Liquidity and Fast Traders: Real World Questions

This post is more inference than proof. Take it as you will.

The following are share counts executed on a trade I tried to fill
today, with times. "limit" indicates that those shares were executed
at my limit price. "limit - .01" means that I bought at the limit
price less one cent. The trade was posted at about 1:30pm, adjusted at
2:01pm for a higher limit price, and did not fill in its entirety. My
limit was top of the book for most of the afternoon (meaning I was
offering the best price of anyone displaying the shares they were
willing to buy).

This is a stock with a very low share price, for those who note the
large number of shares.

09/13/2010 02:01:13 PM    300.000        limit - .02
09/13/2010 02:01:13 PM     345.000       limit - .01
09/13/2010 02:09:20 PM    200.000        limit

09/13/2010 03:46:28 PM    381.000        limit
09/13/2010 03:50:05 PM    17.000          limit
09/13/2010 03:50:28 PM    1,000.000     limit
09/13/2010 03:50:40 PM    13.000         limit
09/13/2010 03:51:10 PM    26.000        limit
09/13/2010 03:51:40 PM    23.000     limit
09/13/2010 03:52:15 PM    17.000     limit
09/13/2010 03:52:50 PM    11.000     limit
09/13/2010 03:52:52 PM    100.000   limit
09/13/2010 03:53:20 PM    6.000     limit
09/13/2010 03:53:55 PM    3.000     limit
09/13/2010 03:54:25 PM    2.000     limit
09/13/2010 03:55:00 PM    2.000     limit
09/13/2010 03:59:50 PM    98.000    limit

Note the large number of small share-count trades executed after 3:45pm.

The market closes at 4pm. The last 15 minutes of the trade day are
filled with traders trying to unload shares they've picked up over the
course of the day so they avoid overnight exposure. Notably, these
traders picked up the shares after the day started, and ditched it
before the end of the day.

Interestingly, the bid-ask spread was just a penny, but if I had
executed my entire (small) order of 3333 shares, I would have spiked
the price 52% without hidden liquidity (which I'm sure would have
stopped the ascent at under 10% - still a big spike, but far less).

What does this tell us? Three things:
1) bid-ask spread is a crappy measure of liquidity
2) fast traders are a HUGE percentage of the market. Look at what
percentage of my shares came from traders unloading!
3) there is a lot of liquidity hidden from the market (largely either
fast traders themselves, or hidden by institutions to prevent fast
traders jumping ahead of large orders).

It's also instructive to note that fast traders are largely very
profitable, and they're making their money off of longer-term
investors (like me). I did a back of the envelope based on the depth
and change-frequency of the various bids and asks in the order book,
and I'd be surprised if HFT cost me less than 1%, and possibly 2%, of
my trade value - this was a micro cap (well under $70mm mkt cap), so
this shouldn't be a shocking number.