Let me offer an example of a correlation that I thought of today that is very real, yet is almost useless as a predictor for specific events.
I hike in the dessert quite a bit and often run into rattlesnakes. I probably scare up a few dozen each year. If the weather is below about 65 - 70 degrees, rattlesnake sightings are very rare. Similarly, if the temperature is above 105 - 110 degrees, rattlesnake sitings are rare. There is a clear correlation between daytime high temperatures and rattlesnake sightings. In this particular example, the reason for this correlation is understood, but it doesn't have to be.
On any given day, however, I can look at the temperature and still not have any idea of whether or not I will see a rattlesnake on my outings that day. I've seen them when the temperatures are cold. I've seen them when the temperatures are hot. And I've been out on many days when the temperature is perfect for rattlesnakes and not seen a one.
There is a correlation between temperature and rattlesnake activity. If I collected the data sets on all my outings that included daytime temperatures and rattlesnake sightings, a correlation analysis would find a significant correlation. But temperature is not the only factor. In fact, the factors that contribute to the sighting event are so complex and numerous, that temperature alone will not allow me to make reliable predictions.
We're missing. I completely understand what correlation means. What I'm asking is WHAT items do you believe are correlated in historic investing data. In specific, you said specific correlations are used to assemble monte carlo situations (as opposed to the simple random walks that I thought they all were).
Just to revisit and perhaps explain my fascination with this, my former life as I may have mentioned once or twice was as the marketing droid responsible for driving higher microprocessor usage tools and products. I drove the investment of about $300M into companies that built products based on bayesian and neural nets, AI, smart handwriting recognition, active group collaboration tools, simulation tools (like those monte carlo simulators), etc. I sat on the board of a handful of these investments. I ran $40M worth of programs to encourage universities to come up with better ideas. My efforts were responsible for as much as $3B in incremental revenues to the company.
In otherwords, I dont need a lesson in what these are...
Some of the companies we worked with were developing tools to analyze stock market data looking for ANY correlations, past present or future. Particular interest in the future.
At the time I retired, NOBODY had ANYTHING that could hook what happened in any one year to the ones before or after with any strong correlation in the way that (i think) you described.
What you described was that there are correlations between SOMETHING (and THATS what I want to know) and SOMETHING ELSE (and that too!), that work better historically to connect returns and movements from one year, and those that went before and after. And that these are used by some monte carlo simulators to put together strings of years in an investment returns run, and that they therefore arent random, and that this somehow explained why monte carlo simulators return worse results than simple historic analysis.
Since nothing like that existed at the time I quit 3 years ago, and you say stuff like this exists (even if in theory only), I wanna know what the heck it is!
So what I want to know is:
#1 - What two or more things create correlations in any kind of market data, current past or future, that connect that years returns or behaviors to years past...ex: when interest rates are high, bond prices fall sort of correlations.
#2 - How are these applied to monte carlo simulations to create the run periods and/or produce their data walks.
#3 - Where are these monte carlo simulators. The one gummy pointed to appears to create its data runs through random walk.
