"Traditional" RS vs "Weighted Liquid Rank" on Realistic markets

[akolonin]

Need to verify the claims:

• Use of traditional reputation system with explicit ratings under studied market conditions is misleading in any market conditions providing neither “security” and “equity”. Use of financial as implicit ratings is almost as good as use of explicit ratings weighted by financial values, but still provides less “security” and “equity”.
• For extremely unhealthy market conditions it may be more safe not using any reputation system at all relying on everyone personal experience as any reputation system may be cheated if volume of scam-driving transactions is comparable to volume of fair transactions.

Cassio wrote: "this flies so much in the face of real world markets that it needs much more serious justification than we've provided. We just studied a tiny set of unrealistic markets." Need to do that after #113 and #110 - for line of few selected market conditions, we need to have 4 runs: A) No reputation system is used
B) Using explicit ratings with neither aggregation nor weighting C) Using explicit ratings with financial weighting D) Using explicit ratings with financial weighting and aggregation Then gather results in individual sheet with LTS, PFS, weighted pearson and confusion matrix by category

[akolonin]

@deborahduong , in the current runs:

We need 5 with default parameters: 1) No RS 2) Plain Vanilla RS (weighting=False) 3) Weighted Liquid Ranks 4) TOM-based 5) SOM-based

And then, we need to explore space for 4) TOM-based 5) SOM-based

• following pattern in https://docs.google.com/spreadsheets/d/1HbAjXZzj6E1meXyxPLS8f62LbhLRIv3lCwpSLgACt8I/edit#gid=2064997300 plus add periods like 10, and 30 days for 1000X183 run.