Marc -

Research question

Each member looked at it

Jan - 

Tradeoffs in sensitivity of the algorithm. Develop an algo to see if it is collusion. Play with specificity to identify where false negatives and false positives 

Would probably need 

Very computationally expensive

The more grants we have, the more computation that is needed.

It has to do with a threshold. Compare healthy nodes 

Whats a good way to look at a grant from different perspectives to see thats

Need to cap the minimum amount of money that someone can put into an individual grant. 

Marc

Became a trusted seed of the governance group. 

 - Looking at raw conribution data to see how it evolves over time. 

Through the round is we could see

We should really look into the QF algo and enhance it to better. 

We should 

As a working question, we hypothesize that contributor behavior will change over time and certain patterns emerge during a round and in between rounds:

Contribution to Notebook

https://github.com/marc4gov/gitcoin_cadcad_model/blob/te-gitcoin/Gitcoin_grants_Round_data_research_Marc.ipynb

Write up

Apart from the cadCAD simulation data, it's interesting to look at the raw contribution data to see if there is some strategic behaviour pattern evolving over time. In the aforementioned notebook we look at two temporal aspects of the Gitcoin Round 9 data: the total amount a grant gathers and the node degree of grants in the graph-based network of contributors-to-grants over the duration of the Gitcoin Round. We already mentioned that the best strategy to extract funding according to the QF-method is to have substantially small amounts spread over multiple contributors to a grant. So, one definition of strategic behaviour is to look for grants accumulating lots of small contributions over time by different contributors.We calculate the degree of a grant (x-axis) to its total_amount/degree ratio (y-axis) in a 'Gapminder'-style plot, the width of the circle is the total_amount a grant collects during the round. We presume that the smaller the ratio and the higher the degree (circles moving to the right bottom corner), the more a grant is suspect to strategic behaviour. Stepping through the timesteps, other interesting patterns submerge like 'whale' contributors suddenly moving in and making a grant moving fast to the right, suggesting a 'kickoff' for maximizing QF-results. We use the same style plot for the contributor's orientation, however results are somewhat blurry here.Just another way of plotting, we use a spring-layout for mapping contributor's to grants, to see how node degrees and contributor's amounts (width of the vertex) are evolving over time, and again we notice some big contributors moving in at a certain point in time. By visualizing what's happening during a Gitcoin Round, we get a better picture of emerging behaviour, both from a grant's and a contributor's perspective.

Marc -

Research question

Each member looked at it

Jan - 

Tradeoffs in sensitivity of the algorithm. Develop an algo to see if it is collusion. Play with specificity to identify where false negatives and false positives 

Would probably need 

Very computationally expensive

The more grants we have, the more computation that is needed.

It has to do with a threshold. Compare healthy nodes 

Whats a good way to look at a grant from different perspectives to see thats

Need to cap the minimum amount of money that someone can put into an individual grant. 

Marc

Became a trusted seed of the governance group. 

 - Looking at raw conribution data to see how it evolves over time. 

Through the round is we could see

We should really look into the QF algo and enhance it to better. 

We should 

As a working question, we hypothesize that contributor behavior will change over time and certain patterns emerge during a round and in between rounds:

Contribution to Notebook

https://github.com/marc4gov/gitcoin_cadcad_model/blob/te-gitcoin/Gitcoin_grants_Round_data_research_Marc.ipynb

Write up

Apart from the cadCAD simulation data, it's interesting to look at the raw contribution data to see if there is some strategic behaviour pattern evolving over time. In the aforementioned notebook we look at two temporal aspects of the Gitcoin Round 9 data: the total amount a grant gathers and the node degree of grants in the graph-based network of contributors-to-grants over the duration of the Gitcoin Round. We already mentioned that the best strategy to extract funding according to the QF-method is to have substantially small amounts spread over multiple contributors to a grant. So, one definition of strategic behaviour is to look for grants accumulating lots of small contributions over time by different contributors.We calculate the degree of a grant (x-axis) to its total_amount/degree ratio (y-axis) in a 'Gapminder'-style plot, the width of the circle is the total_amount a grant collects during the round. We presume that the smaller the ratio and the higher the degree (circles moving to the right bottom corner), the more a grant is suspect to strategic behaviour. Stepping through the timesteps, other interesting patterns submerge like 'whale' contributors suddenly moving in and making a grant moving fast to the right, suggesting a 'kickoff' for maximizing QF-results. We use the same style plot for the contributor's orientation, however results are somewhat blurry here.Just another way of plotting, we use a spring-layout for mapping contributor's to grants, to see how node degrees and contributor's amounts (width of the vertex) are evolving over time, and again we notice some big contributors moving in at a certain point in time. By visualizing what's happening during a Gitcoin Round, we get a better picture of emerging behaviour, both from a grant's and a contributor's perspective.