Here is a question. Should I make the time-frame selectable? In other words, what option would this ideal indicator have? It would need a time-frame of course. If we looked at the past 50 contracts exchanged, that would be a very small and noisy time frame. So I think the indicator should have a time-frame value that the user can select. Timeframe = 1000 would mean that the past 1,000 contracts would be used to determine what the result would be. If 80% of them were the ask being hit, the indicator might read 80% -- if, on the other hand, it was exactly equal to one another, the indicator might read 0%. This indicator might also benefit from another line called "speed." Speed would measure the average X number of contracts per a specific time period. If volume suddenly increased, this line would move upwards to 100% or downwards to -100%. This would be a stochastic of sorts that measures the rate of how quickly contracts are accelerating in changing hands. Now, if the first indicator was peaking at 80%, and suddenly the speed line shot higher and the first indicator started dropping, that would be a perfect signal to sell. You could even set the timeframe for whatever trend period you wanted to trade from. aphie
In my earlier pursuits I had this 'indicator' programmed into my system. I wasn't able to garner much useful info from it BUT i might not have given it enough time. I eventually moved onto higher timeframes systems ...
Interesting. I had the time frame selectable but in a different sense. 5 min meant the number of contracts hitting the bid or ask within that bar. Heck why not do both ... have one for time delimited and another for #of contracts...
Deep, What are you mainly using now to trade off of? You sound like you've been through all of this before. aphie
Hindsight I moved to a longer timeframe daily ,30 60. What you are working on is more 'micro'. I don't want to dissuade you from your research but my belief is that the smaller timeframes are dominated by extremely large institutions like DE Shaw that do arbritrage all decked with super fast execution - quotes ( lot of them do cross market arbritage too). Any non-random noise would be captured by them before most 'mortals' have a chance of detecting and acting on that opportunity. I believe the Prediction company (doyne farmer et al.) backed by UBS Warburg managed to find an 'edge' using chaos theory. The people he worked with were extremely bright AND passionate about their work (hell they actually tried to 'model' water dripping patterns from a faucet). So that is the level of committment / resources / talent required in my mind.. I personally know my own limits BUT also realize that is not the only way to skin the proverbial cat .... hence my migration to higher timeframes .... at any rate it may be possible .. i just thought why make it tougher than it needs to be.... DH
aphie, The time frame is not relevant because it gives you a false reading of force in the market. If in the first hour of trade, 1,000,000 contacts trade and you get a reading of +30% for those 1,000,000 contracts, you have a lot more upward force than if you got the same reading during the lunchtime hour, when only 500 contracts trade. The lunch time hour reading could be telling you to go long, when a relatively small number of trades in the last few minutes massively skew the reading. Deephindsight, this is where your system may have broken down. Think of total volume as total force. The higher the volume of contracts in a given period the higher the volatility. hence the reason morning and afternoon have the highest volatility and the highest opportunity for profit. Time is not important only force, the more force the faster and deeper the move. For this reason, it's important to have at least three volume lengths all moving in the same direction, to confirm a trade. And they all have to represent a figure that can remove immediate noise. These lengths will be determined by the average daily volume traded in the particular market your in. Obviously the S&P would have a higher volume than say, pork bellies, so a higher number of contracts will be used in your measures for S&P than pork bellies. The numbers you arrive at will be different for the time frame you choose to trade. If you wish to scalp, then 1000, 5,000 and 20,000 might be good volume figures for the eminis. I don't scalp so Im not sure on exact numbers. When the market reaches critical support or resistance, the indicator will clearly tell you if their is plenty of force to take you through those points. For example if you reach resistance and the two longer term indicators are positive and your shortest indicator is even higher than the longer term ones, then you will probably crash straight through resistance and keep going. This indicator is the ultimate divergence indicator. It never lies. If you wish to see the ultimate example of divergence, then I encourage you to look at the futures price x volume data for the September 17th to 22nd period last year. The divergence allowed you to pick the bottom almost to the point. You will need Course of Sales data to calculate the indicators. Good luck, let me know how you get on. And don't tell anyone I told you. Runningbear
I've spent most of my Sunday programming and reviewing different mathematics for various moving averages. I am code naming my program "Jasper." So far, Jasper is completely hooked into esignals feed and is able to record real-time transactions as they occur. I've spruced up the user interface a bit and added some custom made graphical buttons to give it that cool streamlined feel. I've also put in some data windows and I'm thinking about going overboard and incorporating a "Three Dimensional World" graph of ES movements where you can fly around the graph and look at it from various angles. I have no idea what purpose that would serve, but it would be impressive to view it. aphie