well true stat arb isint always the 'zero risk' i was talking about i guess. it depends how you do it. hedging etc. remember I am new to finance so I have very limited exposer to stat arb.
Arbitrageurs
- Thread starter paulbechard
- Start date
segv, recently I have been focusing on trying to solve this slippage issue on IB. I think they are a market maker for forex but I am unsure for any other equities. I think the solution to this problem will either come in the form of some tricky combination / limit order magic or less latency.
Paul,
In regards to these two issues, your strategy is extremely susceptible to many perils associated with high-frequency trading. Some possibilities:
1. It is possible that you have a sampling error. When you detect a signal at time t+0 it has ceased to exist by the time your order reaches the market at time t+1.
2. It is possible that your strategy creates market impact. The act of attempting to take advantage of small inefficiencies could make the market more efficient.
3. It is possible that your strategy is detectable by other market participants. You are not the first person to develop an automated trading agent of this kind. Some of the adaptive agents can be very clever, and can detect anomalies in their environment very quickly.
4. It is possible that your strategy is trading random noise. It is quite easy to find a fit for some condition in a large amount of random data. Or, your system might be initiating trades based on some condition that is the result of some stochastic process. The condition may never have existed, but seemed to because of the stochastic arrival of information.
5. It is possible that your data is totally invalid. Quotes, trades, and other high-frequency market data are notoriously unreliable. Institutions that profit from this form of trading go to great expense in an attempt to sanitize their data in real-time.
I think that the field of high-frequency microstructure trading, in general, has reached the stage of diminishing returns for all but the most competitive institutions. My recommendation to you would be to explore other quantitative methods that are not dependent on such high-frequency data.
-segv
If you believe slippage or latency is the source of your problems, and your system have yet shown consistent period of profitability, then you are looking at the wrong place. Slippage can make a big difference, but not enough that would make your system unprofitable in the long run. Latency is something that can be improved, it is a CS problem, not a strategy problem, a few ms can not make an unprofitable strategy profitable.
You know a lot more than most new grads, which is great. But I believe that you owe it to yourself to see what a real-world automated trading environment looks like, all the data analysis goes on, the model maintenance, etc, etc. I did pure AI (knowledge representation and default logic) research as an academic, and I use some algorithms from those days, but to do high frequency trading requires some deep knowledge about the exchange market data / order routing systems, things that it is difficult to gleam from books.
As for high frequency direct API brokers, two good ones that have mentioned often here are Lime Brokerage (contact Kevin Snow over there), and RedSkye securities. I have not had direct experience with them (I am a member of a few exchanges), but I know the founders of both firms, and they are all very well respected industry people.
You know a lot more than most new grads, which is great. But I believe that you owe it to yourself to see what a real-world automated trading environment looks like, all the data analysis goes on, the model maintenance, etc, etc. I did pure AI (knowledge representation and default logic) research as an academic, and I use some algorithms from those days, but to do high frequency trading requires some deep knowledge about the exchange market data / order routing systems, things that it is difficult to gleam from books.
As for high frequency direct API brokers, two good ones that have mentioned often here are Lime Brokerage (contact Kevin Snow over there), and RedSkye securities. I have not had direct experience with them (I am a member of a few exchanges), but I know the founders of both firms, and they are all very well respected industry people.
Not intended to discourage you but going in and out of a trade withing only one second means you are catching one or two cents if you are trading stocks.
If this is the case then all the odds are stacked against you and doing so using market orders is a real killer, this is trading frame for market makers and very advanced institutions.
Automation is great but you need to find other edge.
If this is the case then all the odds are stacked against you and doing so using market orders is a real killer, this is trading frame for market makers and very advanced institutions.
Automation is great but you need to find other edge.
yeah, those are the most lucrative...
Yeah you're right. Iv started only buying lotto tickets. 1$ in, 20 mil payout!! now thats lucrative!
But you know what they say. A bird in the hand is worth 20 million in the bush. Now I just need to figure out how to arb the natural bird/bush ratio.But seriously. I know that small low risk trades will definately not maximize the profit on my capital. I'm just hoping to get something consistant to work, and I'll research more risky trading strategies later.
I have a new question I am hoping people with some intimate exchange knowledge would know. Would there be any difference between putting two limit orders and dealing with what happens if they dont both get filled compared to placing two market orders? Do limit orders get priority or are there micro fluxuations on the exchanges that could hit reasonable limit orders that I do not see? I notice that sometimes my market orders get filled at prices which I do not see coming through the broker.
Thanks
Of course there is a difference. The difference is the guarantee of having the order execute, and the cost of the guarantee is the bid-ask spread. As to "what happens if they don't both get filled", you tell me. What happens when one order gets filled but not the other in your strategy? You now have risk, thats for sure. I thought you were trying to decrease slippage.
-segv
Sorry, I did'nt make myself clear. I was'nt asking about how to do a duel limit only order, and I understand the fundemental difference between limits and market orders.
What I was really curious about was if anyone uses limit orders in place of market orders because they find they fill 95% of the time as fast as market orders or whatever. And if the exchanges treats limit orders different as in the priority they get filled, and wether or not on average if I were to send the same order as a limit and a market during a point of high volitility, would the limit tend to fill at the price and the market slip greatly due to fast fluxuations at the exchange?
My main goal right now is to reduce slippage, if i can place two limit orders and have them filled as quickly as market orders enough of the time, having guarenteed prices and not being a victim of large swings might be worth dealing with the issue of being stuck with only one or possibly both orders not filled.
What I was really curious about was if anyone uses limit orders in place of market orders because they find they fill 95% of the time as fast as market orders or whatever. And if the exchanges treats limit orders different as in the priority they get filled, and wether or not on average if I were to send the same order as a limit and a market during a point of high volitility, would the limit tend to fill at the price and the market slip greatly due to fast fluxuations at the exchange?
My main goal right now is to reduce slippage, if i can place two limit orders and have them filled as quickly as market orders enough of the time, having guarenteed prices and not being a victim of large swings might be worth dealing with the issue of being stuck with only one or possibly both orders not filled.