Using Data Science Competitions to Practice Quantitative Research Skills
Summary
The document considers how team based quantitative challenges can build skills in programming, residual analysis, portfolio management, allocation, statistics, and decision making under changing objectives. The original prompt seeks research exercises using substantial data and realistic historical settings, rather than conventional simulated investing games. It describes a game in which teams analyze information and choose actions as a way to practice rapid analysis and synthesis.
The response suggests data science competitions as an adjacent form of practice, including a historical patient admission prediction challenge and Kaggle contests. These projects can develop statistical and machine learning skills that overlap with actuarial work and financial time series analysis. The examples are recommendations, not documented trading strategies or evidence of investment performance. The cited healthcare contest may be highly competitive, and the response notes that its subject is not investment itself; its relevance comes from transferable analytical methods. The discussion does not compare competitions or provide a curriculum for building finance specific historical data games.
Key ideas
- Team challenges can exercise programming, statistics, allocation, and portfolio management skills.
- Data science competitions offer practical prediction tasks with supplied historical data.
- Healthcare admission prediction can build methods that overlap with actuarial and financial time series work.
- Competition experience may transfer analytical skills without directly teaching an investment strategy.
Tags
Full text
# Quantitative Analysis Games on Investing? # Quantitative Analysis Games on Investing? I have been playing a quantitative investing game (instructions here) that is actually quite interesting, teams have some time to decide their next moves. It requires all kind of knowledge such as spreadsheet -programming, residual -analysis, portfolion -management, allocation, XYZ -- variables and in the end the challenge is to develop the politics that maximize the utility in the game (and the utility functions change unexpectedly in the game) -- a real dynamic programming problem at the best. I feel it is quite fun and useful way to practice quantitative skills (besides extremely fast programming during sprints and good preparation drives analysis skills such as R and statistics/over-fitting/etc, besides a lot of fun-hectic time in a group). - Can you come up with more `"research labs"` (aka games) like that perhaps more realistic with historical data? - They do not need to be like the above, they could be on a more specific segment such as speculation with inverse yield curve with historical-data-approximated values or anything like that, ideas? - I am thinking that there may be some sort of competitions in real-life on this kind of issues, are there? Experience? I find this kind of games/challenges could be quite driving agent to teach and learn different kind of usually-low-profile quantitative issues, very well monopoly works but it is a bit too static to learn quantitative skills to analyze and synthesize information. P.s. I am now not looking for mock `investing or financial` games but games where you get a massive amount of data and you have some goal like in the example game, actually the massive amount of data is not essential -- just on some challenging quantitative issue. ## Answer by BlueTrin (score 3) https://quant.stackexchange.com/a/4103 Try this: if you win this, you are really good ;) - The Heritage Health Prize, is a contest to predict the admission of patients. The company provides you historical data. This is a very interesting exercise where you can apply AI and statistics/econometrics http://www.heritagehealthprize.com/c/hhp - if you are looking for different projects, why not join Kaggle (http://www.kaggle.com) ? Kaggle is a crowdsourcing website trying to match people/companies which have problems to solve with data scientists. There are often prizes. The Heritage Health Prize has started a long time ago and the level of competition is quite high, so I would not expect to even have a chance to win the prize if I started now, but it can be interesting nevertheless especially that your research on the patients admissions will have real world impact if useful ! If you are serious about this and preferably living in London, drop me a message, I would like to do some research but I would like to be part of a group. The Heritage Health Prize is not an investment game per se, but is kind of related to actuarial sciences and the skillset require overlaps with financial timeseries analysis.
Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.