Simulation and the Stock Market
Simulation is only as good as the realism it inspires.
-----------------------------------------------------------------------------------------------
---------------------------------------------
What does it mean to simulate the markets?
To try to get a handle on the requirements, let's compare two systems that have very different operating conditions and simulation dynamics:
In 1 they pretend to fly airplanes, and in 2 they pretend to process real information.
In a market simulator, should we pretend to trade, or pretend to process real market data?
The difference between the flight and ground simulations is that the former represents the flight of an airplane in some air space, whereas the latter represents the real information interchange among personnel of the air defense system (possibly to include intercepts by fighter aircraft).
By means of aircraft controls, like a flight stick and panel gauges, the flight simulator provides a way for the pilot to interact with and fly a seemingly real airplane. And by means of elements like radar screens, computer consoles, and phone lines, the SAGE simulation provides a way for the personnel to perform seemingly normal operational duties.
The pilot trainer uses a very realistic aircraft cockpit but strictly symbolic (mathematical) aircraft dynamics to fly in a controlled but purely fabricated air environment. Alternatively, the SAGE system uses the real operating system but only representational data to direct inquiries, initiate intercepts to identify, and possibly force, or even shoot down, purely fabricated bogie aircraft.
In other words, the aircraft pilots train in a flight chamber that has the look and feel of a real aircraft, but is a symbolic aircraft flying in a purely symbolic air environment. And the defense personnel train in a physically real system with fabricated information about enemy aircraft.
--------------------------------------------
In my view of a market simulation, the market trader would be the pilot, and the market would be the air environments of these organizations.
One approach would then be to make real trades using real tools immediately to hand (like a computer or telephone) and to simulate everything else, namely the whole market. But this isn't feasible. The operating environment is far too complex.
Instead, let's use the market itself as the real part of our simulator. So then everything would be real, including the trader's orders and the market information. The only part that wouldn't be real would be the interaction of the trader with the real market world. Our job would then be to find a realistic way to represent that interaction.
The analogy would be to make paper trades; it would be the real thing but without monetary obligations. You could use any of the available models/indicators to try to project prices of the real market and act on those predictions. It would be like back-testing theories, except it would be done in real time. Any trial data for simulated exercises, as well as test data for predictions based on market models, would be provided by real-world market values -- the real stuff on the real ticker tape.
The trader would be like the medical doctor who takes the patient's temperature, makes a diagnosis, and prepares a treatment. Everything but applying the treatment to the real patient would be real.
This kind of simulation assumes you actually make a trade at a specific price and time, that shares are available at the indicted price, and that the trade occurs as represented. For example, a trade based on a limit order below the market might assume that, on a percentage basis, say, enough shares are available at or better than the limit price to complete the trade.
--------------------------------------------