Variance Lab: Four YaeBot Tools for Risk and Evidence

Variance Lab is YaeBot’s analytics product for risk, edge, sample-size uncertainty, and statistical confidence. It focuses on one practical question: does this strategy have enough statistical support, and can I survive the variance on the way there?

Instead of presenting one attractive equity curve as destiny, Variance Lab gives you four ways to pressure-test a strategy: two forward-looking stress tests, and two backward-looking reality checks.

Two Modes, Four Tools

Strategy Stress Tester

Forward-looking tools for planning position size and edge resilience before you deploy or scale a strategy.

Drawdown Stress Test
Edge Decay Test

Edge Reality Check

Backward-looking tools for auditing observed results from a backtest, paper-trading period, or live sample.

Sample Size Reality
Luck Horizon

Use the forward-looking mode when you have a strategy hypothesis and want to understand the path it may force you to endure. Use the backward-looking mode when you already have results and need to know whether they are meaningful or still dominated by sampling noise.

How to Use Variance Lab

Open the GoemonYae Discord server.
Go to 🦊・yaebot-commands, or use the commands invite if you need access.
Type /yaebot.
Select Variance Lab from the product selector.
Choose one of the four tools.
Fill in the modal fields and submit the job.
Review the result chart, headline metrics, warnings, and assumptions.

Use /mytier in Discord to see the exact limits currently attached to your role. Public documentation describes the tier logic qualitatively because account limits may change as the service evolves.

Drawdown Stress Test

Question answered: “What depth of drawdown should I be prepared to tolerate if this strategy behaves like my inputs suggest?”

Drawdown Stress Test simulates many possible trade sequences from your strategy assumptions, then focuses on maximum drawdown rather than headline profit. That makes it useful before increasing position size or committing to a strategy through a live drawdown.

Inputs

Typical inputs include:

theoretical win rate;
theoretical profit factor;
initial capital;
trade size or risk per trade;
number of trades to simulate.

Use conservative backtest statistics. If live execution usually includes slippage, commissions, or missed fills, reflect that in the inputs rather than relying on the cleanest historical result.

Output

The result highlights drawdown percentiles and usually centres interpretation around practical questions:

How deep is the median maximum drawdown?
What drawdown appears in worse-but-plausible scenarios?
Would that drawdown be psychologically and financially tolerable?
Should position size be reduced before deployment?

Edge Decay Test

Question answered: “How much does the median outcome shift if the win rate degrades?”

Markets are not stationary. A pattern that worked in one regime can weaken as volatility changes, competition finds it, or execution quality declines. Edge Decay Test compares your baseline assumptions with a degraded scenario so you can see whether a modest erosion turns the strategy unattractive.

Inputs

Edge Decay Test uses the same family of forward-looking inputs as Drawdown Stress Test:

theoretical win rate;
theoretical profit factor;
initial capital;
trade size;
number of trades.

The tool applies a defined degradation scenario and compares it with the baseline. Treat the degradation as a stress scenario, not as a promise about exactly how future markets will behave.

Output

The result compares the baseline and stressed equity cones and reports the gap between their median (P50) terminal-equity outcomes as Terminal Opportunity Divergence — how much median terminal equity you forgo if the win rate degrades by the fixed stress. A larger divergence means the median outcome is more sensitive to a weakening edge. The tool does not judge whether the decayed edge remains viable; that reading is yours.

Use it:

before scaling a strategy;
after a material market-regime change;
when live results are beginning to lag the backtest;
as part of a quarterly strategy review.

Sample Size Reality

Question answered: “How much uncertainty surrounds my observed win rate?”

Sample Size Reality is for actual results: a backtest sample, a paper-trading sample, or live trades. It treats your observed win rate as a noisy estimate rather than a permanent truth.

The tool uses a confidence-interval approach for binomial proportions, with Wilson Score intervals referenced in the implementation. The important interpretation is simple: small samples create wide uncertainty bands, so a 60% observed win rate over a small number of trades may still be compatible with a much lower true win rate.

Inputs

Typical inputs include:

observed win rate;
sample size;
observed profit factor;
capital and trade-size assumptions for projection context;
number of future trades to illustrate.

Output

Sample Size Reality compares optimistic, nominal, and conservative interpretations of the observed result. The conservative case deserves the most attention because it asks whether the strategy still looks acceptable if your sample has overstated the true edge.

Use it when:

your first live trades look promising but the sample is small;
you are considering increasing position size;
a backtest has a limited number of trades;
you need to explain why more evidence is required.

Luck Horizon

Question answered: “How many trades do I need before profitability is statistically credible?”

Luck Horizon estimates the trade count required for the confidence interval’s lower bound to clear the strategy’s break-even threshold. It is a patience tool: it helps you avoid both overcommitting to early luck and abandoning an edge before enough evidence exists.

Inputs

Luck Horizon uses observed statistics:

observed win rate;
observed profit factor;
current sample size;
optional projection context such as capital and trade size.

Profit factor determines the break-even win rate. The closer your observed win rate sits to that break-even threshold, the more trades you usually need before the lower bound clears it.

Output

Luck Horizon reports whether your current sample is already statistically credible under the tool’s assumptions, or how many more trades are needed. It also helps translate that number into a practical timeline based on how frequently the strategy trades.

Use it:

before scaling capital;
when deciding whether a strategy deserves more data collection;
during drawdowns, when impatience can lead to premature abandonment;
when comparing which strategy to prioritise for live validation.

Recommended Workflows

Before deploying a new strategy

Run Edge Decay Test to see how far the median outcome shifts if the win rate weakens.
Run Drawdown Stress Test to size the strategy around a drawdown you can actually endure.
Deploy conservatively if both tests remain acceptable.
After enough live or paper trades, audit the sample with Sample Size Reality.
Use Luck Horizon to decide whether you can scale, must wait, or should redirect effort.

When auditing existing results

Start with Sample Size Reality to reveal confidence-interval width.
Use Luck Horizon to calculate the remaining evidence required.
If the sample is credible, use the forward-looking tools to plan sizing and stress-test the edge.
If the sample is not credible, keep trading at minimal size or collect better data before making capital decisions.

Feature Comparison

Tool	Mode	Main question	Best used for
Drawdown Stress Test	Forward-looking	”What drawdown might I need to endure?”	Position sizing and psychological preparation
Edge Decay Test	Forward-looking	”How much does performance change if the edge weakens?”	Gauging edge sensitivity before scaling
Sample Size Reality	Backward-looking	”How uncertain is this observed result?”	Auditing small backtests or early live samples
Luck Horizon	Backward-looking	”How many trades until I can trust this result?”	Validation planning and patience calibration

Common Questions

Is Potential Path Modeller the current YaeBot product?

No. Variance Lab is the public YaeBot analytics product. Potential Path Modeller and related internal names are not the public documentation model; Variance Lab splits the work into clearer tools for drawdown stress, edge decay, sample-size uncertainty, and significance timing.

Why avoid exact public simulation limits here?

The command interface is tier-aware, and /mytier is the source for the limits attached to your current role. Public docs describe the access model without freezing numbers that may change as account limits evolve.

Is a statistically credible strategy guaranteed to make money?

No. Statistical credibility only means the observed sample is less likely to be pure luck under the tool’s assumptions. Execution quality, fees, slippage, market regime change, and trader behaviour can still destroy live performance.

Where does Signal Sentinel fit?

Signal Sentinel is separate from Variance Lab. Variance Lab is interactive analysis through /yaebot; Signal Sentinel is the public 🦊・signal-sentinel feed for selected in-house entries and exits. See the Signal Sentinel guide for the feed format and access links.

Next Steps

YaeBot Overview Learn the Discord command workflow and how Variance Lab fits beside Signal Sentinel

YaeBot FAQ Access, troubleshooting, commands, tiers, and the live Signal Sentinel feed

Ready to Master Algorithmic Trading?

Join many traders using GYTS to refine their edge.

Choose Your Path