I discuss in a previous post that Hi Stakes Market Game is under a feature freeze and I am ironing out bugs. One bug relates to asset valuations. Another terrible bug was related to my in memory db library that works like CQEngine. Those bugs were a catastrophic cascading probe, one compounded on another.
My In-Memory DB Library's bug was in the query planner specifically in the cascading probe where filters after two to be ignored and incomplete implementation of the necessary residual filtering. There were fixed and I can report significant correctness and performance boosts. One benchmark dropped to nanoseconds from 40 microseconds. Nuts.
Hi Stakes Markets support multiple asset classes, so asset valuation is done differently for each, but still within each of the classes are stratification based on valuation. How do you value those over time? How do fluctuations feed back into the volatility engine? How much strength factored into the feedback? How do you mitigate amplification loops? It's tough, I've spent this week alone trying to tinker with the volatility engine chasing down mathematical bugs or inconsistencies. These are the kinds of questions that come up when you delve deep into simulated valuation.
One thing I'm glad I have is a circuit breaker system that halts individual assets when they start getting out of bounds. That system alone has helped tremendously with stopping runaway valuations. It's invaluable. It can be aggressive too, so it's not perfect. The markets are miles ahead better than without it. Who says regulators are bad? I guess too much regulation is bad.
Because Hi Stakes Markets is a financial simulator, we don't have all the variables necessary to make it easier to calculate scores and the ones we do have are lower resolution for simplicity. One thing we all come to expect from markets is moving prices, that's the valuation part. Volatility is measured, but how much of it should contribute to the next valuation and the next? How about historical volatility and pricing?
I can't give specifics right now as they are in flux, but my ideas have been rooted in the use of hyperbolic trigonometric functions to determine how much volatility contributes into valuation. Volume is an important factor in the volatility strength contribution. That allows outside input, aka players trading, to impact the simulation. Without volume, daily price motion pretty much serves as the baseline volatility metric. A big daily motion should not impact volatility as much as trading influenced motion.
When it comes to regulations, exploits are going to be a problem so I've been trying to figure out ways to limit exchange exposure to risk. I need to test my margin call system for players that want to use margin for Futures and CFDs.
A ton of brainpower has been dedicated this week at answering these valuation questions and the outcome will be presented soon.
I got preliminary app screenshots on the app store and they look amazing. Incredible stuff really. They really showcase how simple the app is even though there is a ton of complexity that powers it.