I think the bid/ask spread is very misleading. For instance ----ASK------ 2.982 100 3.000 900000 3.010 300000 ---BID------ 2.981 150 2.980 800000 Just looking at the best bid/and ask you would think there is a spread of 0.001 But, if you intend to buy 50,000 shares or so the spread is actually 0.02 There is a very simple calculate you can do to walk the book and find the price paid for X number of shares. Why isn't this used to get a better idea of spread than the traditional best ask and bid?
Presuming that in fact your order is in fact next in the queue, then your 50000 shares would be 100@2.982 and 49900@3.000. Interestingly enough, with some patience, often one could simply chop the 50k order into smaller blocks since the bid/ask spread moves alot slower than the price action itself. In other words, there is always new shares @ bid or ask flowing into the depth. To minimize the unnecessary slippage you note, one could patiently process their order in smaller blocks. Sure, there are transaction costs, but depending on the bid/ask liquidity at each depth, this slippage can add up to a formidable chunk of change that would far exceed the aggregation of multiple transactions fees incurred by using smaller blocks/lots...
You're right about that. I guess my general point was, if you look at the spread in this case to estimate transaction costs, you are going to have major slippage. If you look at the complete depth and absolutely have to dump all shares in one trade then you would basically have no slippage because you would know the entire cost up front. Why do so many people still experience slippage with the availability of full depth? I've read some literature about optimal order splitting, order book resiliency, etc. I'm not so sure chopping orders would incur that much cost because most places charge per share right now, or is that still relatively uncommon?
Well, one possibility would be if you submit a market order close to the same time as someone else (with yours submitted shortly after theirs), you would likely get filled a fair bit higher than you were expecting (given the numbers that you provided). Another issue with slippage is more about liquidity and order sizing - if you decide that you need to get out now, the liquidity of the stock is really going to determine your slippage, particularly if another of other parties simulataneously reach the same conclusion.
Ah hah, slippage from order timing seems to be much more of a problem than that of estimate costs based on available data. In that case, having the fastest data and order entry systems possible is what really helps in that case. As far as slippage from the marker order going deeper than expected, thus could be overcome by placing a limit order at worst price you expect to have filled at, if someone beats you to it you at least won't pay any more than than your worst case estimate.
When they changed to decimals I wondered how market makers were going to make money. I think today they figured out how: hide the true spread. You will never know who is next in line as a retail trader and you don't know the top tier liquidity until you trade. The Level II screen is hilarious today. Computers place and lift all these "orders" in millisecond intervals as if it's a super fast market, but the T&S produces only a few trades. As a result, you can't rely on the disseminated data to determine the true liquidity of the best bid and ask. Futures products carry big spreads, but at least you know where the best bid and ask liquidity is
Today I had to dump 5k shares of a very illiquid stock. There were about 30 transactions total on the stock for the entire day. Nine of them were my own sell executions placed just to make sure that I didn't blowout the bid side. The specialist (at least the one I know) from time to time creates the image of active interest by buying and selling to himself in order to adjust the spread width on Level II. I agree that often info is hidden from your level II. This happened once today to me when I had put a limit order on the ask side which on one occasion did not show up on the Level II. I was trying to drive a tight spread so that I could more easily bleed out. At any point in the day there were never more than 400 shares at a given time on the bbid side and I was intent on unloading 5k. It took 5.5 hours to keep the spread at my exit level (being on both ask with a limit order and hitting the bid when it was within a tight $.01 spread). Net net, I saved on average at least $.09 per share by not slipping past the inside bid. All of the 9 exits were at or above my desired exit. There was never a point where, I could have gotten all out at once without slipping. So to not slip, I combined limit and market appropriately and with patience. Unfortunately, the stocks high never got past my executed transaction price, so the stock didn't go up... I agree, with liquid names, this activity is at worst too rapid. But today, I used the bid/ask activity to get out with no slippage on markets, and gains in several cases via limit orders. My spin is that when possible, slippage can be altogther avoided as long as your willing to exercise some patience and you the bid/ask data to your advantage. I agree that this is most certainly easier done with illiquid stocks as opposed to liquid stocks which can move against you alot quicker should surges of volume pour in.
Good post makosgu. What you said is a non-geek-speek version of this paper. http://arxiv.org/abs/cond-mat/0206280 Abstract: In this paper we demonstrate a striking regularity in the way people place limit orders in financial markets, using a data set consisting of roughly seven million orders from the London Stock Exchange. We define the relative limit price as the difference between the limit price and the best price available. Merging the data from 50 stocks, we demonstrate that for both buy and sell orders, the unconditional cumulative distribution of relative limit prices decays roughly as a power law with exponent approximately 1.5. This behavior spans more than two decades, ranging from a few ticks to about 2000 ticks. Time series of relative limit prices show interesting temporal structure, characterized by an autocorrelation function that asymptotically decays as tau^(-0.4). Furthermore, relative limit price levels are positively correlated with and are led by price volatility. This feedback may potentially contribute to clustered volatility.
Great link... Thanks for the post. This definitely looks like a great read. I try not to talk geeky although I do alot of geeky work. It's easy to read alot of geeky stuff but definitely alot harder to apply since applying often requires intuition, the real hard part since it requires in depth comprehension. Will definitely give it a thorough read thru. Much Appreciated Stephen and Kindest Regards..