Market Makers and Adverse Selection

Discussion in 'Strategy Development' started by clearinghouse, Sep 27, 2011.

  1. I was reading a paper written by one of Don Bright's traders about sub-pennying, where the trader talks about how when the market is hit, the direction of the move is likely to be against the trader who is hit on the whole penny.

    If this is the case (which I believe it to be), is it a sane and reasonable assumption to say that market makers in stocks on places like ARCA or NASDAQ generally almost always (well, let's say the probability p > .50, at the very least) absorb adverse-flow in the short run? Most passive entries will be against them, except in the occasional case where they get to the front and get some small random 100 lot or so?

    As an example, say BAC is (50,000 7.00 x 7.01 50,000), then a market maker at 7.01 who is first in queue gets hit at 7.01. If the trade is small, say 100 shares, maybe the odds are in his favor that the market won't go 7.01 x 7.02; however, if the trade is sizable against the book, it's going to go against him.

    Does this essentially mean that market makers are constantly in the business of taking short-term heat against them and managing that heat based on the assumption that the true price is in their favor in the longer term? Is it adequate to boil down the problem into something like:

    "You might be right in the next 5 milliseconds, but I will generally be right in the next 5 seconds?"

    Thoughts from the crowd?
  2. not rocket science.

    by definition, a passive order is getting hit when flow is against it.

    these days , flow changes about 200 times an hour, so I wouldnt sweat it.
  3. Yeah, so, theoretically though, if you were that guy who got the rebate, got nibbled at, and the flow didn't go against you, you just made bank then, right? Is that possible to do with any sort of reliability on something like nasdaq or arca or one of the primary exchanges where all the volume is?
  4. Because of the terminology used in your post, I am guessing that the paper you read might have been based materially on chapter 14 (“Bid/Ask Spreads”) of Larry Harris’s “Trading & Exchanges – Market Microstructure for Practitioners”. Just a guess …

    This is a detailed theoretical book, well worth a read (IMHO). However, when referred to here on ET in the past, other posters have claimed (and they are probably right) it dates from a pre-HFT world… so may have little relevance to what actually happens today.

    In Larry Harris’s (theoretical) world, prices move against the market maker (i.e. the “pure” market maker, who is not also speculating on direction) mainly when the MM trades with “informed traders”…. “Since the dealer incurs the loss only when he trades with a well-informed trader, the average loss per trade to informed traders is the loss from trading with an informed trader times the probability of trading with an informed trader. This average loss is the adverse selection component”.

    So an alternative (although more long winded) “boiling down” might be “… as most of you are uninformed traders (who are for the most part usually wrong), I will make a market for you and rely on the law of probabilities to keep me ahead when averaged out over time… I have added the adverse selection component to the spread to keep the expectancies in my favor even though I may occasionally trade with a trader who actually knows what they are doing.”

    However, much/most of today’s market making is done by algos/HFTs, which also act as speculators (rather than just theoretical MMs managing their inventory). And their economics consists not just of being paid the bid/offer spread, but being paid rebates, and also trading momentum and mean reversion strategies, etc. In these cases, the entity providing liquidity for the trade (the “MM”?) may have a variety of reasons for believing it should be a good trade without too much/any “heat”.
  5. So these algos/HFTs with passive orders have to assume that price-movement generally goes in excess of where they've decided to provide liquidity, right?

    So for case 1:

    Say I have 500 shares in my inventory and I think a forecast will go to Y. For me to realize my trade with some reasonable probability, I have to offer my 500 shares at some tick increment beneath Y, say, Y-Offset, with the full expectation that the market will take it in excess of Y-Offset.

    So for case 2:

    If I want to enter a given stock at price X, I have to assume that the price will go through or touch X (where I am in the queue) to enter, but that the long term forecast must be in excess of X.

    And that in case 1 and 2, this unavoidable. For a pure rebate trader who only adds liquidity, he would always need to have a long-term forecast and absorb short-term inefficient fills against informed traders, right? If he minimizes heat, all he is really doing is putting some sort of statistical bounds on how far-through the price will really go, right?
  6. you guys are complicating things beyond your understanding.

    hft are 99% arb and frontrunning.


    figure it out yet?

    u will not be able to do what they do.
  7. rosy2


    heres a paper
    Dealing with the Inventory Risk
    Olivier Gu´eant∗, Charles-Albert Lehalle,∗∗Joaquin Fernandez Tapia∗∗∗
  8. your head will explode reading that paper.

    they talk some serious trash, then show a stock that moves 4 cents in 20 minutes as an example.
  9. Thank you for this. This is very useful.
  10. I read it. Sometimes when you read these papers you don't walk away using everything they say, but you lift a few ideas here and there and then convert it into something else. I don't think their methods are compatible with what I am doing, but they did expose me to some new ideas.
    #10     Sep 28, 2011