Gas Fee Forecast for Ethereum

How this forecast is built — and what it weighs

Most “gas predictors” on the web are dressed-up gas trackers — they show you the current gwei with a colored arrow. A real forecast has to commit to a number for a future time. This one blends four signals, weighted by recency:

1. Rolling base-fee trend. The last ~600 blocks (≈2 hours) tell you whether mainnet demand is climbing, cooling, or steady. Trend strength weights heavily for the next 1-3 hour horizon.
2. Time-of-day cycle. Gas follows recurring daily peaks tied to US trading hours (13:00-22:00 UTC). Overnight UTC and pre-US-market hours consistently run cheaper. This weight grows for 6-24 hour horizons.
3. Day-of-week pattern. Weekends (Saturday and Sunday UTC) run ~30-50% cheaper than mid-week US business hours. Friday late-UTC into Saturday morning is the cheapest stretch most weeks. This signal dominates the 24+ hour horizon.
4. Post-spike decay model. After a big base-fee spike (a major mint, a token launch, an exchange listing), prices decay back toward the trailing mean over 2-6 hours. The model accelerates the decay when the spike is more than 2σ above recent average.

The chart above is a likely-range visualization, not a single point. The center line is the median forecast; the band is the rough confidence envelope. Confidence narrows for short horizons (1-3 hours) and widens for longer ones — typical for any pattern- based forecast. Full methodology and data inputs →

Accuracy in practice — what to expect

Pattern-based forecasts work well on the dimensions where Ethereum gas is structured — time-of-day, day-of-week, and post-spike decay. They fail on the dimensions where Ethereum gas is unstructured — surprise mints, off-cycle CEX listings, and macro-level ETH price moves that change inclusion-tip behavior.

• 1-3 hour horizon: high confidence on direction. The forecast knows whether you should send now or wait another hour with usable accuracy.
• 6-12 hour horizon: high confidence on the cheaper window — when in the next half-day will gas be lowest. The exact gwei value at that window is best taken as a range, not a point.
• 24+ hour horizon: good for picking which day to act on (the day-of-week pattern is robust). The hour-by-hour shape past 24 hours is directional only.

When are gas fees lowest?

Across most weeks, the cheapest stretches sit on weekend nights and during late UTC hours on weekdays. For a deeper look at the weekly pattern, browse cheap gas times or check ethereum gas fees now for the live read.

How to act on the forecast

• If the chart trends down: non-urgent transactions can wait for the projected low.
• If the chart trends up: send routine transactions now before fees climb further.
• If the chart is flat: fees are unlikely to move much. Use the standard tier and don't overthink it.
• For anything truly time-sensitive: ignore the forecast — pay rapid and move on.
• For full automation: set a gas alert at the threshold the forecast suggests is reachable.

For deeper savings, see how to reduce gas fees.

When the forecast fails — and what to do

Any honest gas forecast tells you up front when it'll be wrong. This one fails on three recurring scenarios:

• Surprise NFT mints. A popular collection going live without a posted schedule pushes priority fees and wallet quotes up immediately, while the base fee climbs over several consecutive full blocks. A move from 15 to 80 gwei can't happen in a single EIP-1559 base-fee step (base fee changes at most ±12.5% per block), but it can happen within a few minutes if demand stays saturated. The pattern-based model has no way to predict the mint — it shows up as a delta from forecast, then the decay model takes over once the event is in the rolling window. If you see gas climbing fast above forecast, check for a mint name and wait 30-60 minutes before acting.
• CEX listing pump candles. A token going live on Binance or Coinbase often triggers a frontrunning arms race that lasts 1-3 hours and pulls base fee with it. Like NFT mints, these are unpredictable from on-chain data alone. The 2σ post-spike decay model handles the tail but not the spike itself.
• Mainnet upgrade days. Ethereum upgrades (Pectra, Fusaka, etc.) and major rollup events sometimes shift the underlying gas-cost structure overnight — EIP-4844 dropped rollup costs 10-20x and also shifted L1 patterns. Forecasts based on pre-upgrade data can be miscalibrated for the first 1-2 weeks after a hard fork. Treat any forecast across an upgrade boundary with extra caution.

In all three cases, the right move is the same: don't act on the forecast, set a gas alert at the target gwei you actually want, and let the queue resolve.

Frequently asked questions