BlowFish

The thing is the short term trades (if they are trend follower too) will have that similar 60% 65% figure. Of course if you are talking about mixing another strategy with trend following to smooth the equity curve that is a different matter I dunno I think there are fairly fewer ways to 'do' trend following.. If you are taking contracts off along the way I would evaluate them as different trade types (that just happen employ the same entry setup).

My emphasis. Fulfilling the client order may leave you holding some 'working' inventory or even taking some of the clients inventory just to finish their order.

Slightly off topic. I see you mention FDAX which is one of my favourite instruments. I have had problems with DAX in the past. In a nutshell I think it is to do with block trades.Every now and then you will see a trade go off for 1k or even 10k contracts. The depth of the DAX is usulally lower double digits. Eurex has a mechanism where participants can negotiate large blocks outside the auction though they are still reported on the tape, I am pretty sure that is what these are. After all if best ask has 28 offered and 1500 goes off and there is still 28 offered......well makes sense to me. Actually I think I have answered my own question. (what to do about these blocks) It is impossible to tell the direction of these trades (from a delta point of view). So the only sensible solution is to exclude them from delta accumulation? I hate filtering valid data but assigning them incorrectly would not be desirable. Of course not assigning them has implications two. Actually the fact the trade was negotiated suggests that neither participant required immediacy.

The problem is that my testing shows that TS & MC are not the best platforms for this type of work. In a nutshell the mechanisms they use to get bid and ask (insidebid,insideask) do not return the data synchronously. They return the current bid and ask not the value that occurred at the time of the tick that you are processing.

You know what it's on the to do list! I am not sure what it will show but I figure it might provide some pretty interesting metrics! With my renewed interest in 'delta stuff' (3rd time round now) my primary interest is selecting robust tools and data to work with. I guess the obvious choice is IRT/DTNIQ. Neoticker in the spirit of not being obvious Neoticker would fit the bill too it's whole architecture lends itself to this sort of work.

Hi DD, Most simple trend following systems have a general characteristic of a fairly low % of winners. Of course you could add all sorts of filters discretion or what have you to try to improve this. Tough job that. However there is an easier way to deal with this. Sweeping generalisation number two most trend following systems utilise fairly aggressive pyramiding of winning positions to produce acceptable returns. Without this approach to MM they are unlikely to perform well. I should say so far I have only read your original post.......you have been busy this weekend!

Absolutely agree with you there:) Identifying the event and being confident to act on it is all that is important! The rest is idle banter really.

Just going back to flog the dead hedge horse some more:) To those not interested who was at work and why skip to the next post.BTW I am certainly not questioning that someone was getting busy. FT, I wonder if it might be that we have different definitions of what a hedge is. It is my understanding that by definition a hedge is taking a position in an instrument to offset risk. That risk might be in physical goods (or lack of them), another instrument, or even a basket of stocks. In the case of the latter the bet is that the basket will go up. If yo suspect the broad market might not go up you can hedge the position, now you can make make money whichever direction the market goes (providing your basket out performs the broad market). The money is still made on the basket not on the hedge, it can now be made in down markets too. The key thing is not the market dropping its that your stocks don't drop as far. From what you have said it sounds to me that it was essentially a speculative position on the broad market dropping. Maybe I am being pedantic (my girlfriend tells me I am getting increasingly *n*l) but if they are making money on the 'hedge' part of the position its not a simple hedge any more. As an aside If you wanted to hedge your basket (still betting it will outperform) and speculate on the general market dropping then a non linear hedge (buy calls) would be a better option (if you pardon the pun) than a linear hedge(futures).

I guess I could have rolled this into the previous post but separate train of thought. Trouble with arbs is there are so many types. For this I guess it's Ok to leave aside the esoteric types and focus on the S&P. I guess that means we are looking at the cash as the reference instrument? That would make the futures and or the options the hedge. Lets leave aside options for now. I find it hard enough to think about 2 things at once let alone 3! Can we consider arbs in a similar way to the traders we have discussed?....heres my thinking. Arbs will essentially try to provide liquidity in the markets they trade as they are legging in and out of positions. The cash position (being the reference instrument) may leave our Arb requiring inventory for there hedge. Like the trader executing for the elephant they might need to take liquidity form the hedging market (the futures) to balance out there hedging portfolio. Does this sound plausible? I think I can see the pieces of the puzzle sitting there on the table but I am not sure I am closer to putting them together! (i.e detecting things) Might be time to start drawing boxes and arrows.

Do you use other traditional market internal type indicators in the correlated market to try to help detect when these are going off e.g. Prem Tick Tikki etc.? I guess you must try to identify the arbitrage opportunity to get in with/ahead of these programs? Incidentally arbs use the concept of 'value' but not in an MP sense, there assumption is the basis (difference in prices of the components of the hedge portfolio) will return to 'fair value'.

FT thank you for your reply (and your ongoing interesting input). For me it is important to separate the quantitative from the qualitative. Personally I attach much more significance to the former, seems like the only prudent course of action to me

Nothing wrong with being a sceptic! It boggles my mind that people expect to trade consistently based on something that they have been told is the 'truth'. With little or no verification for themselves. My next step is looking at the tools and data I use. Last last time round I was not entirely satisfied with their integrity. That's probably a topic for another thread though.Actually I wonder whether there might have been bias on my part precisely because I was wary of the core assumption!!

In that situation it seems sensible to assign the portion that executes immediately as taking liquidity and the balance as providing it (if it executes). There are numerous situations like this for example when two non-marketable limit orders are crossed within the spread, then both orders could effectively provide liquidity but, because of their timing, the second one reduces liquidity. I think it is probably academic as it will all depend on how the exchange reports the trade.

That's pretty much the standard way of doing it with the participants requiring immediacy (market orders) taking liquidity from those offering liquidity (limit orders).

Exactly! Though I still plan to conduct my own straw pole next week, just for fun. Whilst I find the work of people like FT pretty intresting, I did have some difficulty with the core assumption. (That 'delta' is an adequate proxy for order flow) In fact that was what prompted me to look further into the mater. I should point out most of the work that is published has been done by 'academics' rather than industry pundits. (Somehow I can't imagine GS publishing things like this)! Now I have to say I am equally as suspicious of 'academics' but the reports are thorough with detailed notes on the numbers and the methodology used to process them. There are also numerous independent reports with (slightly) varying results but similar conclusions. There is one particular conclusions that I have made (your mileage may vary). In a nutshell I have to concede that market delta type calculations are a reasonable proxy for order flow. I don't think I was wrong be be sceptical, I now have a firmer view based on what I consider an acceptable body of evidence. With regard to this thread, I would now accept that there was sell side order flow, however I remain sceptical that you can detect that a) it was a hedge and b) whether it was speculators (or what FT would call commercials). I am not saying it was or was not just that I remain sceptical that you can detect that . What is more important of course was whether it provided a quantifiable trading opportunity.

The commitment of traders is about as good as it gets CFTC Commitments of Traders Long Report - CME (Futures Only) Edit worth mentioning that the categories of traders in the report as defined by the CFTC are somewhat different to what FT means by 'commercials'. Commercials would actually be reporting non-commercials in the report. Non-Commercial (”Large Speculators”) These large speculators are mainly hedge funds, banks etc who trade futures just for speculation. Commercial (”Hedgers, Exporters, Importers”) These are people who use the futures contracts for hedging purposes, and these commercial participants are generally producers, farmers, factoris, exporters and importers etc. They use the commodity or futures markets to take a position that will reduce the risk of financial loss in their assets due to a change in price. Non-Reportable (”Small Traders”)

I guess as a bit of fun these things are OK but they can encourage pretty bad habits as you point out (using maximum leverage on the most volatile instruments whilst being as close to always in as you can). There are a whole bunch of 'trading gurus' that started their careers in this fashion.

I think there are some subtleties in JR's work that are easy to miss. These are essentially to do with 'context' though I don't think he uses that term. He doesn't seem to talk about 'S/R' either. He considers things from the point of view of 'where are the orders'. Incidentally it was this way of thinking that lead me me to the idea that S/R could be considered as a 'magnet' attracting price rather than simply as a barrier that price bounces off. This is not discussed in his free material as far as I am aware. He also has pretty robust definitions for congestion and cautions on taking trades there. As to the stop my understanding is that it is initially under the lowest bar of the hook, brought to BE after P1 is hit.

The obvious answer (well to me so it might be half baked) is that they would try to offset the position to other side liquidity takers. If that is not possible they would demand inventory from other side liquidity providers. I guess one might assume the very fact they have ended up with an inventory imbalance would mean they might swiftly move to step two, demanding liquidity? Isn't this a fundamental issue when there is an intermediary the intent liquidity requirements of the original participant is obscured?

Ahh I see, I did not realise there was an 'indicator' column further investigation of other flags that FIX provides reveal.... The Price column displays the price at which the transaction occurred. Some prices are followed by an "A" after the price indicating an offer/ask price which occurred at or below the previous last. A "B" after the price indicates a bid which occurred at or above the previous last. Prior to regular session hours, an indicative price may be displayed with a volume of zero in the size column. I presume you look at A or B to determine whether buy side/sell side liquidity has been taken?

I think I am right in saying "provided the skew is close to zero and the distribution looks symmetric". Looking at larger data sets to provide context for the scalps looks promising to me too, similar to the ideas Jerry presents in the HUP's thread. After all on a 15second chart todays VWAP SD's and PVP are essentially HUP's.

Your right the studies suggest that the algorithms are 70% accurate detecting order flow (nearer 80 actually) that is subtly different to saying that 70% of commercials order flow is at market. I will have the opportunity to do my own straw poll over drinks next week. Actually talk of 'commercials' can confuse the issue, order flow is order flow at the end of day and as the 'commercials' are by far the largest participants there is a good chance there will be a 'commercial' on both sides of the trade.

Pardon my ignorance but could you clarify what you mean by indicative? My understanding is that the term would be used for something like spot FX where you have access to a subset of the data sources and get quotes indicative of the average. My goal is to make some progress towards a working algorithm! And I absolutely agree about matching time and sales with current quotes. Unless you are provided with quote changes absolutely synchronously with T&S that path is fraught with issues (as I have discovered)!

There has been a lot of research on this subject. What can be shown is that the Lee and Ready algorithm (which is essentially using volume@bid/ask to classify trade direction) is about 70-80% accurate. It can be shown that the most inaccuracies creep in when trades are between best bid best ask, or outside best bid best ask. Ellis, O’Hara, and Michaely proposed some modifications and the accuracy rises a few percent. Most studies use the NYSE TORQ database for research as it has an audit trail that can be used to establish trade direction. A positive side effect of the old specialist system! Ellis, O’Hara, and Michaely did some work on the NASDAQ based on a proprietary data set provided by NASD. If you search around you can find several papers on the subject.