Quant analytics: Test of trading strategies on Brownian process

(Last Updated On: September 30, 2011)

Quant analytics: Test of trading strategies on Brownian process

The method testing the trading strategies:

1) Generate a time series generated by the Brownian process. Distribution of the random increments for modeling the market data. In the simplest case, you can use a normal distribution.
2) Optimize and test a trading strategy on simulated data.
3) Repeating the multiple steps 1 and 2 assess the distribution of estimates of profitability of trading strategy.
4) Optimize and test strategy on real market data. Obtain an estimate of profitability strategy.
5) Use distribution statistics (obtained in step 3) for reject the hypothesis of zero (or nonpositive) profitability a trading strategy.

This method used by someone before?

, I did miss the fact the data is generated over 95 years.

I am not doubting how GBM with mean reversion works, I am merely commenting that with no stop and a known distribution, designing an algorithm that makes money is a trivial exercise.

—–

now you got me. coud you say someting more on the tools you use?do you use micro structure or smoother methods PDE on dt, HF econometrics, events?
or at least how many levels in ths book do you use?just bid ask or? thst’s really an interestig info for me. I tried some time ago to make some study on 5 level book on personal collected data from EUREX futures.(they even have 10 levels). I was able to make some trade on that but nothing to be used for algo on the long run.

in the chart the strategy is up 3 the buy or a random entry?

your system seems really cool!!do you have some paper to know something more on the idea?I have read papers on market making in HF, but your strategy seems really cool and you don’t have even the rebate!

Python is used to generate statistics on trade data (time & sales) and 3 levels of order book data. I believe we (there are multiple individuals involved) use what you might term micro structure.

I have not seen anything that works below level 3 because of spoofing and hedging. We simply look for ratios among the levels that give us a slight statistical edge for the direction of the next tick.

When applied thousands of times per day, that slight edge is enough to make money.

NOTE I now post my TRADING ALERTS into my personal FACEBOOK ACCOUNT and TWITTER. Don't worry as I don't post stupid cat videos or what I eat!