Forums  > Trading  > Optimal Trading using Monte Carlo Simulations  
Page 1 of 1
Display using:  


Total Posts: 1
Joined: Apr 2018
Posted: 2018-04-21 12:13
Hello all,

I'm trying to implement/improve a Market Making strategy in a highly non-liquid market. The question is I think a general problem that could also answer how to make an optimal pairs trading strategy for example.

So you have some kind of Monte Carlo Simulation of the market:
1. Bid Ask Spreads
2. Market Orders (which orders where hit)

Now the question is how to find an optimal strategy using the simulation. In general, the final model/strategy would have to make some decisions:
i. Should a limit order be placed
ii. Should a market order be placed
The model, of course, should consider if it is more profitable to wait for a hit on the limit side or just place a market order since the model is predicting better prices in the future.

I hope the problem set is clear. Any ideas how that could be solved?
Found some similar problems but most of them are using analytical methods to solve it. I would like to get an optimal solution/strategy under the simulation.

I looked on stochastic control but not quite sure if that kind of model could fit in.
Alternatively using Q-Learning, but that is more like a black box model so wanted to avoid that in the first place.

Thanks a lot for any hint!


Total Posts: 317
Joined: May 2006
Posted: 2018-04-23 12:32

You are mixing up several issues here.

The "limit order or market order" decison should be driven by your inventory, risk preferences and some benchmarks. If you have no inventory, there is no reason to place market orders. If you have inventory, you need some benchmark along the lines of "I can keep xxx amount for up to 10 minutes, yyy amount for up to an hour and zzz amount until close". Based on that, you try to get out with limit orders while you have time, and cross when you are out of time.

Time is obviously just an example. You pick what ever benchmark is appropriate. E.g., "I can keep xxx amount as long as the price is in this range", etc.

The highly illiquid market will pose another set of issues that you could write books about. Best bid and best ask are probably meaningless - you have to track a bigger part of the orderbook. And then, you have to have some idea what happens when you place an order in various parts of the book, what happens when you cancel it, what happens when you cross the spread, what happens when somebody crosses to you. Etc.

It's a big task, and by the sound of it you have a steep learning curve in front of you. Good luck.

"There is a SIX am?" -- Arthur
Previous Thread :: Next Thread 
Page 1 of 1