Model Builder

Caution

The Model Builder requires a Sharp tier subscription. Free, Pro, and Elite users will see an upgrade prompt when accessing this tool.

The Model Builder lets you create custom scoring models by adjusting the weights of the 7 scoring blocks, preview how those changes affect real picks, and backtest your custom model against the default engine configuration. If you believe certain blocks should carry more or less influence, this tool lets you test that hypothesis with real data.

Note

The app interface is in Spanish. This guide references the actual labels you will see.

Overview

Navigate to “Herramientas” (Tools) in the sidebar and select “Constructor de Modelo” (Model Builder). The route is /tools/model-builder.

The page uses a two-column layout. The left column handles model management and weight adjustment. The right column shows a live preview of how your weights affect a specific pick, plus backtest results comparing your model against the default.

Weight Sliders

The left column displays 7 weight sliders, one for each scoring block. The default weights are:

Block	Default Weight	What It Evaluates
PlayerProfile	25%	Season averages, trends, consistency, minutes, usage, fatigue
Matchup	20%	Opponent DvP, defensive play types, pace, rim protection, H2H
GameContext	15%	Spread, total O/U, home/away, B2B, rival injuries, refs
MarketLine	15%	Line edge, line movement, implied probability, multi-book consensus
Synergy	10%	Teammate injury impact, on/off splits, usage shift
AnalysisQuality	10%	Sample size, hit rate, Wilson CI, mean/median consistency
ExternalSignals	5%	Tank01 projections, insider leads

Drag any slider to increase or decrease a block’s influence on the final score. The weights are relative — the engine normalizes them so they always sum to 100%.

Model Management

Above the sliders, a “Modelo:” (Model) dropdown lets you select a previously saved model or the default configuration “Predeterminado”. To save a custom model:

1. Adjust the weight sliders to your preferred configuration
2. Click “Guardar” (Save)
3. Enter a name in the “Nombre del modelo” (Model name) field
4. Click “Crear” (Create) to save, or “Cancelar” (Cancel) to discard

When a saved model is selected, two additional buttons appear: “Actualizar modelo” (Update model) to save changes to the existing model, and “Eliminar” (Delete) to remove it.

Preview and Backtest

The right column has two sections:

Pick Preview

A “Pick de vista previa:” (Preview pick) dropdown lists recent picks from the last 7 days (up to 50). Select a pick to see how your custom weights would score it compared to the default engine.

The preview displays:

• Spider chart — A radar chart showing the weighted scores from each of the 7 blocks using your custom weights
• Score badge — The calculated final score with color coding: green for 63 or above (PREMIUM), amber for 60-62 (STANDARD), and the numeric score displayed prominently

Backtest Results

Below the preview, the backtest section shows a side-by-side comparison of your custom model versus the default engine over historical data. This includes:

• Hit rate comparison between your model and the default
• Number of picks each model would flag as PREMIUM or STANDARD
• A quick visual showing whether your custom weights improve or degrade overall accuracy

A share button lets you export or share your model configuration.

Tips

Tip

Start with small adjustments. Moving one or two sliders by 5% at a time lets you isolate the impact of each change. Large sweeping changes make it hard to understand what improved or worsened results.

Tip

If you have strong convictions about a specific factor — for example, that matchup data matters more than the default 20% — raise Matchup while lowering a block you trust less, then backtest to see if the data supports your theory.

Tip

Pay attention to the backtest sample size. A model that backtests well over 10 days might not hold up over a full month. Use the broadest date range available for the most reliable comparison.

---

Next: return to the Dashboard to apply your insights, or check the Prop Evaluator to test specific picks with the default engine.