tommaso

I use the SCX index See here: Metrics for Algorithmic Trading by Tom Gastaldi and also here: G-Bot Algorithmic Trading Project - Strategies You need to invest in "essentially different" price moves. Or else you are just "overloading" the same investment (and can easily blow up without even realizing it).

Massive scalping can also be used as a form of "hedging". To create a "cushion" to protect larger trades. See for instance this thread: Forums - Algorithmic Trading a large folio with small orders or this post on Linkedin: I would be curious to share experiences: For you, what are 1- the most efficient hedging strategy you use? 2- the most efficient way to reduce drawdown? Thank... | LinkedIn

Best way is to trade a large folio of ETFs and use the cointegrations for hedging.

Hi Xiao, yes i real trade with it, and so do some of my users/friends (which include large funds too). I am not so fool as to make public claims about my income, but i let anyone interested to freely try the plaform on their machines for as long as they wish. Diligent traders have to be able to take their own determinations, based on their capital and intelligence :-) I dont even need to "retire", as i don't consider this work, but pure fun. You are right, if you are intelligent, you love coding and have a lot of time at your disposal, the best way is certainly to develop your own system. It's a full time commitment though, so it requires real passion and love. ;-)) Tom

Hi Xiao. I m also with IB and i have developed a platform which i am also distributing free (a working project). Take a look a it. I think it's pretty stable and reliable and might solve some of your headaches: my IB platform Then let me know if you found it useful and, especially, if you have suggestions for further improvements for the project. This will allow both discretionary and automated trading. T

Using the 1 realization of past data will only lead to curve fitting. "Prediction" makes no sense as far as trading is concerned. People easily forget their observations are submerged in an ocean of variance... Tom

Hi onesmith thanks for asking. Since you are interested in a recent value of the "direction", you limit your computations to a recent number of price variations ("moves" from a gridline to another). The (k-1) would represent how many latest distinct price changes you are considering to assess the recent "direction". When you use equispaced gridlines (and do not use weighting), you can just simplify as follows (which is immediately intuitive to visualize): Number of "UpMoves" - Number of "DownMoves" sdx = -------------------------------------------------------------------- * 100 Number of all "Moves" intending here "UpMoves", "DownMoves", "Moves" as the distinct price "moves" from a gridline to the adjacent one. [ The distance between gridlines depends on your "timeframe" (similar to when you measure time, you may be interested in microseconds variations, or you may just be intested in minutes, depending on what you are doing). (I often use a distance between gridlines equal to 0.1% of the price.) ] Let me know if something unclear here. Tom

another effective (and more robust) way to determine direction (actually a form of "slope") can be the SDX - signed direction index SDX Tom

Do Or Die congratulations for your magnificent work and posts. You are a great contributor! Tom

Right, that is the attitude! It's is there just to be tried by people and gather and implement as much feedback and constructive criticism as possible. [ Many fund managers are already testing it severely. See for instance: Forums - Trading FUTURES with IB ] not claiming anything, clearly. Just hard work and dedication. But putting together all the suggestions from many traders around the world, something good has hopefully to come out ;-)) Right ?

Try this G-Bot Algorithmic Trading - Platform Screenshots G-BOT Algorithmic Trading - Project and Platform G-BOT Algorithmic Trading - Project and Platform (100% auto, also with possibility of manual intervention). A beast!

You should first define the notion of Randomness. One of the most famous probabilist, De Finetti, started his famous book on probability with the sentence: "probability does not exist". So, go figure. As we know nothing about future prices, and they are obviously unpredictable, we could say prices are random. Or better, not deterministically predictable. And, actually not even probabilistically. ;-) T

Right, coin toss is a too simplified and useless model to make reference to. A more useful model could be the GBM with mean reversion, and then one could just start making some sense, at least with the precaution to test the strat with **different volatility** levels (as mkt as a non-costant volatility obviously): ModelAssist for ModelRisk Algorithmic trading stretches much beyond a naive immagination can reach. There are ways to "overlay" different strategies (for instance a trender against a countertrender) in such a way that one strategy "protects" the other one, and both run without stops. There is also the concept of "artificial option", a conceptual device created using the folio instruments. No need to say that this is much better, and can be proved experimentally, than attempting to use stops at single trade level, which is an absolutely sure way to leave all the money in the mkt, in the long run (i am talking of algorithms). As said, smart traders can still trade with stops and survive a little longer (sometimes more than their lifetime), due to a better choice of high probability "trades" (news processing, past experience, instinct, understanding of international economy, local high/low perception, etc.). Tom