I'm currently in search of an algorithm that can determine whether or not it's time to buy something (an item, a stock, a service, etc.) given an history of prices (30, 50, 100, ...).

My idea is something like this :

Given the history of prices, you should buy now because the price is likely to be going up.


Given the history of prices, the price is likely to drop even lower, hence you should wait to buy.

I've been Googling to find this, but i think I might not have entered the right keywords, because I couldn't find anything. That, or this does not exist because it's not reliable.

Statistics are not my main strength, but algorithms are. If you have knowledge of a mathematical study on this topic, I'll take it too.

  • 1
    $\begingroup$ You've asked an extraordinarily broad question. In its current form, it is unlikely that anyone can give you a meaningful answer. Please consider narrowing the focus. In particular, if there is some specific problem you are trying to solve, it would be best to detail that instead. $\endgroup$ – cardinal Dec 30 '11 at 13:32
  • $\begingroup$ Actually, what I'm trying to do, is predict the price tendency in the upcoming days, given a (long) history of the prices for that same item. But statistics is not in my area of expertise. $\endgroup$ – 3rgo Dec 30 '11 at 17:33
  • $\begingroup$ Is there any data that models things that might influence price? There is little to none meaning in past price curves to predict new price. $\endgroup$ – clyfe Dec 30 '11 at 19:03
  • $\begingroup$ Try "optimal stopping" (there are a couple of links under optimal-stopping-from-an-unknown-distribution. $\endgroup$ – denis Jan 1 '12 at 13:39

Sorry, this may seem like a simple question but in quantitative finance this is almost unanswerable. Even at places like algorithmic trading companies where they would love to know exactly when to buy a stock and when to sell it, they do not have solid models to answer this question. Very rarely can previous prices predict future prices.

Speaking for trading companies, instead of trying to predict the future price and claiming the present is a good time to buy or sell, quants will focus on trading volatility which, some argue, have better mathematical properties. Trading volatility is interesting because the goal is to make a set of transactions that have a mean 0 outcome, but high variance. The higher the variance, the higher the risk, and the higher reward


Not the answer you're looking for? Browse other questions tagged or ask your own question.