For this coursework, you will have to develop a short yet complete agent-based model (ABM) of the
early 2021 GameStop (GME) short squeeze (https://en.wikipedia.org/wiki/GameStop_short_
squeeze). You can use any library you want, and any resource you want, but everything you use
must be appropriately referenced. The main goal of this coursework is to make sure you know how
to develop an ABM for a real-world financial scenario. In fact, after the GME short squeeze, several
investment funds started to recruit people who where i) familiar with financial subreddits, and ii)
capable to model events based on Reddit activity.
If you are not familiar with the GME short squeeze event, I suggest you read about it before starting
the coursework. Its Wikipedia page is an excellent starting point (and probably enough for what you
need to do).
Details of the model and assumptions you can make:
1. Your ABM should aim to model the period between 01-December-2020 and 04-February-2021.
2. You can use any research result (Google Scholar is your friend!) to model the agents and the
environment, and to validate your model.
3. Make good use of the Lucchini et al. paper uploaded on Learn. No need to reinvent the wheel.
Task 1 – 15 marks
Build the hedge fund class. Part of the reason the market behaved as it did, is that some large
institutional investors where betting on GameStop to go fail, and GME (its stock) to go to $0. To
account for this, build an agent class that models this type of agent. Specifically, this agent-type should
be a fundamentalist with a target fundamental price of $0. Also, this class of agents should be able
to short sell. Short selling means that it is possible to sell shares that are not owned, by borrowing
them from the market. In practice, this means that the agent can open a ’sell’ position and close it
anytime, but if the current price is higher than the price at the position opening, the agent will incur a
loss. Because of this, potential losses are infinite. Assume this class of agents can switch to a different
behaviour. Model the value function of this agent-type by using prospect theory.
Describe this agent by using the four steps we have seen in the lectures. You can add to the agent
behaviour outlined above, but clearly motivate any further assumption you use to model this agent
Task 2 – 30 marks
Build Reddit traders. Develop agent classes that reflect the findings in Lucchini et. al. Describe
each agent class you build by using the usual four steps. You can use any additional resource you like,
but you must reference it properly and justify its use.
Task 3 – 15 marks
Setup the ABM experiment. Model the market environment in which agents interact by buy
ing/selling. Model any additional interaction if you specified any in the previous steps.
Task 4 – 20 marks
Validate your model. Validate your model using the results by Lucchini et al. and any other
resource you like. Motivate your choice of parameters for the calibration and discuss the results.
Task 5 – 10 marks
Sensitivity analysis. In the interest of time, you are not required to run a sensitivity analysis.
However, explain how you would do it for this ABM, including a discussion of which parameters you
think are the most important to analyse.
Task 6 – 10 marks
Discussion of your results. Discuss your results and how good is your ABM to explain the
GameStop short squeeze event. Discussion should include any limitations of the model (anything
you think affected your results).
EasyDue™ 支持PayPal, AliPay, WechatPay, Taobao等各种付款方式!
E-mail: email@example.com 微信:easydue