Ladies and Gentlemen the plumbing of the market is a maze constructed to remove money from traderâs pockets....That said if and when you can afford to place your brand new IBM Z main frame next to the oneâs belonging to the exchanges (yes that's right inside the data centre) you will have solved your routing problems
sorry to be such a killjoyI've looked at Tsay's book. It is exactly the sort of dreck that feeds delusional quant dreams of predicting price levels. Show me one person who has successively traded using ARMA, ARIMA, GARCH, etc. and I'll show you a figment of your imagination. It's hard enough just to accurately follow price (hint: most traders can't), let alone trying to accurately predict it.
The models presented in that book is the sort of stuff "grown-up" (institutional) quantitative traders actually use. People who trade cross-sectional statistical arbitrage use PCA or other factor models, cointegration or threshold chaining etc. People who trade longitudinal statistical arbitrage use all sorts of noise models, autoregressive models and many more. This is more or less where these things have been for years, so it's not necessarily even the cutting edge.
The real grown-ups are those of us who realize price can't be modelled. Modelling price is a fool's errand, or a childish fantasy: take your pick. Price isn't random but that doesn't mean it's predictable, and the whole point of modelling is prediction.
The book does not say anything about predicting the price. It talks about either modelling the dynamics of price (in univariate case) or modelling the relative value (in multi-variate case).Quote from kut2k2:
The real grown-ups are those of us who realize price can't be modelled. Modelling price is a fool's errand, or a childish fantasy: take your pick. Price isn't random but that doesn't mean it's predictable, and the whole point of modelling is prediction.
If your "grown-up" quants are so capable, why are they scared to death of directional trading? Everything with those wimps is hedges, arbs, swaps, and other pussy market moves. Present them with anything that isn't market-neutral and they need a change of underwear.
Like I said, the whole point of modelling is prediction.
I've had statistics and advanced calculus. I've designed my own custom adaptive smoother and a very accurate custom trend indicator.Quote from sle:
Two questions for you:
(a) Do you have any math education beyond grade school?
(b) Did you actually read the book and try to do the exercises?
Don't bother answering (b) if you have difficulties answering (a)
Prediction of what? Of dynamics, volatility or relative value? For example, if I have a model to forecast changes in implied volatility as a function of the spot price, does it contradict your view of the world?
Ok, so you are doing some simple noise analysis and some simple autoregressive modelling and are saying that it has actually worked for you. That, however, does not mean that more involved methods are trash, as you seem to indicate.
A PhD in CS-oriented physics and 15 years institutional trading experience, as a market maker and now as a portfolio manager. I trade mainly volatility products.
Something about just doing what's profitable, while facilitating the edgeless outright directional punts of excited degenerates..Quote from kut2k2:
... If your "grown-up" quants are so capable, why are they scared to death of directional trading? Everything with those wimps is hedges, arbs, swaps, and other pussy market moves. Present them with anything that isn't market-neutral and they need a change of underwear.
