What are algos actually doing?
- Thread starter booked
- Start date
It does get killed in a trend. MM HFTs prefer slow moving instruments with large ticks.
Well it doesn't get killed completely. They still get a bid/ask range to stick around for a few seconds, but how long they stay at it depends on their algos and how much market orders they are willing to sustain before bailing out due to loss of symmetry. Those are HFT strategies for market making, they aren't all exactly the same. Many involve arbitrage in the same go, allowing them to keep a bid/ask position for longer even if the symmetry is lost because they can assume it will restore itself due to the arb info they posses. There are many variants but the logic is about the same.
Even with short trends, the spread stays steady long enough for them to make fractions of a cent at a time, and this adds up over time.
For instruments like AAPL where the tick is rather small (the spread can be a dozen ticks) it gets harder for them also. This is because non-hft directional traders can afford to tighten the spread by posting a buy limit above the bid. If the spread is too tight, the HFT won't make money because they still have to pay fractional commissions. That's why they like stocks with higher volume, less trending, and lower absolute price (ie in the 20s).
+1
Someone "gets it"
Rebate trading and front running. They use predatory orders to scan liquidity in dark pools and the market and then trade accordingly. They mostly use EDGX and BATS since both these offer a wide selection of order types designed for algo trading.
I think something like 51% of all BATS rebates go to one algo trading firm.
Other than that I think like 8% of volume is straight arbitrage. A few news HFT programs.
Also latency trading is a big HFT share of the pie. This goes with predatory trading strategies. Since the US uses a system wide NBBO that creates latency between direct market feeds from all the exchanges. An algo can trade ahead (microseconds) when it sees an order from a direct feed but the NBBO hasnt yet updated system wide. Co-location required here....
I think something like 51% of all BATS rebates go to one algo trading firm.
Other than that I think like 8% of volume is straight arbitrage. A few news HFT programs.
Also latency trading is a big HFT share of the pie. This goes with predatory trading strategies. Since the US uses a system wide NBBO that creates latency between direct market feeds from all the exchanges. An algo can trade ahead (microseconds) when it sees an order from a direct feed but the NBBO hasnt yet updated system wide. Co-location required here....
I never had the cost structure to facilitate this sort of trading, but I wanted to investigate this style of trading so I designed some models around speed to understand my counterparties and spent a lot of time obtaining that speed to prove my ideas.
What I found was that I was slowly losing money against the market, but slowly gaining money on just pure rebates and the occasional spread-making. On some days, I'd get completely beat up in the trend. I needed a sort of portfolio of stocks to keep the rebate profits to keep the overall pnl positive (without my fees that is; ultimately my broker made the most profit on my efforts. I would need a radically lower cost structure to make money doing this.)
It's a non-trivial effort to enter this space. I'm not sure how many firms can enter this space without eating short-term losses on the way in. The barrier to entry is fairly steep, and connectivity is critical because what you don't know is what ends up hurting you the most. What you don't know can be expensive to obtain also. There are a lot of fast firms, but I think there are maybe only a few firms total that have any real market-making abilities.
Some may disagree with me, but I think anyone entering this space with visions of grandeur is probably better off looking for subtle quantitative edges over speed -- at least until one has a solid revenue base to finance expansion into true market making.
Can you elaborate on this point further and provide an example? The reason I'm asking is because, let's say I see someone trade out EDGA. It doesn't help me to go hit NASDAQ just because I saw someone trade out EDGA. That alone is not reason to go running off to an exchange where the remove fees are an arm and a leg.
Most people use smart routing, so seeing NASDAQ clear before EDGA or NasdaqBX almost never happens.