Block Catalog
113 blocks across 6 categories.
model-head
AdaBoost Regressor
Frame → ModelArtifact
Adaptive boosting of weak learners weighted by residual magnitude.
try in builder 0 pipelines
eval-metric
ATS Percentage
CountsInput → Metric
Against-the-spread cover rate: wins / (wins + losses).
try in builder 0 pipelines
eval-metric
Average Edge
BetSlate → Metric
Mean of the edge column across all applied bets.
try in builder 0 pipelines
ops
Backfill
Frame → Frame
Backfill a source over a date range. Browser preview processes the first 100 rows and reports the total; server processes the full range.
try in builder 0 pipelines
bet-sizing
Bankroll-Aware Kelly
Predictions → BetSlate
Kelly scaled by current bankroll relative to a target: shrink stakes during drawdown, expand during streak.
try in builder 0 pipelines
model-head
Bayesian MLP / MC-Dropout (TF.js)
Frame → ModelArtifact
Bayesian neural network approximation via Monte Carlo Dropout. Yields both a point estimate and a calibrated uncertainty — high-variance predictions can be sized down or skipped. Toy task: noisy sine regression.
try in builder 0 pipelines
ensemble
Bayesian Model Average
Predictions[] → Predictions
Weight each base model by posterior probability proportional to exp(−loss).
try in builder 0 pipelines
feature-transform
Bayesian Ridge Features
Frame → Frame
Run BayesianRidge on engineered features and emit posterior-mean predictions as a derived signal column.
try in builder 0 pipelines
calibration
Beta Calibration
Predictions → Predictions
Calibrates probabilities via a beta-distribution warp; often generalizes better than Platt for skewed score distributions.
try in builder 0 pipelines
model-head
BNN MC Dropout (TF.js)
Frame → ModelArtifact
Bayesian NN via MC dropout: variance from N stochastic forward passes gives a calibrated CI.
try in builder 0 pipelines
model-head
Bradley-Terry (pairwise)
Frame → ModelArtifact
Browser-only pure-JS Bradley-Terry rater fit by Minorization-Maximization. Estimates latent multiplicative strengths so that P(i beats j) = π_i / (π_i + π_j).
try in builder 0 pipelines
eval-metric
Brier Score
PredictionsWithActuals → Metric
Mean squared error between predicted probability and binary outcome.
try in builder 0 pipelines
ops
Cache
Frame → Frame
Cache the inbound body by key/content-hash with a TTL. Browser uses an in-memory map; server reads/writes the run_cache store via ctx.cache.
try in builder 0 pipelines
model-head
CatBoost Classifier
Frame → ModelArtifact
Ordered-boosting classifier built for categorical features; strong out-of-the-box without tuning.
try in builder 0 pipelines
eval-metric
Closing Line Cents
ClvCentsInput → Metric
CLV in American-odds cents: priceAtPick − priceAtClose.
try in builder 0 pipelines
eval-metric
Closing Line Value
ClvInput → Metric
CLV in points: (pickLine - closingLine) × side. Proxy for true edge.
try in builder 0 pipelines
ops
Conditional
Frame → Frame
Route a pipeline by a predicate. Evaluates once and takes exactly one branch (the other is never evaluated). Behaves identically in browser and server.
try in builder 0 pipelines
bet-sizing
Confidence-Weighted
Predictions → BetSlate
Stake scales with prediction confidence: stake = bankroll × scale × |p − 0.5|. No-edge-aware.
try in builder 0 pipelines
ensemble
Confidence-Weighted Blend
Predictions[] → Predictions
Each model contributes proportional to |p − 0.5|. Confident models dominate the blend, uncertain ones drop out.
try in builder 0 pipelines
calibration
Conformal Bias Correction
Predictions → Predictions
Shift predicted probability by the residual-quantile bias measured on calibration set. Distribution-free.
try in builder 0 pipelines
feature-transform
Conformal Prediction Band
Predictions → Predictions
Empirical residual quantile for 90% coverage prediction interval.
try in builder 0 pipelines
model-head
Constant Baseline
Frame → ModelArtifact
Predicts the training-set mean (or a configured constant) for every row. Sanity check: anything that beats this is doing real work.
try in builder 0 pipelines
feature-transform
Correlation Pruning
Frame → Frame
Drops features whose pairwise |correlation| exceeds threshold (default 0.95).
try in builder 0 pipelines
feature-transform
Cover Probability
Frame → Frame
Computes binary cover label and spread-derived signals (f_spread_line, f_is_big_favorite, f_is_pick).
try in builder 0 pipelines
model-head
Decision Tree
Frame → ModelArtifact
Browser-only axis-aligned binary decision tree fit via greedy Gini-impurity splits, max-depth + min-samples stopping. A single interpretable tree.
try in builder 0 pipelines
ensemble
Diversity-Weighted Blend
Predictions[] → Predictions
Reduce weight on highly-correlated base models: w_i ∝ 1 / (1 + mean correlation with others).
try in builder 0 pipelines
ops
Drift Monitor
Frame → Frame
Monitor feature drift for a model. Browser preview measures drift on a 200-row sample; server measures over the full history. Flags when drift exceeds the threshold.
try in builder 0 pipelines
eval-metric
Edge Persistence
EdgeSeries → Metric
Correlation between edge in window t and edge in window t+1. Measures whether edges decay over time.
try in builder 0 pipelines
bet-sizing
Edge-Weighted Linear
Predictions → BetSlate
Stake proportional to edge: stake = bankroll × scale × max(edge, 0).
try in builder 0 pipelines
model-head
Elastic Net
Frame → ModelArtifact
Linear model with L1+L2 penalty; balances Lasso and Ridge for sparse, stable fits.
try in builder 0 pipelines
feature-transform
Elo Power Ratings
Frame → Frame
Maintains team Elo ratings and exposes home/away rating pair as features.
try in builder 0 pipelines
eval-metric
Expected Calibration Error
PredictionsWithActuals → Metric
Mean per-bin gap between predicted probability and observed frequency, weighted by bin size.
try in builder 0 pipelines
feature-transform
Exponential Moving Average
Frame → Frame
Compute an exponentially weighted moving average of a numeric column, optionally partitioned by a group key and ordered by a sortable key. Pure, leakage-safe; the EMA at row t uses only rows ≤ t within its group.
try in builder 0 pipelines
model-head
Extra Trees Classifier
Frame → ModelArtifact
Ensemble of extremely randomized trees; cheaper variance reduction than Random Forest.
try in builder 0 pipelines
bet-sizing
Fixed Unit
Predictions → BetSlate
Edge-independent flat stake (default 1% of bankroll).
try in builder 0 pipelines
bet-sizing
Flat With Edge Gate
Predictions → BetSlate
All-or-nothing: bet a fixed unit only when edge ≥ threshold; otherwise skip. Simple, transparent.
try in builder 0 pipelines
bet-sizing
Fractional Kelly
Predictions → BetSlate
0.25× Kelly stake by default; lower variance than full Kelly.
try in builder 0 pipelines
bet-sizing
Full Kelly
Predictions → BetSlate
Full Kelly stake: bankroll × (b·p - q)/b. Maximizes log growth; high variance.
try in builder 0 pipelines
feature-transform
Games Tabular (TF.js)
Frame → Frame
TF.js-compatible tabular feature transformer for browser-side use.
try in builder 0 pipelines
model-head
GAT Matchup (TF.js)
Frame → ModelArtifact
Graph Attention Network for team matchups. Each game is a 2-node graph (home, away); multi-head attention learns pairwise interaction weights between team features. Toy task: predict home margin from two team feature vectors with an asymmetric interaction term.
try in builder 0 pipelines
ensemble
Geometric Mean
Predictions[] → Predictions
exp(mean(log(p))). Pulls down disagreement; preserves zero. Robust when one model is wildly wrong.
try in builder 0 pipelines
model-head
GNN Matchup
Frame → ModelArtifact
Graph neural network over a team-vs-team graph; captures schedule-strength + transitive matchup effects.
try in builder 0 pipelines
model-head
Gradient Boosting Regressor
Frame → ModelArtifact
sklearn GradientBoostingRegressor: sequential weak learners minimizing residuals; robust default.
try in builder 0 pipelines
model-head
GRU Sequence (TF.js)
Frame → ModelArtifact
Browser-only gated recurrent unit for sequence regression. Lighter and faster than LSTM with comparable accuracy on short-window time series.
try in builder 0 pipelines
bet-sizing
Half Kelly
Predictions → BetSlate
0.5× Kelly stake. Roughly halves variance vs full Kelly for ~75% of the growth.
try in builder 0 pipelines
calibration
Histogram Binning
Predictions → Predictions
Binned calibration: each predicted probability is replaced with the empirical outcome rate in its bin.
try in builder 0 pipelines
eval-metric
Hit Rate by Confidence
PredictionsWithActuals → Metric
Cover% within high-confidence bucket only (default p > 0.6 or p < 0.4). Tells you whether confident picks are reliable.
try in builder 0 pipelines
feature-transform
Home/Away Split
Frame → Frame
Separates a stat into home-only and away-only rolling averages (capturing venue-dependent skill).
try in builder 0 pipelines
calibration
Identity Calibration
Predictions → Predictions
Pass-through. Use when downstream consumers must always run a calibration block.
try in builder 0 pipelines
calibration
Isotonic Regression
Predictions → Predictions
Non-parametric monotone calibration via pool-adjacent-violators.
try in builder 0 pipelines
eval-metric
Kelly ROI
BetResults → Metric
ROI computed against Kelly-sized stakes (not flat units). Reflects compounding when staking varies.
try in builder 0 pipelines
model-head
KNN Classifier
Frame → ModelArtifact
Browser-only k-nearest-neighbors classifier. Stores the training set, predicts via majority vote of the k closest rows. Useful when the decision surface is local and non-linear.
try in builder 0 pipelines
feature-transform
L1 Feature Selection
Frame → Frame
LassoCV-based feature selection on 70% train; keeps features with |coef| > threshold.
try in builder 0 pipelines
model-head
LightGBM Classifier
Frame → ModelArtifact
Histogram-based gradient boosting; fastest of the boosters on large tabular data with categorical features.
try in builder 0 pipelines
model-head
Linear Regression
Frame → ModelArtifact
Ordinary least squares; transparent baseline for spread/total margin prediction.
try in builder 0 pipelines
feature-transform
Linear Regression Features
Frame → Frame
OLS-fit predictions of a target column from a feature subset; appended as f_lr_predicted.
try in builder 0 pipelines
eval-metric
Log Loss
PredictionsWithActuals → Metric
-mean(y·log(p) + (1-y)·log(1-p)). Penalizes overconfident wrong predictions.
try in builder 0 pipelines
model-head
Logistic Regression
Frame → ModelArtifact
Linear classifier with L2 regularization; baseline for binary outcomes.
try in builder 0 pipelines
eval-metric
Longest Win Streak
BetResults → Metric
Maximum consecutive wins in the result series.
try in builder 0 pipelines
eval-metric
Loss Streak
PredictionsWithActuals → Metric
Longest run of consecutive losing picks, ordered by input position. Useful for stress-testing bankroll resilience and detecting bad-luck clusters.
try in builder 0 pipelines
model-head
LSTM Sequence (TF.js)
Frame → ModelArtifact
Stateful sequence model trained client-side via TF.js. Toy task: next-step sine-wave prediction. Production task: PBP / drive-level sequence regression.
try in builder 0 pipelines
feature-transform
Market Efficiency
Frame → Frame
Spread vs. consensus arbitrage signals and line movement features.
try in builder 0 pipelines
model-head
Massey Ratings
Frame → ModelArtifact
Browser-only pure-JS Massey rating system. Solves the linear least-squares system MᵀMr = Mᵀb via Cholesky factorization. Output: per-team additive rating recovered from observed game margins.
try in builder 0 pipelines
eval-metric
Max Drawdown
BankrollHistory → Metric
Largest peak-to-trough drop in bankroll. Wraps analytics2/metrics/formulas.drawdown.
try in builder 0 pipelines
eval-metric
Maximum Calibration Error
PredictionsWithActuals → Metric
Worst-bin gap between predicted probability and empirical accuracy across N bins.
try in builder 0 pipelines
ensemble
Median Combiner
Predictions[] → Predictions
Per-row median of base model predictions. Outlier-resistant; non-differentiable.
try in builder 0 pipelines
feature-transform
Min-Max Scale
Frame → Frame
Scale a column to [0, 1]: (x − min) / (max − min). Adds <col>_mm to the frame.
try in builder 0 pipelines
feature-transform
Minimal Signal
Frame → Frame
Sparse feature extraction for low-data regimes; defaults safe in cold start.
try in builder 0 pipelines
model-head
Mixture Density Network (K=3, TF.js)
Frame → ModelArtifact
Browser-only MDN. Predicts a 3-component Gaussian mixture (π, μ, σ) per row via custom log-sum-exp NLL loss. Handles bimodal / boom-or-bust outcomes that a single-Gaussian regressor must average over.
try in builder 0 pipelines
feature-transform
MLB Feature Store
Frame → Frame
MLB team-level rolling runs, OPS, ERA, ballpark factor; schedule-aligned for game prediction.
try in builder 0 pipelines
feature-transform
MLB Player Feature Store
Frame → Frame
MLB PA/AB, LHP/RHP splits, ballpark factor, 10-game rolling stats.
try in builder 0 pipelines
model-head
MLP Classifier (TF.js)
Frame → ModelArtifact
Browser-runnable multi-layer perceptron via TensorFlow.js.
try in builder 0 pipelines
model-head
Multi-Task (TF.js)
Frame → ModelArtifact
Browser-only multi-task head with a shared dense trunk and three specialised heads (cover / total / moneyline) trained jointly via weighted-sum loss. Exploits target correlation when the same game produces all three lines.
try in builder 0 pipelines
model-head
Naive Bayes (Gaussian)
Frame → ModelArtifact
Browser-only Gaussian Naive Bayes. Estimates per-class mean and variance per feature, predicts argmax of log P(x|c) + log P(c). Cheap and fast when features are approximately Gaussian within each class.
try in builder 0 pipelines
feature-transform
NBA Feature Store
Frame → Frame
NBA team-level rolling stats, pace/eff splits, schedule-alignment features (game-level, not player).
try in builder 0 pipelines
feature-transform
NBA Player Feature Store
Frame → Frame
NBA per-player minutes, scoring avgs, team pace, position norms.
try in builder 0 pipelines
feature-transform
NFL Player Feature Store
Frame → Frame
Per-player rolling stats, opponent strength, snap counts.
try in builder 0 pipelines
feature-transform
NHL Player Feature Store
Frame → Frame
NHL TOI, PP/PK minutes, line matchups, goalie pairing.
try in builder 0 pipelines
ops
Notify
Frame → Frame
Send a notification. Browser preview fires a UI toast only; server delivers via the real channel (Slack/email/toast).
try in builder 0 pipelines
feature-transform
One-Hot Encode
Frame → Frame
Expand a categorical column into k binary indicator columns named <col>__<level>.
try in builder 0 pipelines
feature-transform
Opponent-Adjusted
Frame → Frame
Subtracts opponent baseline from raw stat: f_adj = f - mean(f | opponent). Surfaces matchup-specific edge.
try in builder 0 pipelines
feature-transform
PBP Sequence for Transformer
Frame → Frame
Play-by-play windowing for attention/transformer model inputs.
try in builder 0 pipelines
model-head
Plackett-Luce (ranked outcomes)
Frame → ModelArtifact
Browser-only pure-JS Plackett-Luce rater for full rankings. Fit by Hunter (2004) MM updates. Output: per-item log-strengths that imply the probability of any rank order — useful for DFS lineups and tournament finish predictions.
try in builder 0 pipelines
calibration
Platt Scaling
Predictions → Predictions
Two-parameter logistic-regression calibration; classic Platt 1999.
try in builder 0 pipelines
model-head
Poisson Regression (TF.js)
Frame → ModelArtifact
Browser-only generalized linear model with log link and custom Poisson NLL loss. λ = exp(W·x + b). Built for non-negative count outcomes like team goals or play counts.
try in builder 0 pipelines
eval-metric
Profit Factor
BetResults → Metric
Gross wins / gross losses. >1 is profitable; >2 is rare and good.
try in builder 0 pipelines
model-head
Prophet Trend
Frame → ModelArtifact
Facebook Prophet time-series model: trend + weekly seasonality. Use for game-tempo or scoring drift.
try in builder 0 pipelines
model-head
Quantile Regression (TF.js)
Frame → ModelArtifact
Browser-only quantile regression with three heads (Q10 / Q50 / Q90) trained via pinball loss. Outputs full per-row interval — predicted lower / median / upper.
try in builder 0 pipelines
model-head
Random Forest
Frame → ModelArtifact
Ensemble of decision trees with bagging; robust default for tabular data.
try in builder 0 pipelines
ensemble
Rank Averaging
Predictions[] → Predictions
Combines multiple model outputs by averaging per-row ranks; robust to scale differences across models.
try in builder 0 pipelines
feature-transform
Rest Days
Frame → Frame
Days between consecutive games for each team. Adds f_home_rest, f_away_rest, f_rest_advantage (home − away).
try in builder 0 pipelines
ops
Retrain
Frame → Frame
Retrain a model on a dataset. Browser preview reports what would be retrained and on how many rows; server enqueues the job on the existing nflbets-training app.
try in builder 0 pipelines
model-head
Ridge Regression
Frame → ModelArtifact
L2-penalized linear regression; stable on collinear features.
try in builder 0 pipelines
eval-metric
ROI at -110
CountsInput → Metric
Standard sportsbook ROI: (wins × 0.909 - losses) / (wins + losses + pushes).
try in builder 0 pipelines
feature-transform
Rolling Average Window
Frame → Frame
Computes a rolling average over the last N rows for a given column.
try in builder 0 pipelines
ops
Schedule
Frame → Frame
Schedule a target job on a cron/interval. Browser preview fires the target once now and shows the next-run stub; server registers the recurring trigger.
try in builder 0 pipelines
model-head
Set Transformer (TF.js)
Frame → ModelArtifact
Permutation-invariant set encoder. Element-wise embed + self-attention + mean-pool. Toy task: sum-of-set binary classification on N=10 reals.
try in builder 0 pipelines
eval-metric
Sharpe Ratio
Returns → Metric
Mean return / std-dev of returns. Higher = more reward per unit volatility.
try in builder 0 pipelines
feature-transform
Shrinkage to Mean
Frame → Frame
Empirical-Bayes shrinkage: stabilizes small-sample estimates by blending toward the global mean with weight 1/(n+k).
try in builder 0 pipelines
calibration
Sigmoid Stretch
Predictions → Predictions
Reshape predicted probabilities through a temperature-controlled sigmoid: pulls extremes toward 0.5 or pushes them out.
try in builder 0 pipelines
feature-transform
Sport Feature Store
Frame → Frame
Cross-sport ETL: rolling avgs, exponential moving avgs, schedule alignment.
try in builder 0 pipelines
ensemble
Stacking
Predictions[] → Predictions
Meta-model trained on base model predictions; learns optimal blend per row.
try in builder 0 pipelines
model-head
TabNet (TF.js)
Frame → ModelArtifact
TabNet applies sequential soft attention to select which input features are most relevant at each step. Interpretable attention masks reveal which features the model used.
try in builder 0 pipelines
calibration
Temperature Scaling
Predictions → Predictions
Single-parameter calibration: divides logits by T. T>1 softens overconfidence; T<1 sharpens.
try in builder 0 pipelines
model-head
Temporal Convolutional Network (TF.js)
Frame → ModelArtifact
Browser-only causal dilated conv1d stack (dilations 1·2·4). Receptive field grows exponentially, so the model sees lag-k dependencies an RNN of equal depth would miss.
try in builder 0 pipelines
ensemble
Top-K Voting
Predictions[] → Predictions
Keep only the top K base models by training-set metric; average their predictions. Drops weak voters.
try in builder 0 pipelines
model-head
Transformer Encoder (TF.js)
Frame → ModelArtifact
Encoder-only transformer for PBP / drive-level sequences. Stacks multi-head self-attention + FFN with residual connections and layer normalization, mean-pools the sequence.
try in builder 0 pipelines
bet-sizing
Variance-Capped Kelly
Predictions → BetSlate
Fractional Kelly with hard cap at maxBetPct (default 5%) of bankroll.
try in builder 0 pipelines
feature-transform
Vegas Line Movement
Frame → Frame
Open-to-close spread/total delta and implied-probability shift. Positive delta = home line moved toward favorite.
try in builder 0 pipelines
ops
Webhook
Frame → Frame
POST a JSON payload to a URL. Browser preview echoes the payload without sending; server POSTs for real (https-only, SSRF-guarded).
try in builder 0 pipelines
ensemble
Weighted Average
Predictions[] → Predictions
Linear combination of model outputs with explicit weights, normalized by sum |w|.
try in builder 0 pipelines
