If I recall correctly, the CME didn't set the 5:1 ratio, nor the per cancel fee - IB did. If I remember correctly, GLOBEX was to charge something like $2K/product/day if a broker exceeded the threshhold. It would be useful if, when/if these fees are decided to be charged, IB could reveal their statistics to show people how they were calculated.
I think the modification / cancellation fee is a good thing for the reasons mentioned above. I did not, however, like IB's original proposal for two reasons: 1. Credit ratio of 1:5 is too low, triple it and I am happy 2. Credits do not carry to following day, make them good for the next 2 trading days and I am happy. Trent
If you look at what floor traders do, they continually are bidding and often modify their orders. They dont get a fill most of the time. Exchanges and brokers would like their clients to be 'good' customers- to send in 10 market orders a day with no cancellations at then go home. Well if electronic exchanges really want professional 'at home' remote floor traders, then they have to allow floor trading behavior which means frequent modifications. If you think an automated trader is using a method which is wasteful- such as continually entering and modifying orders that arent marketable, then you can use some metric to detect this rather than some expensive $1 fee that hits everyone. Many of these wasteful automated systems are trying to game the first in first out order queue. Maybe a good alternative would be for the exchange to mix in random order matching along with FIFO. This would allow some form of front running and reduce the benefit of being early in the queue at a given price level.
We are on the same page. We will continue to try and influence the CME fee policy. Our implementation will reflect the exchanges final decision, and will be skewed towards charging users who over-use the routing capacity. For now I am off this subject.
Can any of the more solid programmer types quantify how much more bandwidth we are really talking here and what kind of additional processing power would be necessary to manage the increased order flow? It would be interesting to see what the cost in time and bandwidth of processing one additional instuction/second would be and scale from there.
LMAO. I agree completely. Its a shallow attempt to create an edge for seatholders and members. The problem of overutilization of bandwidth will otherwise be self correcting. Latency caused by bandwidth restriction would probably cause executions to occur on orders that are trying to cancel. That will likely cause some problems for rapid fire order canceling bots. The FIFO creates a free option for those early in the que on the "correct side" of the B/A in a market which is very deep and thick. Free options are ultimately the problem IMHO. One solution not discussed by CME is to thin the que by reducing the tick size on the ES to .10 or better yet .05 and enable "penny upping". Of course market makers and those relying on structural edge such as an arbitrary .25 tick will never go for that. And they rule the CME.
You can guess exactly what people are doing. Do assure your place in line, they'll scatter bids across 6 or 10 price levels. Then they'll pull the bids at the initial price levels until they think the market is turning, then they'll leave that bid in place and cancel the rest once they are filled. In one aspect they are like floor traders which is that they bid the bid. But in another aspect, they have an assured place in line and a guaranteed fill at the bid with the option to cancel at no cost. This is an advantage that no floor trader has. The floor has a tradition by which those who make the market get the first shot at fills. However among those those have been making the market, there is no fixed place in line. And even within this floor tradition, front running is still possible. The early bidders have an advantage with respect to getting filled, but they do not have a guaranteed fill at the bid nor a fixed place in line. To replicate this in electronic markets, exchanges would have to modify the strict FIFO queue and allow some random order matching regardless of the place in line. The would mean that the above strategy would not result in a guaranteed fill at the bid and the free option would suddenly have a cost to it. However there still would be a statistical advantage to the early bidders, giving a benefit to those who make the market. Partial order queue randomization has another advantage because it adds a cost to those who use high bandwidth strategies without imposing a $1 fee on users who normally modify and cancel orders. Also the electronic exchanges and the API brokers could develop a metric to catch those users whose bots are poorly designed and waste server resources. It could even run real time so that if someone's bot flipped out or someone attempted a DOS attack, the trading system as a whole would not be impactedl
If you also feel the "thick que problem" should be addressed by a lower MPV, vote with your feet and trade YMs instead ES