What techniques to market makers use to manage operational and 'massive' adverse selection risks?

Gambit · Dec 14, 2015

An integral part of market making is trading against uninformed flow. I know of one darkpool which allows this explicitly: pdq. Turqoise may allow this as well. Do any of you think these venues have value?

Gambit · Dec 14, 2015

cjbuckley4 said:
i belueve he's referring to a Calendar Spread Option.

Calendar spread option. Makes sense. Thanks.

cjbuckley4 · Dec 14, 2015

IAS_LLC said:
CJ,

For what its worth, the open order risk you're talking about has had me a little edgey too. My solution (if you can call it that) has been to be conservative, and only place orders to establish queue position at one book level (usually the top of the book, or the 2nd level... depends on market conditions). I basically run an adverse selection simulation for my entire portfolio before my risk management algos will approve an order to be released to the market. If the simulation results in my risk thresholds being violated, I will reject/cancel open orders and liquidate positions until everything is back in check. This probably results in fewer opportunities to profit, but should allow me to survive more bad days (I hope).

Thanks,
Brad

Hi Brad,

Thanks for jumping in. Can you comment on how you've done for Q position using that tactic? Joining the Q on the BBO is kinda a special case that is situational because incoming order flow is constantly altering the state of the first levels, but if you look at the shape of the book, you usually have your greatest amount of size on levels 2-3. Do you find there's alpha consistently adding size on the second level?

Thanks.

cjbuckley4 · Dec 14, 2015

Gambit said:
An integral part of market making is trading against uninformed flow. I know of one darkpool which allows this explicitly: pdq. Turqoise may allow this as well. Do any of you think these venues have value?

I gotta bow out on that topic. I'm not appropriately capitalized or technically equipped to deal with equities. I'm gonna have to restrict my efforts to CME.

IAS_LLC · Dec 14, 2015

cjbuckley4 said:
Hi Brad,

Thanks for jumping in. Can you comment on how you've done for Q position using that tactic? Joining the Q on the BBO is kinda a special case that is situational because incoming order flow is constantly altering the state of the first levels, but if you look at the shape of the book, you usually have your greatest amount of size on levels 2-3. Do you find there's alpha consistently adding size on the second level?

Thanks.

I guess I should clarify: I end up canceling orders before execution more often than not. I constanly (usually) have an order on both sides, but cancel before execution if the proper conditons dont exist once my queue position estimate nears the top. After canceling, i immediatly will go back to the end of the line (usually). I'm not trying to constantly capture the bid-ask spread, and am currently only doing 1 lots, so would not call what I'm doing market making, per se. When I target the 2nd level, its because I think the market is about to briefly move away from a mean and than revert for a quick pip or two on one or two contracts. I really couldnt say if there is profit to be had in adding real "size" to the levels off BBO. My gut says yes, but i think youd have to start modeling things like market response to your orders to really know the viability of a strategy where "size" is being put on....or just test it on live markets out of the gate...and be sure what you're doing can't be interpreted as spoofing..

Full disclosure : I'm still in simulation mode, haven't taken it live yet.. I like the way my simulations look so far, but the real market is the only judge that matters...so time will tell

brownegg · Dec 15, 2015

Gambit said:
Thank you for your comments brownegg. Would you mind elaborating on the arms race? Do you think 20 to 40 us is competitive enough to make any impact?

It's hard to define "impact". Anything sub-100us is usable in HFT. Maybe not straight-up vanilla arb, but you can make money. The problem is that about five years ago, too many people showed up and bought lots of shiny toys, and there are a LOT of players at that speed. The pickings are slim, and the costs are very high. If you fall behind in any aspect, including WAN / microwave / whatever, all of your costs get you nothing. So your "reward" for surviving is that you get to spend more.

Also, it's pretty unlikely that you have access to a system as fast as you're describing if you have to ask that question. If you had that, you'd know it's fast. It's also possible someone is telling you the system you can have access to is that fast, but I'm skeptical of that too.

Gambit · Dec 15, 2015

brownegg said:
It's hard to define "impact". Anything sub-100us is usable in HFT. Maybe not straight-up vanilla arb, but you can make money. The problem is that about five years ago, too many people showed up and bought lots of shiny toys, and there are a LOT of players at that speed. The pickings are slim, and the costs are very high. If you fall behind in any aspect, including WAN / microwave / whatever, all of your costs get you nothing. So your "reward" for surviving is that you get to spend more.

Also, it's pretty unlikely that you have access to a system as fast as you're describing if you have to ask that question. If you had that, you'd know it's fast. It's also possible someone is telling you the system you can have access to is that fast, but I'm skeptical of that too.

I admit that I am wholly out of my element here. The latency numbers I quoted are straight from vendors. Let's suspend disbelief for a moment. Could liquidity provision in a single market be viable? How about trading the soybean crush?

Gambit · Dec 15, 2015

2rosy said:
pick your market. I wouldn't make markets in front month crude oil but would in csos or any rfqs. there are still people in the options pits making markets: grains, meats. on the screen with average software you can't stream quotes on anything but otm without getting picked off

Thanks for the input.

cjbuckley4 · Dec 15, 2015

IAS_LLC said:
I guess I should clarify: I end up canceling orders before execution more often than not. I constanly (usually) have an order on both sides, but cancel before execution if the proper conditons dont exist once my queue position estimate nears the top. After canceling, i immediatly will go back to the end of the line (usually). I'm not trying to constantly capture the bid-ask spread, and am currently only doing 1 lots, so would not call what I'm doing market making, per se. When I target the 2nd level, its because I think the market is about to briefly move away from a mean and than revert for a quick pip or two on one or two contracts. I really couldnt say if there is profit to be had in adding real "size" to the levels off BBO. My gut says yes, but i think youd have to start modeling things like market response to your orders to really know the viability of a strategy where "size" is being put on....or just test it on live markets out of the gate...and be sure what you're doing can't be interpreted as spoofing..

Full disclosure : I'm still in simulation mode, haven't taken it live yet.. I like the way my simulations look so far, but the real market is the only judge that matters...so time will tell

Of course, a lot of cancellations are a foregone conclusion. No one wants to provide liquidity at a lost.

At risk of sounding like a jerk questioning a part of your approach on two straight posts, I gotta ask this both for my understanding and as a scientific method type of deal. You know, one guy proposes something, the others attempt to disprove it.

When you say "After canceling, i immediatly will go back to the end of the line (usually)", can you help me understand your logic behind this? It may be a good idea, but I can't really see why immediately looking at it. My thought is that if you're last in the Q, you have really high adverse selection risk, because the guy who took you out just got in on the new BBO without paying the spread. Furthermore, you've lowered your probability of capturing the spread (or whatever your strategy is, because I'm sure it's not to see prices move against you), because the probability of the Q behind you holding up against opposing order flow pressure has become lower than it was at your initial Q position.

2rosy · Dec 15, 2015

Gambit said:
I admit that I am wholly out of my element here. The latency numbers I quoted are straight from vendors. Let's suspend disbelief for a moment. Could liquidity provision in a single market be viable? How about trading the soybean crush?

yes or in similar markets and with rfqs