Leaderboards

univariate-k_002univariate-k_034univariate-k_003residual-k_021univariate-k_008residual-k_001univariate-k_013univariate-k_005residual-k_003residual-k_002residual-k_034residual-k_005residual-k_013residual-k_008univariate-k_021univariate-k_001

Time-Series Elo Ratings

Produced by timeseries-elo-ratings and based on k-step ahead prediction duels using live time series data,

Algorithms here can be found in the TimeMachines Python package, and these in turn draw on a subset of functionality from popular Python time-series packages. Examples include Facebook Prophet, Statsmodels TSA, Flux, PmdArima and more. If you have a suggestion please file a issue or, even better, add a skater and make a pull request.

Wins and losses are based on RMSE with 400 training points and 50 out of sample predictions. Residual streams use data from probability integral transforms, as explained in An Introduction to Z-Streams. Models high up on the residual leaderboards might be worth tacking on after you've already applied your favourite model.

Some algorithms are deployed to predict live data. See the example crawlers. Further motivation for the project is explained at microprediction.com.

NameRatingGamesActive
thinking_slow_and_fast1968.0107yes
precision_ema_ensemble1903.0108yes
tsa_p2_d0_q01896.042yes
tsa_p3_d0_q11857.038yes
quick_precision_ema_ensemble1842.045yes
tsa_p3_d0_q01814.046yes
slow_precision_ema_ensemble1812.046yes
slow_aggressive_ema_ensemble1792.047yes
divine_univariate_hypocratic_fast1754.076yes
aggressive_ema_ensemble1754.0110yes
tsa_p1_d0_q11753.052yes
tsa_p2_d0_q11752.053yes
fbprophet_cautious_hypocratic1750.080yes
quick_aggressive_ema_ensemble1744.060yes
slow_balanced_ema_ensemble1719.043yes
dlm_univariate_a1719.061no
quickly_moving_average1709.0141yes
nprophet_p2_hypocratic1707.019yes
slowly_moving_average1698.0169yes
balanced_ema_ensemble1696.0107yes
thinking_slow_and_slow1693.097yes
thinking_fast_and_slow1684.0105yes
quick_balanced_ema_ensemble1675.047yes
nprophet_p11630.018yes
divine_univariate_hypocratic_slow1601.079yes
regress_change_on_first_known16000no
fbprophet_chaser16000no
tsa_p1_d1_q016000no
tsa_p2_d1_q016000no
tsa_p3_d1_q016000no
sluggish_moving_average1593.0127yes
thinking_fast_and_fast1570.093yes
divine_univariate1564.0131yes
fbprophet_univariate_hypocratic1560.080yes
tsa_p1_d0_q01551.045yes
nprophet_p5_hypocratic1523.028yes
fbprophet_cautious1520.0130yes
fbprophet_recursive1512.0135yes
nprophet_p51503.018yes
nprophet_p1_hypocratic1495.025yes
nprophet_p31493.037yes
pmd_exogenous_hypocratic1467.093yes
fbprophet_exogenous_hypocratic1439.092yes
nprophet_p81437.029yes
nprophet_p3_hypocratic1424.029yes
fbprophet_univariate_univariate_hypocratic1418.077yes
rapidly_moving_average1402.0123yes
dlm_univariate_b1399.045no
fbprophet_exogenous1380.0136yes
nprophet_p8_hypocratic1375.018yes
fbprophet_known1367.0143yes
pmd_univariate1354.091yes
fbprophet_univariate1338.0140yes
nprophet_p21306.025yes
empirical_last_value1299.0161yes
fbprophet_exogenous_exogenous1218.082yes