WINNER... Really all they do is match a buyer and seller. They buy at the bid and sell at the ask and make the penny on millions of shares per day. They can do this b/c they have the necessary nasdaq capital requirements and have paid substantial sums of money to make markets in the stocks they are interested in. If anyone wants more info or has $10 million to committ to such a strategy pm me.
Why? Faster transaction times without being knocked out of the queue when providing liquidity? I'm running a high-freq sub-second arb strat at the moment, but I'm removing liquidity. Are you suggesting other players are sitting there offering liquidity in various locations in anticipation of the arbitrage -- and using cancel/modify on their orders? I suppose they have less technical risk than I do (I still go through a broker), since ARCA and BATS do things like flush and cancel orders on disconnect. Question is -- what happens if ARCA and BATS don't detect your disconnect when you have serious size on the books?
There's no strategy to give away. It's based on: Say Bid was $10 and the ask was $10.01. Assume they knew the exact moment buys and sells would come, which L2 mm's have. They will simultaneously buy from the seller at $10 and sell to the buyer at $10.01 making the spread for which they are entitled to for crossing the transaction. No magic, but requires exorbitant sums to carry out, and a lot of infrastructure to implement.
His explanation assumes they are running index arbitrage strategies. In today's market, it depends on what index/etf you are running those arbitrage games on. If you think you can play arbitrage games using the S&P500 indexes, you are going to need transaction times of under 1ms, colocation, and direct feeds from both NYSE, NASDAQ, ARCA, BATS, etc. It's a tough game to play without some serious startup capital, on account of the technical cost alone.
Are most high-frequency strategies either stat-arb or market making? My general idea about arb is that you enter when a price discrepancy is detected, and exit when the discrepancy disappears. So in this kind of stat-arb strategies, the price discrepancy can really disappear in sub-seconds? That would be quite impressive if that's really how the strategies work. Market making strategies are simpler to understand conceptually, but I guess they are just very infrastructure and technology-intensive, only playable by the big IBs/hedge funds.
You don't need to do all trades in one symbol, your system might be doing 1 trade ever 2 seconds on all of Xetra and you got yourself 9000 trades per second... Any strategy that's profitable in low frequencies can be made into a high frequency system... you just need to run it on thousands of symbols simultaneously...