Short-term volatility model GARCH autoregression, mean-reversion?

[John TRD]

Hello what are you using,or can reccomend for volatility prediction (in forex market 10-30seconds)
For example Time series 10, 11,12,13,14,13,11,10,10,9,10,11,10 (1 second time period)
We can calculate EMA (exponent moving average) (for example 11.5) and difference between current price S and MEANPRICE
First 4 numers 10,11,12,13,14 looks like autoregressive time series with covariance and standart deviaion
We can calculate R=S(n)-S(n-1) Rn,R(n-1) R(n-2) R-price momentum
We can use GARCH MODEL
WE can se Mean reversion with http://en.wikipedia.org/wiki/Standard_score
We can use Autoregressive model formula http://en.wikipedia.org/wiki/Autoregressive_model

Problem Autocorellation say to use R(n) and R(n-1) are highly corellated and if R increasing so R(n+1) will increase too
BUT Mean reversion says that Next value should Decrease When price will be 14 R (n+1)should go down until Price S will be 10
So there are 2 diffrent short term models /First predict R(n+1) wil lbe higher But Mana reversion Say to use that R(n+1) should descrease

HOWTO combine thoose models, r what i shold use for prediction short term 10-30seconds in market
Thanks