First test simulation for combine AMMPS/SHARKFin SPARK run

[sbenthall]

Static parameters

SHARKFin --quarters 4 --expectations InferentialExpectations --expectations ClientRPCMarket

(use the default CRRA, DiscFac, dividend growth rate and dividend std, p1, p2, AttentionSimulation, LUCAS0 population)

AMMPS -days parameter to 240 to simulate a "sharkfin year".

Grid - Variation:

Seeds: 10 ? (SHARKFin -d)

AMMPS parameters:

• mm_lucas_factor =[0.0,0.15,1.0]
• inst_val_std = [0.1,0.2]

SHARKFin parameters:

--attention [0.015, 0.05] --dphm [1000, 10000] --zeta [0, 0.3, .7, 1.0]

Expected results:

Less MM Lucas factor, more II variation, more attention, higher zeta, higher dphm all contribute to the collapse of the financial system.

Other notes

We will continue to use the stripped down version of AMMPs with only MM, II, the Broker, and noise traders. This should have 30 minute runtime, but won't reproduce stylized facts of the stock market.

[wjt5121]

960 Simulation Parameter Grid has been initiated. 571 Simulation have completed the remaining are still running. I will update once complete. I've also attached the parameter grid csv. lucasShark2parametergrid.csv

[wjt5121]

Simulation Update: Simulations 517 through 960 failed tom complete due to the data vol becoming full. I re-staged the environment and restarted the simulations which are about 10 mins in. I will update again once the sims are complete and data is available. I have attached the

[wjt5121]

Simulation Update: With the second run of the experiment ll 960 simulation completed. The longest taking about 47mins. I'm uploading the data to the NFS share now. I will update @gms158 once the transfer is complete.