if your strategy needs to trade in and out a portfolio of assets in high speed,(essentially it's a generalized pairs trading...)how do you make sure all those legs have good and simultaneous execution?anybody cares to shed some lights?thanks!

a multi-asset FOK-type execution rule?

FIX protocol allows for contingent legs, so if one leg doesn't transact where it needs to be, you can unroll and only lose transaction costs on one leg of your transaction. I personally use multiple feeds and internal logic to insure I have both legs. If I end up missing on one leg of a transaction (ie. arbitrage where one fill is correct, but the price slips on the second or third leg, hence erasing your profits) you can then fall back onto secondary logic that tries to unwind your position in as favorable way as possible. One example it to move that bad transaction to a separate watch, then monitor for spreads to align in such a way as to let you unwind with as minimal a transaction cost as possible.But I haven't found a way to insure perfect execution for multiple leg transactions, and just consider those losers. Then try to minimize the damage. (This from a retail perspective with the best CPU on servers possible, and the lowest latency datacenters as possible.) So an institution might have a different perspective and a whole lot more resources to throw at the problem.But I always strive to improve my "hit" rate, so any ideas are always welcome.-cjforex-

it is quite trivial to say that you can't.most of the time, on the exchange side assets are managed by several engines/servers/locations. so even if you're the first to send orders on a specific set of assets, fills will be totally asynchronous.

Try looking up the autospreader from TT.http://www.tradingtechnologies.com/

mizhael, the only thing you can do is provide on the primary entry and exit and take on the hedge entry and exit. this will generally tend to offset your execution costs but also increase your slippage on the hedge - which quite often the slippage on the hedge (on entry and exit) will be equal to your edge or spread.

The way this is generally done is to use 1) limit orders with TIFs that cancel replace or send markets (depending on whether it's the first leg or the second)2) some sort of maximum expected size per position based on either risk levels and/or the number of pair legs currently firing3) then optimize 1 and 2 by latency and expected/actual fill levels (aka slippage).In tradelink there are components to do #1 :http://code.google.com/p/tradelink/wiki/TifTrackeras well as #3 out of the box :http://code.google.com/p/tradelink/wiki/LatencyTracker2 is pretty easy to do, if you join the user list somebody would help you with it.http://groups.google.com/group/tradelink-usersTradeLink is 100% free and open source. Works with 15+ brokers and data feeds. Super-fast tick-based backtesting. http://tradelink.googlecode.com