Overview: To date, the G2F initiative (www.genomes2fields.org) has collected data on over 180,000 corn field plots including 2,500 hybrids and 162 unique environments.
Objective: Develop models to predict corn yield in the 2022 G2F trials based on the current G2F dataset and other publicly available data.
Outcome: The winner of the competition will receive a cash prize or equivalent compensation of $4,000 and expenses paid travel and registration (up to $1,200) to present their research at an international meeting (see About tab for details). Additionally, participants will be invited to co-author a paper with the competition organizers presenting the findings of the competition and the different modeling strategies (subject to organizer approval).
- Any individual or team can participate. One individual must be designated as team leader and all other team participants must be listed.
- Each individual can be involved in only one entry. You cannot be on multiple teams.
- Companies or individuals are encouraged to participate.
- In order to win and receive the cash prize, participants must commit to publishing (or allowing us to publish) their models with open public access after the competition.
- Competition participants agree to have their models and results included in a joint publication on which they will be co-authors if they desire. The organizers reserve the right to include or exclude participants based on the quality and creativity of their submissions.
- The joint publication will be delayed for up to one year from the submission date to allow participants to author their own publications on their models if desired.
To ensure fairness, G2F cooperator labs will be required to designate one or more individuals to take care of their 2022 data. These individuals will be ineligible to participate in this competition.
Competition participants from G2F cooperator labs can help gather the data in field, but only the designated individual(s) in that lab is allowed to access that data.
The organizers reserve the right to disqualify any participants from the competition if they judge that participant to have cheated or attempted to improperly gain an unfair advantage over others.
A member of the winning team is expected and encouraged to present their results at the 2023 Maize Genetics Meeting in St Louis, MO, USA.
- Any modeling strategy is permitted!
- G2F will provide registered participants with trial data from 2014-2021 for use in model development. G2F will also provide participants with weather and other test set input data from the 2022 growing season.
- Any other data which were publicly and freely available by February 1, 2022 are permitted and encouraged for use in model development. Data that became publicly available after February 1, 2022 is not permissible (with the exception of any DOI releases from the G2F initiative itself).
- Participants will be provided with the G2F data from 2014-2021 for use in developing and training their models. This will include Phenotypic data (yield), Field GPS locations, Weather data, Soil data, Genotypic data.
- Test set input information will also be provided for use in generating the predictions for submission and evaluation. This will include: 2022 Field locations, 2022 Field information, Genotypic data, and Weather data.
- By the submission deadline participants will submit:
- Final predictions of absolute grain yield for each hybrid in each test environment in Mg/ha adjusted to 15.5% moisture. A template document will be provided for participants to submit these results. Submissions for the leader board can be made prior to the deadline and the highest scoring prediction (lowest average RMSE) from each team will be used for the final evaluation.
- While not included in the results submission, participants should be prepared to provide their well organized code and any non-G2F provided data they have used upon request of the organizers. The organizers must be provided sufficient information to re-run the model themselves and validate the results if desired. Participants who desire to keep their model proprietary will not be required to provide this information, but they will not be eligible to win the competition or the cash prize.
- The winner will be determined based on having the lowest average root mean squared error (RMSE) across locations compared with the actual yield data obtained in 2022.
- Other ways of evaluating the models will also be employed for publication purposes, but they will not be used to determine the winner of the competition.
- Some locations or hybrids from the 2022 year may fail or have extremely low or unusual values. The organizers may drop locations and/or hybrids in these cases. If that occurs, the corresponding predicted value from the participant will also be dropped.
The winning team will receive:
- A cash award of $4,000 or equivalent compensation (generously donated by the National Corn Growers Association, www.ncga.com, and others). We will do our best to provide this prize to the winning team (i.e., through check, bank transfer, etc.) but cannot guarantee this will be possible in every country. It is the winner’s responsibility to pay any bank transfer fees and applicable taxes.
- An invitation and expectation to present their model and results at a G2F session held concurrently with the 2023 Maize Genetics Conference.
- Expenses paid (up to $1,200) for the 2023 conference (for one member of the team only).
- Invitation to be the first author on the joint G2F publication if desired.