Benchmarks · Flashbots · Polygon

Measured KPIs across FRB’s simulated routes

Every dashboard build ships with the same telemetry harness. We run nightly simulations across Ethereum, Polygon, and BNB to keep tabs on inclusion, latency, and gas overhead. Below is the public snapshot you can cite when asked “why FRB?”.

Flashbots inclusion (median)

92.1%

vs 78.4% for generic bots

Polygon relay latency

148 ms

FRB auto-pins to healthiest WSS endpoints

Gas / tx (weighted)

$1.84

Private bundles avoid PGA bidding wars

Headline KPIs vs generic bots

Metric	FRB (AI Agent)	Generic bots	Notes
Median inclusion rate (Flashbots/Mainnet)	92.1%	78.4%	Simulated bundles with identical gas caps, Nov 2025 window.
Polygon private bundle inclusion	87.3%	69.8%	FRB routes auto-rotate WSS endpoints and retry late slots.
Average latency (node → relay)	148 ms	226 ms	Node clients auto-pin to healthiest WSS, fallback on jitter >60 ms.
Gas / tx (weighted avg.)	$1.84	$2.73	Flashbots-only execution avoids public PGA bidding wars.
Day-one PnL ramp per $100 funded	$14.20	$6.70	Based on sandboxed contracts with identical budgets and risk guards.

Relay & builder benchmarks

These are the inputs we feed into the dashboard “Health” gauges.

Flashbots

92%

Variance window: ±4%

Default path for Ethereum backruns and protection routes.

Builder0x69 (Polygon)

87%

Variance window: ±6%

Best results when paired with FRB latency monitor + compact mode.

BSC Protect / Eden

73%

Variance window: ±9%

Still beta—FRB throttles bundle submissions when congestion spikes.

MEV-Share / SUAVE

65%

Variance window: ±12%

Great for collaborative order-flow; FRB tags these as “research” routes.

Methodology & data hygiene

Every benchmark stems from a repeatable telemetry harness. We replay identical bundles on multiple FRB releases, log relay selected, latency, refund behavior, and whether the attempt was simulation or live. Runs that deviate beyond tolerance trigger an investigation and a public note on/releases.

Latency samples include p50/p95 readings from Frankfurt, Singapore, and Virginia POPs.
Gas-per-tx figures already incorporate refunded slots so finance sees true cost.
Dataset rows ship with timestamps and hash references so you can rerun the job internally.

Want to validate? Pull the JSON, match the run ID to your own telemetry export, and flag any drift greater than 5%—we’ll help reproduce it.

Download the raw dataset

Need to plug these figures into your own pitch deck? Grab the JSON export straight from the telemetry job.

See it in the dashboard

Metrics refresh live inside `/app/dashboard`. Connect your wallet, pair one node client, and watch the gauges update in real time.

Launch dashboard

How to apply these numbers

Use inclusion deltas when pitching FRB vs legacy bots—cite the dataset and link to this page.
Feed latency targets into your WSS rotation policy; anything above the published medians should trigger an alert.
Cross-reference per-chain benchmarks (e.g., BNB metrics, Flashbots metrics) when tuning gas caps.

Remember: these are simulated yet reproducible snapshots. Always rerun on your own stack before committing changes.

For board decks, attach the JSON export, paste a screenshot of this page, and include your own commentary so reviewers see both the data and the narrative.