Datathon 2020 Jury Results



The teams were assigned in groups of 5-6 (working on the same case, and one mixed group with various cases). Every group gets to be judged by three members of the jury with closest expertise. The two teams which got ranked at the top of their group qualified for the final. The jury vote was based mainly on the content of the articles of the teams. The Jury members provided feedback below most articles for which we are thankful.


So the finalists were:



After the final 6 teams were announced, they had time to polish their article and to create a video presentation to the Jury. One team failed to do so and was disqualified (which is a pity because there could have been other semi-finalist in their place).

Final ranking with concordance measure (Kendall W) and significance measure (Chi Sq)

The procedure for ranking follows Spearman & Kendall Rank correlation method (with the assumption of the disqualified team ranked as 6th in all rankings). The team with the best collective rank wins. Kendall concordance (0,1) is used for measuring consensus. Chi-square is used for measuring statistical significance.

Pairwise Spearman correlation among Jury members’ rankings and Jury group rank

As additional analysis the pairwise Spearman correlation (-1,1) renders similarities between each pair of Jury member and each Jury member and the Jury group ranking. Values (-.35,.35) are considered no correlation, values (|.35|,|.70|) are considered weak correlation, values (|.70|,|1.0|) are considered strong correlation.


Sociogram based on Jury members’ rankings

As one more additional visualization, the sociogram renders (with arrows) correlation >.35 (at least weak consensus). Note: topological distance does not reflect correlation distance, as the configuration is settled by push-pull algorithm.

Highest network ranks (jury members with most consent over all):
1. Vladimir Labov
2. Liad Megen
3. Demir Tonchev

And the winners are


Share this

Leave a Reply