How to determine if one player moved a price

I'm trying to understand what caused certain price movements (aren't we all!) in per-minute data for major NYSE stocks. In particular, I'd like to determine whether a given price movement of X% in either direction was due to a single entity making a large trade (maybe their order-splitting algorithms leave traces, or maybe they skipped that and had to buy/sell in a hurry) versus an aggregation of many small orders.

What is the best way to approach this analysis?

-
I guess you have to look at the ladder. But I'm not sure will give you here solutions for your problem – RockScience Apr 12 '11 at 6:29

This could be very difficult to determine in practice, because the axe (who controls the supply and demand) wants to hide his tracks.

Also consider the axe's aliases. I mention this because you would need to take into account the axe disguising his trades through another market maker (for example, Goldman trading through ARCA, or even showing sales between himself and his conferderate).

If it is the player's task to accumulate or distribute a lot of stock, his job will be to do so with as small a change in stock price as is possible. He may achieve this through a number of different methods, including head fakes (selling a small quantity when his direction is to accumulate), patience (accumulating over several days or weeks) or refreshing his bid size.

-
frankly speaking, even without assuming that the other participant is trying to hide his move, do you think that, without calling the broker and or having an insight within the participant, you can algorithmically determine who is trading? or simply that an order is triggered by a single underlying participant? – SRKX Apr 14 '11 at 7:52
@JSmaga, good question. The reason I picked Goldman is because traders recognize Goldman as an axe in many of the large cap, NYSE stocks (the OP's area of interest). They closely monitor the market maker GSCO to attempt to augur GSCO's true intentions. If Level II says GSCO, you can be more certain that it is a single, large participant. This does not hold true if the Level II says FIDC (Fidelity) or any other retail broker. I suppose one way to take my answer is: If the axe can hide his behavior and successfully fool many traders, what hope is there for an algorithm? – rajah9 Apr 14 '11 at 13:14
@rajah9, very interesting it didn't know that. thx – SRKX Apr 14 '11 at 13:24
@rajah9 That's an interesting tip - if comparing e.g. GSCO vs FIDC in the order tells you that trades are more likely by a large participant, that's information. It's not a definitive answer, but it does skew a probability distribution a bit in either direction? – Greg Apr 14 '11 at 18:22
@Greg, yes, I think that if you see GSCO in the Level II trade then you're looking at a large participant. Conversely, I do not know of any cases where the axe will use a retail market maker. (Just thought of a confounding factor: Perhaps Fidelity Mutual Funds use FIDC?) – rajah9 Apr 18 '11 at 13:22

If you look at the ladder, you might have some insight, but it's mainly speculation.

The only way to be really "sure" in my opinion would be to have some insight from a broker.

Otherwise, what I'd try to look for is to recognize execution schemes, but again you have to know the algorithms of all the participants in order to determine "who" it was.

In my opinion, even "fat fingers" are difficult to detect, because I think the algos are reacting to the move in the market (probably using some kind of momentum) and kicking in as well.

-