Selection Strategies¶

The library provides two selection strategies that control how much of each asset the algorithm selects relative to the minimum required.

SelectionStrategyOptimal¶

The default strategy. For each asset, the algorithm targets approximately twice the minimum required amount:

\[ \text{target}(a) = 2 \times \text{minimum}(a) \]

This means that after paying for the required outputs, the remaining value is returned as change that is roughly the same size as the original outputs.

When to use

Use for most transactions. This strategy helps the wallet's UTxO distribution evolve to match the user's typical payment patterns, increasing the likelihood that future selections succeed and lowering amortized transaction costs.

Example¶

If the user sends 100 ada, the algorithm targets selecting ~200 ada worth of inputs, producing ~100 ada in change. This change output is similar in size to the payment, maintaining a balanced UTxO set.

SelectionStrategyMinimal¶

Selects just enough of each asset to meet the minimum required:

\[ \text{target}(a) = 1 \times \text{minimum}(a) \]

The selection terminates as soon as the minimum is covered.

When to use

Use only as a fallback when the optimal strategy fails. Regular use of the minimal strategy leads to small, fragmented change outputs that degrade the wallet's UTxO set over time.

Example¶

If the user sends 100 ada, the algorithm selects ~100 ada of inputs, producing very little change. The small change outputs are less useful for future transactions.

The Round-Robin algorithm¶

Both strategies share the same underlying Random-Round-Robin algorithm. The algorithm processes each asset in turn, selecting UTxOs randomly until all assets reach their target.

flowchart TD
    A[Build selector for each asset] --> B[Reverse list, ada selector last]
    B --> C{Any selector can improve?}
    C -->|Yes| D[Run next selector]
    D --> E{Improvement?}
    E -->|Yes| F[Keep selector in queue]
    E -->|No| G[Remove selector from queue]
    F --> C
    G --> C
    C -->|No| H[Selection complete]

Selection step logic¶

For a given asset, a single selection step works as follows:

If the current selected quantity is below the minimum, select another UTxO containing that asset.
If the current selected quantity is at or above the minimum, select another UTxO only if it brings the total closer to the target (but not further away).

\[ \text{improvement}(s') = |q(s') - \text{target}| < |q(s) - \text{target}| \]

where \(q(s)\) is the selected quantity of the asset in state \(s\).

UTxO selection priority¶

When selecting a UTxO for a given asset \(a\), the algorithm tries these filters in order of priority:

Singleton: a UTxO containing only asset \(a\) (and ada)
Pair: a UTxO containing asset \(a\) and exactly one other asset
Any: any UTxO containing asset \(a\)

This priority order minimises "collateral damage" -- the selection of unwanted assets alongside the desired one.

Ada is selected last¶

The ada selector is run last in the round-robin. Since every UTxO necessarily contains ada, selecting UTxOs for other assets first increases the probability that the ada requirement is already satisfied without needing additional ada-only inputs.