client, docs, install, help
microprediction- leaderboards include links to algorithm code.
- glossary
- examples
- slack
- videos including an introduction to uses of microprediction.
- book and mp3 for Chapter 1.
Just go to setup to instantly participate and grok microprediction. Or ping me a cal invite for Fridays 11:30am EST and I'll walk you through getting started deploying your models. See meet. This was supposed to be the microprediction client page btw but Github made this a "user page" and they know best. Hi, this is my blog and, oh, so is this. I've always worked in the private sector, though very occasionally publish. My book is out now (buy, read, then complain at /building_an_open_ai_network) and here's other other stuff I've written. I've been annointed "influencer", aka LinkedIn jackass.
I'm dynamic and fascinating. Not really but I run a slack channel for those interested in time-series, covariance prediction, optimization and other things enabling semi-autonomous collective microprediction. Most Friday's at noon I make myself available to anyone interested in these things (see meet), even people selling me stuff on LI who read this far (most don't). I'm not big on creating other deterministic future work interruptions even though I do realize there is excellent tooling for this.
TimeMachines, Precise, and HumpDay packages
TheI 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!
microprediction.org platform
TheJust go to setup, really.
I also maintain a live exchange (high velocity prediction market of sorts, in the broad category of colliders, per my book's outrageous neologism) where distributional time-series prediction algorithms (Python, R, Julia mostly) duke it out, trying to predict future values of streams. Why? So my hedge fund can forever be at the forefront of open-source algorithmic advances that help predict stocks and related quantities.
But there's a bigger picture too per the book and let it not go unemphasized that there is $50,000 in prediction prizes up for grabs to foster progress in that direction. A prediction web will give clever people the opportunity to let their algorithms help someone without their getting in the middle, should they wish to. It remains a social experiment to determine whether that's sufficient motivation. You can see examples of what I've let loose and those will attack any problem you pose too. I'll send you to the better organized documentation that supercedes this README, should you wish to either consume or supply prediction.
No? Prefer to read an entire page before following links. Totally respect that. TLDR:
- I publish live data repeatedly, like this say, and it creates a stream (like this stream or this one or any from the listing).
- Algorithms like this guy compete to make distributional predictions 1 min ahead, 5 min ahead, 15 min ahead and 1 hr ahead.
In this way I can:
- Get live prediction of public data for free (sometimes I make the "public" data obscure)
- Indirectly benefit from hundreds of packages from Github of uncertain quality, and not just Python.
- Sip pina coladas while the accuracy magically improves over time. Read about this example.
Again, should you be interested in helping I'll point out that due to its obscurity, running this script makes a heck of a lot more economic sense than bitcoin mining - try it out or see the docs.
Just to lean on the differences between this and the other benchmarking efforts, the TimeMachines package is traditional open-source software for point-estimates and confidence, whereas the Microprediction client offers live crowd based distributional prediction and also, in theory, exogenous data search.
But the connection is provided by the fact that among the algorithms that compete at Microprediction quite a few use the TimeMachines algorithms (see /skaters), drawn from packages like river, pydlm, tbats, pmdarima, statsmodels.tsa, neuralprophet, Facebook Prophet, Uber's orbit, Facebook's greykite and more. Some are open source (look for CODE badges on leaderboards) but others are private to their author.
Memorable Unique Identifiers and other platform repos
- 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.
Some of my other packages:
- 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!
My book is here!
Hope you like it.