView and GoCharting alerts, leverage and margin, fees and funding, portfolio mode, and the real results of our own testnet books.

• How TT-Autotune works: one alert, six regime champions

  TT-Autotune tunes the Lorentzian Classification strategy per token, per timeframe and per market regime, then syncs the winning settings into a single TradingView alert. Here is how the optimizer, the regime switch and the extension fit together, plus a live testnet ROI chart.

• DCA across timeframes: how margin splits by weight

  How TensorTrader’s DCA Across Timeframes spreads one strategy across a spectrum of chart timeframes, weights margin toward the base, and fits your alert quota.

• How fees and slippage destroy trading bot ROI, with real numbers

  Why a strategy that looks profitable before costs can lose badly after them: round-trip costs, compounding over thousands of trades, and fill-based accounting from TensorTrader’s own books.