Skip to content
Take the guessing out of predicting
Loading
Loading the game…

—

Make a pick

Odds lock when you tap · no cancelling

Live game

Quant models: every model's record

Saabi runs a library of NBA prediction models, from Elo and Poisson to gradient boosting and Bayesian samplers. Each records its picks before tip-off; they're graded after the game, losses kept like wins. Here's every model with 20 or more graded picks, best return first.

ModelPicks onWon–lostReturn a pickBeat the close
Dynamic Bradley–Terry (on Saabi's chances)Games20-7+25.7%+0.3%
Dynamic Bradley–TerryGames19-7+23.8%-1.0%
ARMA (on Saabi's chances)Games21-9+21.1%+0.4%
ARMAGames21-9+21.0%+0.3%
TrueSkill (on Saabi's chances)Games19-7+19.7%+0.1%
TrueSkillGames19-7+19.5%+0.2%
Pair tradingPlayer props27-19+19.4%-0.6%
Oblivious-tree boosting (CatBoost-style)Games29-15+18.1%-0.7%
Dynamic Bayesian networks (on Saabi's chances)Player props23-13+17.9%-3.8%
Margin-of-victory EloGames29-18+16.0%+0.1%
Margin-of-victory Elo (on Saabi's chances)Games29-18+16.0%+0.8%
Geometric Brownian motion (on Saabi's chances)Player props142-86+15.7%-4.0%
Histogram boosting (LightGBM-style)Games28-17+15.2%-0.1%
Geometric Brownian motionPlayer props141-87+14.9%-3.9%
Massey ratingsGames17-7+14.7%-0.9%
Oblivious-tree boosting (CatBoost-style) (on Saabi's chances)Games28-16+14.0%+1.5%
Histogram boosting (LightGBM-style) (on Saabi's chances)Games27-17+13.7%+2.1%
Bayesian networksPlayer props22-14+12.8%-4.2%
Mixture density network (on Saabi's chances)Player props182-121+10.5%-3.6%
Hierarchical panel regression (on Saabi's chances)Player props61-41+10.4%-3.1%
Hierarchical panel regressionPlayer props61-41+10.4%-3.1%
Mixture density networkPlayer props184-123+9.9%-3.5%
Massey ratings (on Saabi's chances)Games17-8+9.8%+3.1%
Davidson model (on Saabi's chances)Games33-13+9.0%+2.0%
Exponentially weighted moving average (on Saabi's chances)Games30-11+8.8%-0.4%
Exponentially weighted moving averageGames30-11+8.6%-0.5%
Volatility drag (on Saabi's chances)Player props162-117+7.7%-3.7%
Volatility dragPlayer props162-117+7.6%-3.7%
Extended Dixon–ColesGames34-18+6.3%+0.8%
Extended Dixon–Coles (on Saabi's chances)Games33-18+5.9%+1.5%
Thurstone–MostellerGames35-10+5.2%+1.3%
Bayesian Elo (on Saabi's chances)Games16-14+4.4%-0.2%
Prediction-market consensusGames29-29+4.4%+0.4%
Transformer (on Saabi's chances)Player props168-131+4.3%-3.2%
TransformerPlayer props168-131+4.3%-3.2%
MicropriceGames31-26+4.2%+1.8%
Bivariate Poisson (on Saabi's chances)Games26-11+4.0%+2.7%
Bivariate PoissonGames26-11+4.0%+1.7%
ARIMAPlayer props167-138+3.3%-2.4%
SARIMAPlayer props167-138+3.3%-2.4%
Gradient boosting classifier (on Saabi's chances)Player props161-127+3.2%-4.0%
ARIMA (on Saabi's chances)Player props167-139+3.0%-2.4%
SARIMA (on Saabi's chances)Player props167-139+3.0%-2.4%
Gradient boosting classifierPlayer props160-127+3.0%-4.0%
Bayesian neural network (MC dropout) (on Saabi's chances)Player props174-145+2.3%-2.0%
Multivariate Poisson (on Saabi's chances)Player props76-63+2.1%-3.8%
Multivariate PoissonPlayer props76-63+2.1%-3.9%
Variational inferencePlayer props167-128+2.1%-3.7%
Bayesian neural network (MC dropout)Player props173-145+2.0%-2.0%
Prediction-market consensus (on Saabi's chances)Games30-31+1.8%+0.8%
Monte CarloPlayer props21-23+1.6%+0.3%
Microprice (on Saabi's chances)Games32-28+1.5%+2.1%
Bayesian EloGames16-15+1.1%-0.2%
Logistic pairwise rankings (on Saabi's chances)Games35-16+1.1%+1.4%
Dynamic linear model (on Saabi's chances)Games28-13+1.1%-0.5%
Structural time-series model (on Saabi's chances)Games28-13+1.1%-0.5%
Variational inference (on Saabi's chances)Player props156-125+1.0%-3.7%
Logistic pairwise rankingsGames35-16+1.0%+0.7%
Capital asset pricing model (CAPM)Player props133-114+0.2%-3.6%
Random forest classifier (on Saabi's chances)Player props171-148+0.2%-3.1%
Random forest classifierPlayer props171-148+0.2%-3.1%
Score-driven/GAS modelPlayer props181-157+0.1%-2.9%
Thurstone–Mosteller (on Saabi's chances)Games35-10+0.1%+1.9%
Monte Carlo (on Saabi's chances)Player props119-1230.0%-0.8%
Exponentially weighted moving average (on Saabi's chances)Player props168-146-0.1%-2.6%
Exponentially weighted moving averagePlayer props168-146-0.1%-2.6%
Maher model (on Saabi's chances)Games16-9-0.2%+1.5%
Independent Poisson goals model (on Saabi's chances)Games16-9-0.2%+1.5%
Dixon–Coles (on Saabi's chances)Games16-9-0.2%+1.5%
Score-driven/GAS model (on Saabi's chances)Player props181-158-0.2%-2.9%
Capital asset pricing model (CAPM) (on Saabi's chances)Player props133-115-0.2%-3.7%
Deep ensemble (on Saabi's chances)Player props142-124-0.7%-2.8%
Deep ensemblePlayer props142-124-0.7%-2.8%
Bayesian state-space modelsPlayer props141-126-0.8%-3.2%
Quantile random forest (on Saabi's chances)Player props71-62-0.8%-3.6%
Quantile random forestPlayer props71-62-0.8%-3.6%
Gaussian process time series (on Saabi's chances)Player props170-149-0.9%-2.6%
Gaussian process time seriesPlayer props170-149-0.9%-2.6%
Heikin-Ashi trend (on Saabi's chances)Player props219-218-0.9%-3.1%
Expectation propagation (on Saabi's chances)Player props61-56-1.0%-4.3%
Expectation propagationPlayer props61-56-1.0%-4.3%
GeometricPlayer props136-145-1.2%-2.1%
Bayesian state-space models (on Saabi's chances)Player props141-127-1.2%-3.2%
1D convolutional network (on Saabi's chances)Player props141-124-1.3%-2.2%
Gaussian process classification (on Saabi's chances)Player props57-51-1.4%-4.0%
Gaussian process classificationPlayer props57-51-1.4%-4.0%
Heikin-Ashi trendPlayer props223-225-1.4%-3.0%
GARCH (on Saabi's chances)Player props156-141-1.6%-2.7%
Neural ODE (on Saabi's chances)Player props144-130-1.7%-2.0%
GARCHPlayer props156-141-1.7%-2.7%
Bayesian change-point detection (on Saabi's chances)Player props138-123-1.7%-2.9%
Bayesian change-point detectionPlayer props138-123-1.7%-2.9%
Metropolis–Hastings MCMC (on Saabi's chances)Player props17-11-1.8%-3.2%
Gaussian processes (on Saabi's chances)Player props52-47-1.8%-3.6%
Gaussian processesPlayer props52-47-1.8%-3.6%
Stochastic volatility (on Saabi's chances)Player props135-124-1.9%-3.0%
Temporal convolutional network (TCN)Player props110-99-1.9%-3.7%
Stochastic volatilityPlayer props135-124-2.0%-3.0%
Generalized least squares (GLS)Games29-22-2.1%+1.5%
Hidden Markov model (on Saabi's chances)Player props140-122-2.2%-4.4%
Sharpe ratioPlayer props165-150-2.3%-3.6%
Temporal convolutional network (TCN) (on Saabi's chances)Player props110-100-2.4%-3.7%
DiversificationPlayer props168-152-2.4%-3.2%
Sharpe ratio (on Saabi's chances)Games41-28-2.5%+7.8%
Sharpe ratio (on Saabi's chances)Player props165-151-2.6%-3.7%
Diversification (on Saabi's chances)Player props168-153-2.7%-3.3%
Hidden Markov modelPlayer props138-122-2.8%-4.4%
Dynamic linear modelGames27-13-2.8%-0.8%
Structural time-series modelGames27-13-2.8%-0.8%
EGARCH (on Saabi's chances)Player props155-144-2.8%-2.8%
EGARCHPlayer props155-144-2.9%-2.8%
Ensemble (props, stacked)Player props14-16-3.1%+0.2%
Geometric (on Saabi's chances)Player props139-154-3.2%-2.4%
Histogram boosting (LightGBM-style) (on Saabi's chances)Player props128-119-3.2%-3.2%
Histogram boosting (LightGBM-style)Player props128-119-3.2%-3.2%
Kalman filter (local level) (on Saabi's chances)Player props166-154-3.3%-2.4%
Generalized least squares (GLS) (on Saabi's chances)Games30-23-3.3%+1.3%
Kalman filter (local level)Player props166-154-3.3%-2.4%
Generalized least squares (GLS)Player props185-173-3.3%-2.6%
Extended Kalman filterPlayer props111-102-3.6%-3.2%
Dynamic generalized linear modelPlayer props111-102-3.6%-3.2%
Extended Kalman filter (on Saabi's chances)Player props111-102-3.6%-3.2%
Dynamic generalized linear model (on Saabi's chances)Player props111-102-3.6%-3.2%
Generalized least squares (GLS) (on Saabi's chances)Player props185-174-3.6%-2.7%
Random walk (on Saabi's chances)Player props102-94-3.7%-3.7%
Random walkPlayer props102-94-3.7%-3.7%
GJR-GARCH (on Saabi's chances)Player props146-139-3.9%-2.7%
GJR-GARCHPlayer props146-139-3.9%-2.7%
Metropolis–Hastings MCMCPlayer props17-12-4.2%-3.5%
Lognormal (on Saabi's chances)Player props139-130-4.4%-4.6%
LognormalPlayer props139-130-4.4%-4.6%
ARIMAX (on Saabi's chances)Player props141-134-4.5%-2.3%
SARIMAX (on Saabi's chances)Player props141-134-4.5%-2.3%
ARIMAXPlayer props141-134-4.5%-2.3%
SARIMAXPlayer props141-134-4.5%-2.3%
Particle filterPlayer props117-109-4.6%-3.5%
Particle filter (on Saabi's chances)Player props117-109-4.6%-3.5%
Ensemble (props, median) (on Saabi's chances)Player props214-206-4.8%-2.8%
Linear regression (on Saabi's chances)Player props158-149-4.9%-2.3%
Linear regressionPlayer props158-149-4.9%-2.3%
Efficient frontier (on Saabi's chances)Player props143-136-5.0%-3.1%
Efficient frontierPlayer props143-136-5.0%-3.1%
Implied volatilityPlayer props138-133-5.1%-2.3%
Causal adjustment (on Saabi's chances)Player props22-21-5.1%-3.7%
Causal adjustmentPlayer props22-21-5.1%-3.7%
Implied volatility (on Saabi's chances)Player props137-133-5.3%-2.3%
Size factorPlayer props139-134-5.4%-3.6%
Low-volatility anomalyPlayer props126-120-5.5%-2.6%
Home-adjusted Elo (on Saabi's chances)Games20-19-5.6%+0.5%
Home-adjusted EloGames20-19-5.6%+0.2%
Make your own picks against Saabi and your friends. Free: points to play with (no cash value), your own record, head-to-head challenges with friends and private leagues.