Here is a quick update about the situation. The G-L drift has been working ok so far. We have started taking profit on the gas instruments (dgaz), while most of the investment is currently in metals. In particular, JDST (very volatile) has absorbed quite a bit of it. On the other hand, the scalping action we have been carrying out is sustaining the "investment", limiting at very acceptable levels the DD. So far the DD has never been particularly worrying (we have been in the range of 10-15% of the margins, which most often have been fluctuating in the range 40K-80K for this instance.) Two of the 4 layers of TNA have gone in positive (over 1K each) and I have flattened and put them in "manual mode" (to "suspend" them): the remaining 2 are more than enough <img src="http://www.elitetrader.com/vb/attachment.php?attachmentid=144630&stc=1&d=1395790873" />
Hi fullautotrading, could you please comment on the tendency toward loss of the profit / blue line that is displayed in the figure? I would expect that the profit/blue line would display either a positive drift or an horizontal drift with larger or more frequent peaks toward positive values. Also I would expect that the height of the peaks would increase as the use of the margin increases. Instead it seems that the profit line is not altered by the amount of capital invested? If we look at the Unrealized Gain and Unrealized Loss lines their spread increases over time. Which makes sense in terms of an increase in the number of open games. This does not seem to impact on the profit line behavior. Do you have any explanation for this phenomena? Thanks!
Tom, Could you also comment on how much commission you are paying? What about fills? ...are you using market orders or limit orders? Given the number of trades, it seems that reduced commission from adding liquidity and trying not to pay the spread might cause the PNL 'drift' to be substantially more positive. mj
Hi AItrader I think that the timespan we are looking at (a few weeks) is clearly "too short" to make evident the macro phenomena you mention. Especially because the variance itself of the underlying price trajectories is large, and that may definitely "mask" them on a single "realization" (a sample path) and relatively short timespan. In general terms, we have the PNL (blue line) which can be seen as follows: PNL = (G + L) + Unr What here is "drifting" is the part (G + L). The other part, the Unr, depends on: - Price curves (usually the more volatile they are, the higher would be the Unr) - Size and direction of orders used to trade We do not have, obviously, control on the price trajectories, and what we can do is to intervene on the order sizes and order directions (buy/sell), in order to fight the growth of the Unr, so that, in the meantime, the (G + L) can work steadily "pulling up" the PNL. The Unr somehow acts as a (significant) "random component" which (unavoidably) comes to "disturb" (or, actually, actively "fight"), mostly downward (because the positive scalps are closed and go in the Gain component), the drift we establish. It's pretty analogous to when you generate a GBM with a given drift. Irrelevant of the its amount, you can of course still be pulled "down" (or further up, in that case) by the accidental component over a given timespan. (This is more evident when the volatility is high.) In our case, the effect is even more marked on the downside because essentially the Unr is (practically) always negative. Clearly the drift will still be "acting" in statistical terms, but you might not be able to "see" it in a given realization or on a limited timespan (that is why I plot the <b>"G-L" curve</b> as strong visual aid to the manager about the scalping progress). See for instance this picture: linked from this page; or this picture: from this</a> page In general, larger is the volatility, more you can deviate from the avg behavior. Clearly, just as the so called Kelly criterion suggests, it's not enough that you have something (some advantage) working in your favor (for instance a probability, or a drift, or whatever) but it also necessary to keep under check the rate of growth of the accidental component (Unr), to avoid "blowing up" before the effects of the "advantage" start have significance. The fact that the PNL has remained almost flat here, still while we are heavily investing in metals is a positive thing. Imagine what we would have if we had not scalped away those 30K of the "G-L" (denoting the sum of the G, L components). Those would "show" in the PNL, and it would be way below -30K. So the "G-L" is actual $$$ that we do have made. The fact that the PNL can be "above or under the water", is somehow accidental, depending on volatility, order sizes, etc. (and somehow immaterial once we accepts this logic), especially when starting up a session. It depends on how much we invest of the available $$$ and how much volatility have the instruments. In the long term the objective is obviously to pull up the PNL, which from a certain point on, will possibly remain "above the water" (but clearly always with fluctuations), except for possible periods of "strong investment", where the Unr can even pull it down again. Clearly, we want to keep under check the Unr (and that is the job of the "scalping/hedging game"), so that we allow some new investments, but hopefully it does not escape from our control, so that the drift can keep working.
Hi monkeyjoe, sure with pleasure. The current setup is the so called "cost plus". It's this plan: https://www.interactivebrokers.com/en/?f=commission&p=stocks1 Within this plan we are paying the highest rates, being at the lowest volume levels. [The current session (60 days) numbers are: Fills: 3,296 Volume traded: 304,438 Comms: 2,951.03 ] Our margin scheme is "portfolio margin", see: https://www.interactivebrokers.com/en/?f=margin&p=pmar Orders are always LMT (in case the manager can override this, but there is usually no point). > it seems that reduced commission from adding liquidity and trying not to pay the spread might cause the PNL 'drift' to be substantially more positive Right, with a larger fund you would be paying much "less" in commissions (in relative terms, clearly) due to both larger orders and larger volume. (As a "mkt taker" you will be paying the spread.) (What bothers me more are actually the interests paid (short positions) )
First day of the week. We are still "loading up" on metals. To ease a bit the situation I have "flattened" a few layers which were in profit, just in case I need some extra margins later to hold on the current investment: DGAZ 1.3K SCO (flattened because I don't like the instrument) TNA 1K TNA 1.5K UVXY (previously flattened because useless, having VXX) Drift is fine. "G-L" figure is almost around 40K, which is a good "cushion" for this phase. Margins used are about 90K. For the short position interest, the details would require access to Account Management. Anyway, the fees can be found through the "Tools" item of the account management menu. Some more info are : "Interest credit on short stock proceeds" http://ibkb.interactivebrokers.com/article/41 "The Short Stock Availability Tool" http://ibkb.interactivebrokers.com/video/1073 The fees are published on <b>Support > Tools > SLB</b> Here are some examples I found there: DGAZ @ ARCA Current Rebate Rate* -4.5942% Current Fee Rate** 4.6342% DGLD @ ARCA Current Rebate Rate* -4.835% Current Fee Rate** 4.875% DUST @ ARCA Current Rebate Rate* -5.46% Current Fee Rate** 5.5% GASX @ ARCA Current Rebate Rate* -8.835% Current Fee Rate** 8.875% GLL @ ARCA Current Rebate Rate* -5.085% Current Fee Rate** 5.125% JDST @ ARCA Current Rebate Rate* -3.854% Current Fee Rate** 3.894% QID @ ARCA Current Rebate Rate* -2.085% Current Fee Rate** 2.125% UCO @ ARCA Current Rebate Rate* -4.21% Current Fee Rate** 4.25% TBT @ ARCA Current Rebate Rate* -1.335% Current Fee Rate** 1.375% VXX @ ARCA Current Rebate Rate* -2.585% Current Fee Rate** 2.625% UVXY @ ARCA Current Rebate Rate* -6.71% Current Fee Rate** 6.75% ZSL @ ARCA Current Rebate Rate* -11.19% Current Fee Rate** 11.23% SDS @ ARCA Current Rebate Rate* -1.5432% Current Fee Rate** 1.5832% (the picture is still not entirely clear to me. For instance, what happens when one holds a position for a few seconds or minutes and what are the actual details of computations.) If anyone knows more details, of course it would be very nice to know them. It would also be nice and transparent that the details about interests, for each and every instrument, would be provided <b>via API</b>. For instance the cumulative interests from a date, the average %s, and so on.
The short fee is charged daily by IB. So if the fee is 3% then you will be charged (3 / 365) each day. In addition you have margin interest, which is also charged daily. I remember studying this extensiely some time back were I thought I had found a decent strategy, but it was proven to be extremely sensitive to interest charges & transaction costs so I had to abandon it. Those expenses can rack up pretty quickly in my experience.
End of the week. Here is a quick update on the situation. Everything ok so far, we continue to increase the "net gain component" keeping a good drift. We still have a relatively "large" amount $$$ tied with metals, which is however being sustained well by the gain component "G-L" (which is now about 1/3 of the peak margins used). I will probably get rid of ZSL due to the interest figures (and, more often than other ETFs, pretty wide spread) and probably keep DUST and JDST. JDST is indeed *very* volatile, but it's a kind of "good" volatility that so far I don't dislike at all. Hi TskTsk, thank you for the info and your experience. There is no doubt that interests are something that we all would rather pay less, just as commissions. In general, the larger the capital traded, and the better is the treatment from broker, so that trading expenses become relatively smaller for larger funds. Anyway, I still rather prefer to use these instruments and pay the relative fees (of course the smaller, the better), as I feel that it's much more comfortable to trade them (and also having on your side the long term decay). On the other hand, also with futures, which would require a much larger capital anyway, some $$$ is eroded by contango, etc. After all, the achievement of <b>consistently profitable</b> trading methodologies is so hard, because we not only need to have some source of real edge, keep carefully hedging, etc. but we must also continuously fight against steady and inexorable fees which do stack up against us. With this kind of capital, order size and frequency, we are practically in the "largest fees" zone. Worst than this would be not using the "cost plus" commission plan, and having the RegT margin (which would probably render it a "mission impossible"). After all, when capital is "small", ETFs seem still the best way to go. (I would have a few recommendations about maximizing the scalping action. Maybe one day will make a "list of recommendations" or suggestions, for those interested in this approach.) A few ETFs have not been shortable the last couple days, like ERY and DGLD for instance. When I see instrument in a not shortable state, and look at the "volumes" claimed, I don't know why but I get that weird feeling which makes me associate some of the mkt behavior with the 3 shell game )
A quick folio update. Situation still ok. Metals still need to "come back home". This week I should make some game improvements I have in mind. In particular the possibility to differentiate the maximum order size, for what we have called the "scalping reversing" condition, and for the "hedging-protecting" entries.