Gas Fee Trends: Forecasting Network Congestion On-Chain

DeFi Intel Research · Updated 2026-07-02 · intermediate

Gas fees on Ethereum and other EVM-compatible chains can feel like a capricious tax—one minute you pay pennies, the next you forfeit a small fortune for a simple swap. Yet beneath the surface, fee spikes are not random; they follow predictable patterns driven by network demand, block space constraints, and mempool dynamics. By reading on-chain metrics, you can forecast congestion and time your transactions to save up to 80% on costs.

This guide moves beyond basic “low gas hours” advice. We’ll dissect the three foundational metrics that professional traders and DeFi power users monitor: mempool depth, gas price percentiles, and block utilization. You’ll learn how to combine these signals into a practical forecasting model, which tools to use, and how to integrate this knowledge into your own workflow.

By the end, you’ll be able to spot congestion buildup before it hits the mempool, avoid bidding wars during NFT mints, and execute time-sensitive operations with confidence. No more guessing—just data-driven fee management.

The Importance of Gas Fee Forecasting

Every Ethereum transaction competes for limited block space. When demand spikes—due to a popular NFT mint, a DeFi launch, or a market crash—fees can soar 10x in minutes. For frequent transactors, even a single poorly timed transaction can erase a day’s trading profits. Forecasting is not a luxury; it’s a risk-management essential.

Smart contracts, MEV bots, and arbitrageurs constantly monitor the mempool and preemptively adjust bids. Retail users who send transactions blindly often overpay by 2-5x. By understanding fee dynamics, you can:

• Schedule token swaps during predictable low-activity windows (e.g., early Sunday mornings UTC).
• Avoid the final seconds of a Dutch auction or a hyped NFT reveal.
• Set optimal gas prices that confirm within a reasonable time without overpaying.

Forecasting also helps when gas fees are high: you can either wait for the mempool to clear or use L2 solutions. The key is recognizing that the fee market is a self-correcting system—sustained high fees eventually price out non-urgent users, demand falls, and fees settle back down. Metrics tell you where in that cycle you are.

Mempool Data: Where Fees Are Born

The mempool (memory pool) is the waiting room for all pending transactions. Each node maintains its own list of unconfirmed txns; validators (block proposers) choose from this pool which to include in the next block, typically prioritizing higher fees. Mempool depth—the total number of pending transactions—is the earliest indicator of congestion.

A sudden surge in mempool size (e.g., from 50,000 to 150,000 pending) signals that more users are competing for the next few blocks. But raw count alone isn’t enough. You must filter by gas price tiers. A mempool full of low-fee txns (below 10 gwei) indicates many cheap attempts that will wait; a mempool stacked with high-fee txns (above 100 gwei) means a premium bidding war is underway.

Key on-chain metrics to track:

• Mempool tx count — total pending, split by fee range.
• Mempool growth rate — is it accelerating or leveling off?
• Weight — some txns consume more gas (e.g., complex swaps vs simple ETH transfers).

Platforms like Etherscan’s Mempool Explorer, Blocknative, and Dune dashboards provide real-time mempool visualizations. Watching the mempool fill is like watching a pressure gauge—once it hits a critical threshold, fee spikes follow within minutes.

Gas Price Percentiles: The Fee Market Thermometer

Gas price percentiles reveal what the majority of users are paying and what they’re willing to pay. The most commonly watched are the 5th, 50th, and 95th percentiles. The 5th percentile shows the floor—the bare minimum to eventually get a transaction through during lulls. The 50th (median) is a good indicator of “normal” activity. The 95th percentile reflects the urgency of power users and bots.

A widening gap between the 5th and 95th percentiles indicates market stress: those who must transact now are paying a huge premium, while patient users can still pay low fees if they wait. Conversely, a narrow spread suggests a calm market where even urgent transactions are cheap.

How to use percentiles in forecasting:

• When the 95th percentile spikes while the 5th stays low, expect fees to drop back to median once the urgent wave passes.
• If the 5th percentile also rises above 50 gwei (illustrative), that’s a wholesale congestion event—consider delaying.
• Track the curve over the last 24 hours. A steep upward slope in the 50th percentile signals building demand.

Data sources: Etherscan’s Gas Tracker (which shows “Low/Average/High” often based on percentiles), or Dune queries that plot hourly percentile bands. A moving average of the 90th percentile is a reliable leading indicator for fee spikes.

Block Utilization: Measuring Network Capacity

Each Ethereum block has a gas limit (tens of millions of gas units; it is validator-adjustable and has been raised over time). Block utilization is the percentage of that limit actually used. When blocks are consistently above 95% full, the network is saturated—every slot is contested. Low utilization (e.g., 50-70%) means ample space and cheap fees.

Block utilization trends are a lagging indicator compared to mempool depth, but they confirm whether high fees are due to genuine demand or temporary bot activity. A block consistently 99% full means the base fee (EIP-1559’s mechanism) will rise automatically, squeezing all transactions.

Watch for these patterns:

• Full blocks + rising base fee = congestion sustained for hours or days.
• Full blocks but base fee stable = capacity is perfectly matched (rare).
• Empty blocks suddenly filling = a new wave of demand is hitting.

Ethereum’s EIP-1559 introduced a base fee that adjusts per block target utilization (50%). If blocks are consistently above target, base fees increase. This makes block utilization a self-reinforcing predictor: high utilization today forces fees up tomorrow.

“Base fee is a direct function of how full the last few blocks were. If you see six consecutive blocks at 100% utilization, the base fee will have roughly doubled—and the next wave of users will pay even more.”

Building a Simple Forecasting Model

You don’t need a PhD to turn these metrics into a usable forecast. The following heuristic, combining mempool, percentiles, and utilization, can give you a reliable congestion score.

Create three thresholds:

1. Mempool Pressure: Pending txns above 100k? +1 point. Above 200k? +2. Growth rate >10% per minute? +1 extra.
2. Percentile Gauge: 95th percentile above 150 gwei? +1. 50th percentile above 60 gwei? +1. Spread between 95th and 5th >100 gwei? +1 (stress).
3. Block Utilization: Average utilization over last 20 blocks >90%? +2. Base fee rising? +1.

A score of 0-2 indicates low congestion—safe to transact with low fee settings. 3-5 means moderate—consider waiting or using a medium priority. 6+ signals heavy congestion—delay non-urgent transactions or move to an L2.

This model is intentionally simple. You can refine it with weighted averages or integrate it into a Telegram bot that alerts you when conditions shift. The key is consistency: track the same metrics at the same times daily to learn your chain’s rhythm.

Tools for Real-Time Gas Analysis

Manual monitoring is possible but tedious. Several platforms aggregate the metrics we’ve discussed into dashboards:

• Etherscan Gas Tracker — provides percentile-based “Low/Average/High” estimates, plus a mempool size chart. Good for quick checks.
• Blocknative Gas Platform — shows real-time mempool by fee tier, plus historical fee curves. Offers API for automation.
• Dune Analytics — user-created dashboards like “Ethereum Gas Market” plot block utilization, base fee history, and mempool depth over time. Highly customizable.
• GasGuzzler — tracks the top gas-consuming contracts (often NFT mints or DeFi launches) and their fee impact.
• MEV Blocker / Flashbots protect — while not a forecasting tool, integrating MEV protection can prevent you from overpaying during sandwich attacks.

For the hands-on reader, creating a custom Dune dashboard with your chosen metrics is the most powerful approach. Many open-source queries exist; clone them and set email alerts for threshold breaches. This transforms passive observation into active fee management.

Timing Your Transactions Like a Pro

Forecasting is only useful if it changes behavior. Here are concrete timing strategies derived from on-chain patterns:

Also consider using EIP-1559’s max fee + priority fee (tip) correctly. Set max fee to current base fee + tip + a 20% buffer for base fee increases. Many wallets auto-calculate, but if you manually set too tight, your txn may get stuck during a base fee rise.

Finally, if you see a pattern of extreme congestion every Thursday at 6 PM (a routine weekly event like a major game launch), schedule your transactions around it. Over time, you build a personal fee calendar.

Conclusion: The Future of Fee Forecasting

Gas forecasting is becoming a competitive advantage. As Ethereum moves toward Danksharding and more L2s proliferate, fee patterns will evolve—but the core dynamics of mempool, percentiles, and utilization will remain relevant. Already, specialized bots use statistical and machine-learning models to anticipate fee conditions over the coming blocks.

For the intermediate user, the path forward is clear: automate data collection, set alerts, and build models like the one described here. Start with simple heuristic scoring, then graduate to regression on base fee changes. The skill of reading on-chain congestion will separate those who participate efficiently from those who pay the congestion tax.

Remember: fees are not a lottery; they are a market. Learn the signals, and you can always transact at a fair price.

Frequently asked questions

Related reading

• Liquidity Pool
• How To Verify Token Liquidity Lock 2026 Guide