Using Bot Health Check to Audit Your Active Bots

It's easy to set up a bot, let it run, and stop thinking critically about whether it's still configured well. Markets change. Your understanding of a strategy deepens. A configuration that made sense three months ago might have edge cases you've since identified.

Relay's Bot Health Check is designed to interrupt that drift. It runs a diagnostic pass across your active bots, reviewing performance consistency, flagging configuration patterns worth examining, and surfacing anything that looks off. This guide walks through how to run it and how to use what it finds.

What bot health check does

The health check isn't just a performance summary. It reviews your bots across two distinct dimensions:

Configuration Health

• Are the stop loss and take profit levels consistent with the risk profile you intended?
• Are cooldown windows appropriate given the bot's typical trade frequency?
• Are position sizing settings still aligned with your current approach?
• Are there any configuration combinations that historically correlate with worse outcomes for this bot?

Performance Health

• Is the bot performing consistently, or has there been a meaningful shift in recent results?
• Is the win rate stable, or are there signs of drift?
• Are losing trades within expected range, or are any outliers pulling the overall numbers?
• Is the bot executing trades at the expected frequency, or has something changed in signal behavior?

The output combines both lenses into a unified assessment for each bot, with notes on anything that warrants closer attention.

Running the health check

From the capability grid: Click "Bot Health Check" from Relay's capability menu. Relay will run the check across all your active bots by default.

Through conversation:

"Run a health check on all my bots." "Check the health of my SPY bots." "Is my QQQ UT bot healthy?"

You can scope the check to a single bot, a subset of bots by symbol, or your entire portfolio. If you're doing a focused review before a configuration change, running the check on just the bot you're about to modify is a good practice.

Reading the output

The health check report gives each bot an overall assessment along with specific observations. Here's what to pay attention to:

Green flags When a bot's configuration and performance are consistent and within expected ranges, the health check confirms that. This is useful information, it tells you the bot is doing what it's supposed to do and doesn't need intervention.

Observations to consider These are patterns Relay flags for your awareness without necessarily recommending action. Examples:

• "Win rate has declined slightly over the last 30 days compared to the prior period." Not necessarily a problem, but worth monitoring.
• "Average hold time has increased. Exits may be taking longer than intended." This can indicate changed market conditions or an exit configuration that's less efficient than it was.

Items worth investigating These are more specific flags that suggest something may be off and warrants deeper review. Examples:

• "Stop loss is being hit more frequently than the historical average for this bot." This can indicate the stop is too tight for current conditions, or that entry signals are triggering in less favorable conditions.
• "Last 15 trades have a significantly lower win rate than the prior 50." A meaningful shift in recent performance that deserves examination before assuming it's random noise.

Acting on the findings

The health check surfaces information, what you do with it is up to you. Here's a framework for working through the findings:

For configuration observations: Use What-If Simulation to model whether a configuration adjustment would have changed outcomes on recent trades. Don't change settings based on the health check observation alone. Use the simulation to evaluate the change first.

For performance drift: Look at the trade history for the period where performance shifted. Was there a change in market conditions that week? Did the bot's signal source change? Use Trade Replay on some of the underperforming trades to see what happened at the execution level.

For execution frequency changes: Check whether your signal source has changed behavior, or whether entry conditions have become harder to satisfy in recent market conditions.

For configuration questions: Ask Relay directly

"Why is the stop loss being hit more on this bot recently?" "Is the cooldown period on my QQQ UT bot appropriate given how often it's been trading?"

Relay can often surface additional context that helps you decide whether the observation warrants action.

Cadence: how often to run it

The health check is most useful when run on a regular schedule, not just when something feels wrong. A good cadence:

• Weekly — as part of your regular weekly review cycle, run a health check after reviewing overall performance
• Before configuration changes — always run a health check before modifying a live bot to establish a baseline
• After unusual market periods — high-volatility events, extended low-volume periods, or anything that changed conditions significantly

The goal is to catch configuration drift and performance shifts early, before they compound into larger issues. Regular health checks build that early warning into your workflow.