Thank you ofthomas. Absolutely right. The whole concept does apply to funds or large private investors. Since we are exploiting an edge of statistical nature, essentially based on a large number of transactions, I believe it does not make any sense to try to use this approach to trade small accounts or just a couple of instruments. In that case it's just fine to trade manually, using a good logic. Unfortunately many people dreaming about getting rich by trading are often undercapitalized. It's really hard (and inefficient) to try to do this with small capital. Just to give an example today I remained "stuck" and had to give up a very nice move (probably 2K missed) on UGAZ due to order rejection due to margin requirements. Also an higher frequency would allow to greatly reduce commissions: https://www.interactivebrokers.com/en/?f=commission&p=stocks1 And we are trading not a small account and using portfolio margin. Imagine what would happen with Reg T and a small account. However portfolio margin itself requires at least 110K just to be enabled: https://www.interactivebrokers.com/en/?f=/en/software/pdfhighlights/PDF-PortfolioMargin.php So people contact me on a daily basis asking if they can trade with small amount like 20-50K or so (and some want even trade futures btw!). Or if they can trade FX. It takes a lot of patience to repeat always the same things and, further, often they get pretty upset. The problem I had with UGAZ today, actually happened with FAS too last week. Now, in order to avoid that I just (right now) added a new feature, so at least we can still grab part of the move. Now, when I receive the rejection message, I take note of the side and size, and set a flag to reduce the order size until the order can be executed. Then, on fill, I reset it. This should avoid getting stuck on a nice move. Yep, good idea to compare against S&P or whatever benchmarks. Well, that needs not to be done "realtime" however (each new curve processed does suck resources). The PNL series can be exported (eg., excel) and then one can play with it offline, maybe at the end of the year. (I have personally some reservation about the use of S&P as some meaningful "benchmark" for algorithmic trading strategies, but never mind... anyone is free to use what he likes better.)
if the account was $151K and you are using PM, you should have close to a $1MM for buying power... assuming you are conservative and not use it all... .5MM... anyhow, perhaps the position sizing algo needs to be looked at to ensure that one is not trading too big of a size... as to benchmark... while most people go against S&P, they also use the different MSCI indices depending on the fund focus...
>for buying power ... What matters here are margins. If you play a little with this kind of (statistical) approach, you readily see how easy it is to hit the "margin wall" with a small account (especially using several layers). Realistically, I don't think one would be able to trade the ETFs in the RegT regime. Clearly, then there are the futures. But, then, there is a "granularity" problem. In the sense that the value of each contract may be way too big for the account size (in relative terms), in a decently diversified folio. I'd like to hear more about industry benchmark and why the S&P would ever make sense for a strategic comparison. The problem is that the "underlying strategy" does not seem make much sense in itself, and therefore also quite strange that one would compare an algorithmic strategy to an arguably meaningless strategy. But of course, I do get that there could be valid arguments in favor and I will be happy to hear about them. Since I am a complete "outsider" (coming from an academic environment) I often find myself like "coming from another planet" in front of concepts that instead are common places in the so-called "industry". Clearly, it goes the other way too: often people have hard time to see a new proposed perspective, which has not absorbed the most "popular" concepts and tries to reason merely based on what seems logic, and on the actual effectiveness and results.
benchmark wise, one is either compared to buy-n-hold or against peers... so you could look at any three depending on the type of fund... broad market (S&P), MSCI Country Index, Hedge Fund Index... that is what I've seen for funds... lastly, for margin calls using PM, it comes down to position concentration... and I would think, continuing with the assumption of trading a fund, that you want to look at VaR to model any possible margin call...
Thank you ofthomas, that is very interesting, and actually makes more sense to me. (As to the margin issue, I was not referring to margin calls. What happens is that when you make and order, IB checks the "initial margin", and even if you exceed 1 cent, the order is rejected entirely. So anyway there is the need to automatically correct the order size, to avoid "stalling" an instrument. It's just a momentary issue, because margin usage can vary enormously from a minute to another, depending on the hedging orders.)
End of week. Today we recovered some the recent DD, incessantly scalping on various layers. This week was productive. I have added several new features and improvements and pretty happy with them. I am working on some conceptual refinements of the scalping/hedging games which seem to yield superior performances. Here are a couple pictures summarizing the current situation. I made some clean up, removing a few layers with (all) closed players. <img src="http://www.elitetrader.com/vb/attachment.php?s=&postid=3929448" /> The removed layers are: <code> FAS OPT 20131122 81 C SMART 100 PNL: 236.93 FAS OPT 20131122 88 C SMART 100 PNL: -207.65 UVXY STK SMART PNL: 20.39 UGAZ STK SMART_L1 PNL: 600.12 TVIX STK SMART_L1 PNL: 700.19 FAZ STK SMART_L2 PNL: 323.66 FAZ STK SMART_L1 PNL: 286.37 FAZ STK SMART PNL: 532.76 FAS OPT 20140417 85 C SMART 100 PNL: 410.49 FAS OPT 20140417 72 P SMART 100 PNL: 1,975.09 FAS OPT 20140131 84 P SMART 100 PNL: 64.22 FAS OPT 20140131 75 P SMART 100 PNL: 31.04 XOP STK SMART PNL: 38.63 XLB STK ARCA PNL: 14.47 XHB STK ARCA PNL: 40.43 VXX STK SMART_USD PNL: 41.18 UVXY STK SMART PNL: 206.42 TQQQ STK SMART PNL: 195.70 SQQQ STK SMART_L1 PNL: 26.52 SQQQ STK SMART PNL: 22.32 QLD STK SMART PNL: 17.84 PNR STK SMART_USD PNL: 6.98 OIH STK SMART PNL: 5.08 IYR STK SMART_L1 PNL: 12.26 GASX STK SMART_L3 PNL: 479.81 FAS STK SMART_USD PNL: 28.43 ERX STK SMART_L2 PNL: 222.30 EEM STK ARCA_USD_L1 PNL: 6.62 DIA STK SMART PNL: 5.02 CVX STK SMART PNL: 9.82 </code>
For those who are not trading along, here is some pictures to give a flavor of the (current) scalping/hedging game. This is the most successful layers we have currently, showing the order cloud generated by the game applied on the instruments <b>ZSL</b> [UltraShort Silver]: http://finance.yahoo.com/q?s=ZSL This is very volatile instruments with the dynamics of an ultrashort (distance between the green lines represents a 1% move). The top picture shows all orders (after a while it becomes pretty crowded). The second picture shows the <b>open players</b> (this is the "view" I normally use, it's not selected by default.) You can see, while the price seems currently go down, some BUY players have remained "stranded" above. Why there are BUY players over there ? That is the result of the <b>automatic hedging</b>, when the app is protecting the short position, of course some of the protecting players will remain stuck open. That is actually <b>what give us edge</b> (the very reason), in fact while they <b>make the unrealized component run almost flat</b> (or at least actively bounded), they are systematically recovered whenever possible, this feeding the statistical drift. This way there is no need to "predict" anything. The reason for the edge is tangible and concrete. You can actually stare at it, as it is illustrated by the Gain/Loss components dynamics. And this also provides a great insight to the fund manager, so that he does not get completely lost in the <b>fear of a DD</b> (starting doubting if his "strategy" makes sense at all). I always had hard time to convey the <b>non-prediction</b> concept in the various endless discussions in many other places, it seems that many people interested in trading cannot even conceive the existence of <b>nonpredictive statistical methods</b>. That is kind of weird, because on the other end, I (coming from a different perspective) <b>cannot conceive a statistically significant prediction</b> [based on mkt data alone]. The tikdata <b>interpolation exercise</b>, whatever made (financial models, NN, GA, or whatever) is fine with me, and can even be an interesting descriptive, summarizing, tool of the past, but the <b>next step</b> (extrapolation) is arbitrary and has no scientific ground whatsoever, as far as I am concerned. Actually, I would argue that there are very good arguments while even "a priori", it makes no sense at all, let alone the possibility to prove it scientifically. (In addition to all that there is also the MMs algorithmic component, which is a presence actively working (statistically of course) against "you" to dismantle all your "predictions", especially when your orders become relatively significant. You can have an immediate feel of that: try placing some good orders on instruments like UGAZ (1000 or more shares), record the instrument reaction for each order, and let me see what you observe in statistical terms.) <img src="http://www.elitetrader.com/vb/attachment.php?s=&postid=3929605" />
what does the IBKR performance report looks like when compared to the performance of the robot... a lot of us understand that report better, can you post it? it would be great for comparison..
Hi ofthomas, as I have explained in the first post, I am trading the account of a private investor who is kindly sponsoring this strategy demonstration. I have remote access to his machines (running the IB Gateway) via Teamviewer (http://www.teamviewer.com/en/index.aspx), and we are on different continents (and both on continents different from the exchanges!). Clearly, I have no access to account management (nor I would need it). However, if you are interested send me an email with your kind interest and I will forward it to him. What I can make readily available, in addition to the app's reports (it prints also a complete report, in HTML, with all charts), if anyone is interested, is the complete list of all order details (order id, price, qty, time, and mkt info, in excel format). From there you can create <b>any</b> report you like. If anyone is interested, drop me a PM with your email. (Clearly, IB figures are identical to the app. The app is <b>accurate to the penny</b>, at any time.)
Very interesting, but still not sure if I understand this system. My understanding is that you do not take losses, but rather hedge them with correlated instruments and wait until price comes back your way to close the position for a profit. Is this correct or?