Hi everyone, I'm ready for my first question here... every response is appreciated. If I want use some statistics/analysis/whatever in my algo trading/strategy development, do I ever need to really know the math behind it (calculus, combinatorics, probability theory, theorems etc.) OR it's just good enough (from practical point of view) to use well known high level concepts: distributions, regressions, well known models, pricing models etc. And use all this stuff like black box. What do you say? P.S. I do really well software development and all aspects of software engineering. It's not so hard to close the "math" gap, but avoid to waste a single minute.
Please don't trade, we really do not need another smart guy to compete with us. You already have good software engineering skills, use that and get into Google, msft, FB, uber and your income will far surpass most DIY traders with 0 drawdown.
https://en.wikipedia.org/wiki/Combinatorics May be this is not an exact term you expect to hear (I'm not English native speaker). AFAIK this stuff is a basis for probability theory, statistics etc.
first of all, it would be nice for you to realize for yourself why do you want to use certain statistics and not some other statistics depending on your answer you may realize how deep you need to understand how these statistical data was arrived at.... again, depending of what you want to achieve imho all those concept as well as statistics are dead without deep understanding how the market works in the type of analysis you try to utilize for your trading the real knowledge is the knowledge of causes, not effects (c)
I'm not a maths expert, calculation as such doesn't come into my trading style much but I can definitely say its good to understand percentages and decimals. Probability theory is interesting too, as all trading is a probability problem in the end: there has to be a possibility of being incorrect or there would be no possibility of gain.
One statistical calculation that has really helped my trading is the Pearson Correlation, which measures how closely related two series of numbers are. Result is a single number ranging from 1.0 (perfectly correlated), to 0.0 (perfectly uncorrelated), down to -1.0 (inversely correlated). For example if I am looking to hedge against a large AAPL purchase, I would feed the Pearson Correlation algorithm the percent change in daily close prices for AAPL and some other securities; examples vs AAPL: QQQ 0.72; SMH 0.58; IBM -0.06; GLD -0.17
Regarding algo development, the two questions I asked myself most are: "Is it statistically meaningful?" and "What is the probability that it will occur?" The more ways you know to effectively answer those questions, in the many scenarios that you'll face, the better, imo.
It really depends. Knowing a fair bit of stats helps a lot for pretty much anything quant related. Knowing a little bit of linear algebra helps a lot, especially for stat arb type of stuff. Snippets of calculus help for options trading.