Hi Sergio77 Personally, I tend to use these 2 words interchangeably in this context. (But keep in mind I am not a native English speaker: if there are differences I'd like to learn, clearly).
This strategy looks very much like a volatility trading strategy. Saying you're aren't trying to predict volatility is naive.
Hi jb514, if you think I am "predicting" volatility, you are, unfortunately, missing the main idea here. And missing by far. Of course, you are entitled to look at it as you wish and prefer, just, being the author, I think it could be useful to consider the authentic interpretation.
Here is a quick folio update. All proceeding according to plan. We are smoothly growing the G-L, while keeping in good check possible runaways, through hedging entries. The drift (G-L)/L has been decent so far. Today, metals have been loading a bit. (This second session is actually overall significantly "smoother" then previous one, due to better and more "mature" games.) <img src="http://www.elitetrader.com/vb/attachment.php?attachmentid=144289&stc=1&d=1395087447" />
Let me repeat again what is the reason of the edge here, so we don't lose the sense of what we are doing. Here, as in our "reality", all is about information. The idea is all in the fact that we are lowering the system entropy, respect to the naive "stop and forget" approach (which is the reason of the "strategic dominance"). Trust me, there is really no way to survive that kind of systematic <b>loss of trading information</b> and both practice in real trading and well designed simulation environments show that very clearly, and beyond any doubt. Let me make another attempt at yet another pictorial illustration. A couple posts ago (page 31) I showed a trading chart of VXX. (That was, and is, real money and real time.) I showed both open players and actual orders Let's take another look at it now, a 4 days later, and compare the charts. This is the real time situation of the open players. You can see now that some of the open buy players (the 3 squares on the right) are gone. Where are them ? Of course they have been closed, because the price went up and they could be "resolved" This is an example of "loss recovery". Those orders which served as stops, now have been "recovered" and they contribute now to the positive drift. It should be pretty clear that if in the trading "game of chance", where you statistically are 50/50 to lose or gain, if I "give you back" some of your losses, you will start "drifting" and have an edge (which of course you can also spoil through reckless trading, technical errors or whatever, but which nevertheless is there and is real, tangible and keeps actively acting). [Clearly, when we talk of "<b>conservation of the trading information</b>", it should be obvious that we mean conservation <b>and use</b>. In fact, if we merely kept the information, without actually using it in the scalping/hedging game, it would be equivalent to "forget" it (and therefore offer no advantage). ] <img src="http://www.elitetrader.com/vb/attachment.php?attachmentid=144293&stc=1&d=1395090780" />
So maybe not predicting, but certainly it is meaningful to your strategy. In your detailed answer to my last post (thank you ), you mentioned live, "current" volatility. How do you measure that? It seems obvious that it's a trailing window, but how big? 20 minutes? 20 days? Also, do you use one systemwide volatility level, or is it instrument specific? I assume the vol adjustment affects hedging "tightness" but not trade size, since changing trade sizes across instruments might effect the natural hedging relationships from diversification... Also, to which instrument do we owe the recent nice PNL uptick? Thanks!
> certainly it is meaningful to your strategy That's right monkeyjoe. Volatility is certainly meaningful, as we do need the price to move in order to close our players (or, say, our "scalps"). In fact, when doing this in practice it appears also evident that some instruments are more suitable than others. For instance, the leverage of the ETFs (and futures) helps because it "amplifies the fluctuations" and helps getting out from a larger number of scalps. With single, non leveraged stocks, you may find yourself in larger and longer drawdown, due to much less "noise" captured and the need to wait for *much* longer "waves". > "current" volatility. How do you measure that? It seems obvious that it's a trailing window, but how big? 20 minutes? Right. One can simply apply the volatility definition on a recent chunk of tickdata (and annualize it for ease of comparison). The trailing window can be varied at will by the fund manager. I am currently using a periodically maintained "trailing" sequence of recent prices, where the sampling happens about every 10 minutes and the prices "go back" about 500 observations (so say about 3 trading days.) > Also, do you use one system wide volatility level, or is it instrument specific? The usage of the volatility is to provide a quantity (along with the tick value) to "shrink" or "inflate" the order cloud, so that we have a statistical "scale invariance". In other word, while the "shape of the orders cloud" is due to our game rules, it's "size", we might say, must go together with volatility (and tick value), or else for some more volatile instruments you might have way too much activity compared with the less volatile ones, and this would somehow defy the purpose of having a folio to diversify. But also within the same instrument, we can have times of higher/smaller volatility, and we still wish to dynamically adjust the game to that. (Actually, I have in mind to create a more comprehensive and accurate indicator to "govern the order cloud scaling", but for now, let's say that, for the moment, it's ok.) <img src="http://www.elitetrader.com/vb/attachment.php?attachmentid=144304&stc=1&d=1395130433" /> > I assume the vol adjustment affects hedging "tightness" but not trade size, since changing trade sizes across instruments might effect the natural hedging relationships from diversification... That can be chosen by the fund manager, in the game rules. I currently adjust the whole cloud. Which means that "scalp size", trailing sizes, (along with entry spacing, etc.) is also scaled dynamically with volatility. When I talk of "scaling", in particular I do not refer to "absolute sizes", but to % variations of the price. >Also, to which instrument do we owe the recent nice PNL uptick? Comparing the charts, you can see that mostly it was DGAZ, which had been "loading up" position for a while, and was able to close some scalps. Also TNA contributed a little bit. (Currently we are "loading" mostly on volatility instruments (vxx) and metals. )
<b>> For instance if I take the latest open order and single player graph (it is two graphs above), could you explain moving from left to right the logic used by Gbot in making the opening, closing, hedging and so on for each of the points in there?</b> Thank you very much AItrader. I am not sure which chart you are referring to precisely; anyway, in general terms, any action such as to open a player, close it, etc. is specified within a set of rules, which the fund manager can change. These are the "<b>games rules</b>" and they govern the creation of the order cloud. The only purpose of the game rules is to favor the grow of the G-L (net gain component) and keep under check the unrealized. The G-L drift essentially arises from the architecture which is continuously using the past trading information to make available a drift. <b>>how much capital is currently being traded by G-bot? Not the total in the account, but that actually used in buying assets.</b> In this second session margin usage has been mostly around 40K. Now is about 50K. The maximum peaks for now have been around the end of last week (touching about 60K on this instance). Soon we will be increasing this figure. <b>>It summary I believe that for a human being to be able to understand the Gbot behavior, she/he should be able to simulate by hand step by step what is happening on the graphs that you are reporting.</b> That's right. In fact the simulation environment allows the fund manager to examine all players and entry one by one, blocking the flow of trading at each order, and studying the situation (see the check boxes on top of the simulation interface). On the other hand, it should not be a matter of understanding the application behavior, but just the way round: it's the fund manager which must be telling the application what to do, in order to realize an automatic management of the risk, and of the other various aspects of the trading activity.
End of the week. Here is a quick update on the situation. Everything ok so far, we continue to increase the "net gain component", with a decent drift. Last 2 days have seen some massive "load up" on metals. Also gas (dgaz) has a considerable position. In particular, having several metal instruments (DGLD, DUST, JDST, SLV, ZSL) in the folio, the effect is quite marked. JDST, is definitely the bigger "roller coaster", with huge volatility (price ranged from about 14 to about 23 in very short time, about 64%). So, if you use it, be ready for large swings (DUST is also quite volatile). I am planning next week to upload (and redistribute) an update, where I have introduced a new useful mechanism I will discuss in next posts. <img src="http://www.elitetrader.com/vb/attachment.php?attachmentid=144515&stc=1&d=1395437027" />