Exploration Exploitation Ratio Calculator

Work out exploration exploitation ratio instantly with clear inputs, formula shown and shareable results.

About Exploration Exploitation Ratio Calculator

Epsilon-greedy picks a random action with probability epsilon and the current best action otherwise. Linear annealing interpolates epsilon from its start to its floor over a fixed number of steps, so early training explores almost entirely and late training exploits almost entirely. A non-zero floor is kept deliberately so the agent keeps sampling alternatives in a changing environment.

Formula

Linear epsilon decay

epsilon(t) = start + (end - start) x min(1, t / decay steps); greedy share = 1 - epsilon

Worked example

With the default values below, Exploration Exploitation Ratio Calculator returns Current epsilon of 0.76250. Change any input above and every figure here is recalculated in your browser.

Inputs used

Starting epsilonaccepts 0.001–1: 1
Final epsilonaccepts 0–1: 0.05
Steps over which epsilon decaysminimum 1: 100,000
Current stepminimum 0: 25,000

Results

Current epsilon: 0.76250
Random actions taken so far: 22,031
Share of actions now greedy: 23.75%
Steps until epsilon reaches its floor: 75,000

Step by step

1Values used
2Linear epsilon decay
3Current epsilon= 0.76250
4Random actions taken so far= 22,031
5Share of actions now greedy= 23.75
6Steps until epsilon reaches its floor= 75,000

Assumptions and limits

The result is derived only from the 4 inputs listed above. Anything not among them is not part of the calculation.
In particular, this calculator does not take inflation, tax, fees or charges or compounding frequency into account.
Figures are rounded for display to 5 decimal places, so totals built from several rounded parts can differ slightly from the unrounded arithmetic.

How these calculators are built and tested is described on the Methodology page. If a figure looks wrong, please report it.

Frequently Asked Questions

Why keep a non-zero epsilon floor?

So the agent can still discover changes. With epsilon exactly zero the policy freezes and a shift in dynamics is never detected.

Are there better strategies than epsilon-greedy?

Yes for many settings. Upper-confidence-bound and Thompson sampling explore in proportion to uncertainty rather than uniformly at random, which reduces regret substantially.