microprediction docs, client and beer money 
Packages and a platform for effecting autonomous prediction using lightweight markets instead of models because:
- Markets are better at prediction than models (discuss) - just harder to create and wield, until now.
- Small "microprediction" (glossary) markets are surprisingly accurate (prove me wrong).
Provocations (more in the book)
- No timeseries model should ever be called SOTA again (discuss).
- Prediction capability shouldn't be limited by the capabilities of a single mind, algorithm or company (discuss)
- Somebody's algorithm or data will find signal in your model residuals, someday (instructions).
- Most of "AI" will be done analogously, eventually, though this will take work. See the book or challenge me.
If you would like to see how easy it is to wield a new kind of market to effect turnkey distributional prediction, see the docs and, therein, observe that you can receive live help getting started on Fridays, or in the slack channel. Key points:
- The microprediction platform makes it pretty trivial to initiate your own bespoke market. Just ask Thomas Hjelde Thorensen who recently posted about his experience.
- Many algorithms already competing to predict other streams can easily predict yours too.
- Many more will do so in the future. Anyone can launch a new algorithm using anything they like in the Julia, R or Python ecosystem for example (it's a data interface).
- If you have a CSV with historical data (one column per variable) you can just send it to me (chat in slack say).
The TimeMachines, Precise, and HumpDay packages
I also maintain three benchmarking packages to help me, and maybe you, surf the open-source wave.
| Topic | Package | Elo ratings | Methods | Data sources |
|---|---|---|---|---|
| Univariate time-series | timemachines | Timeseries Elo ratings | Most popular packages (list) | microprediction streams |
| Global derivative-free optimization | humpday | Optimizer Elo ratings | Most popular packages (list) | A mix of classic and new objectives |
| Covariance, precision, correlation | precise | See notebooks | cov and portfolio lists | Stocks, electricity etc |
These packages aspire to advance online autonomous prediction in a small way, but also help me notice if anyone else does.
How microprediction.org "house" algorithms use these packages
Advances in time-series prediction funnel down into microprediction algorithms in various ways:
-
The "/skaters" provide canonical, single-line of code access to functionality drawn from packages like river, pydlm, tbats, pmdarima, statsmodels.tsa, neuralprophet, Facebook Prophet, Uber's orbit, Facebook's greykite and more.
-
The StreamSkater makes it easy to use any "skater".
-
Choices are sometimes advised by Elo ratings, but anyone can do what they want.
-
It's not too hard to use my HumpDay package for offline meta-param tweaking, et cetera.
-
It's not too hard to use my precise package for online ensembling.
There are other ways. Look for CODE badges on leaderboards.
- The muid identifier package is explained in this video.
- microconventions captures things common to client and server, and may answer many of your more specific questions about prediction horizons, et cetera.
- rediz contains server side code. For the brave.
- There are other rats and mice like getjson, runthis and momentum.
- winning - A recently published fast algorithm for inferring relative ability from win probability.
- embarrassingly - A speculative approach to robust optimization that sends impure objective functions to optimizers.
- pandemic - Ornstein-Uhlenbeck epidemic simulation (related paper)
- firstdown - The repo that aspires to ruin the great game of football. See Wilmott paper.
- m6 - Illustrates fast numerical rank probability calculations, using winning. However since the rules changed, this isn't that useful for M6 anymore. The precise package is way more useful, and put one person on the podium!