Not any more I don't. And I still feel the same way. The new data is more 'raw', and raw data is better than cooked data any day. Yes, that means that people that 'pulse' lots of small orders to hide their size need to be dealt with. And, they can, since no one can stop me from re-bundling what they've unbundled.
Just because the game changes, there is no reason to complain as so many have done. The markets change and traders must adapt.
The following 4 users say Thank You to Richard for this post:
Your example would increase the trade size from 50 to 500. What I have shown is that the avg order size has decreased by roughly 50% (maybe 2.5x).
So what it'd show now, using your original example, 5 orders to buy 25 shares at $10 and 5 orders to buy 25 shares at $10. In this case you do not know the two orders came from the same trader, and both orders do not meet the volume split minimum of say 50 shares. So your splitter will show two "dumb" transactions and zero "smart" transactions.
I think the correct way to use volume is to compare the volume at the bid & ask. 50 shares is 50 shares, doesn't matter if they come from the same trader or 50 traders. In the market it's the order flow that moves price, not the trade size.
To place more importance on some transactions than others is to assume way too much about those transactions. As I mentioned in a previous post. the S&P is widely used for hedging. Someone could take off a spread or a hedge and you'd see a big transaction. That doesn't mean that the trader had a directional bias on the S&P.
Also the large trader categories are separated into commercial hedgers & speculators. They do not trade together inside each category. The early (smart) speculators are now selling to the late-comer (dumb) speculators. So even in a large trader group there is "smart" and "dumb".
Can you explain how the data was cooked before this change and how it's more raw now? I have the opposite opinion. I have the impression that orders were not bundled before and now they're broken up into smaller pieces.
What I tried to explain to people many times on traderslaboratory is that predicting turning points is not what the splitter is all about. I think you were even participating during that time, cunparis. As larger orders are the majority of the volume anyway, the delta profile of the larger orders looks nearly identical to the overall "unsplit" delta. The most useful analysis also included the intensity of the large order activity (the thickness of the splitter line) in combination with what price did (the old VSA cliche: effort vs. results).
I mean, really, the majority of accumulation and distribution logically happens on the limit-order side of the trade anyway, which isn't what delta indicators measure. So, it's just not a proxy for actual big trader activity, no matter how much people want it to be.
But no one paid attention to me then and I don't expect anyone to listen to me now. Looking for a magic turning point predictor indicator is a fruitless task that no doubt is why so many people get sour on indicators.
The following 5 users say Thank You to Richard for this post:
The CME used to see 1000 1-contract orders pulsed in rapid succession for example, and they would bundle it and report a single 1000 contract trade. They were cooking the data and giving you a summary. Now, they don't bundle those trades anymore, and just give you the data raw.
It's not that the CME takes in a market order for 1000 contracts, and then reports 1000 1-contract trades. At least, that's not the way I read the news. (although, even if they did break up large trades, I still wouldn't cry about it because you would see 1000 1-contract orders coming across your datafeed in very short timeframe, and could "re-bundle" it yourself with ease)
The following 2 users say Thank You to Richard for this post:
Yes I was participating in that thread. I used some of the code to write a volume splitter for Tradestation and also adapted my volume indicators to use only large trade sizes. I found what you have stated, that the majority of the time the large trader activity mimicks the small 1 lot trader activity. And sometimes when one lead another, the small traders would be more correct and sometimes the other large traders would be more correct.
In theory I said it should not work because of the reasons I've mentioned in this thread and others. But I wanted to prove it to myself. So I did. And I found zero edge with it. Maybe it's something everyone needs to prove to themselves. And if that's the case I don't mind helping out by asking for a few days of charts.
Last edited by cunparis; December 5th, 2009 at 01:51 AM.
Reason: Sorry I didn't see that you replied in two posts..
If by 'correct' you mean pointing in the direction price headed next, then to me that's still expecting delta to lead price enough to make money following it. But, delta indicators will never be that kind of leading indication.
But, just like a hammer sucks at making toast, and a toaster sucks at driving nails, that does not mean they are useless tools. It just means you need to use them for their intended purpose. Which in the splitter's case, no one ever seems to want to do.
The following user says Thank You to Richard for this post: