Algo Execution Issues on SPY

T

trade4love

Hello Everyone :)

Hopefully someone here can provide a little insight/advice/help.

My friend and I built an automated system in TradeStation, then sent the trade signals out of TradeStation (via DLLs) to our Sterling Platform (for better comm pricing).

We have run into major slippage issues, and the system is running on SPY. All orders are placed via ARCA and BATS, 500-1000 share lots. We want to receive the rebates for providing liquidity, or at the very least, not be charged for removing liquidity.

1. Should we be using darkpool routes?
2. What other measures can we implement to avoid/minimize slippage?

We will probably have some additional questions. Any and all help is greatly appreciated!

thx Paul
 
1. if you are using dark pools, you would never get any rebates, since most of them charge for add/remove. Therefore should only be used to remove liquidity.

2. Do you have a proximity hosted server for your strategy? If not, what latency do you have to CS (IP selection screen)?
 
LuckyTrade said:
1. if you are using dark pools, you would never get any rebates, since most of them charge for add/remove. Therefore should only be used to remove liquidity.

2. Do you have a proximity hosted server for your strategy? If not, what latency do you have to CS (IP selection screen)?
More...

1. Rebates aside, if we used darkpools, would the orders still slip? Or, would the orders be filled immediately (within reason)?

2. No proximity hosted server. Server/strategy is simply located here, in Toronto. Would that make such a difference on 5-10 trades per day (to have colocation in NYC or NJ)?

***btw, ty for the response :)
 
trade4love said:
1. Rebates aside, if we used darkpools, would the orders still slip? Or, would the orders be filled immediately (within reason)?

2. No proximity hosted server. Server/strategy is simply located here, in Toronto. Would that make such a difference on 5-10 trades per day (to have colocation in NYC or NJ)?

***btw, ty for the response :)
More...

To answer both of your questions, depending on which server you are on with Sterling, you would have a lot of delay. Imagine, your quotes have to travel all the way from where your trade station quote server is to Toronto, then you make your calculations and ship it all the way to either Chicago or New York servers for Sterling to execute. This is very likely to delay the whole process perhaps by even 1 second. When you are talking about trading SPY, its an eternity. You could be thinking that you are sending a marketable order or whatever, but in reality by the time it gets there it no longer is. So here is your slippage.
 
LuckyTrade said:
To answer both of your questions, depending on which server you are on with Sterling, you would have a lot of delay. Imagine, your quotes have to travel all the way from where your trade station quote server is to Toronto, then you make your calculations and ship it all the way to either Chicago or New York servers for Sterling to execute. This is very likely to delay the whole process perhaps by even 1 second. When you are talking about trading SPY, its an eternity. You could be thinking that you are sending a marketable order or whatever, but in reality by the time it gets there it no longer is. So here is your slippage.
More...


Awesome answer!!! Okay, so time to look for colocation options :eek:

And, just to be clear, would the darkpool routes be of any help? Or, waste of time/effort to bother with?
 
Dark pool routes are there to minimize the market impact, however lots of them scrape lit market, you never really know what their algos do. If you want rebates, then you need to post and then ARCA/BATS/NSDQ would be your best bet.

Before you start looking at where you need to place your strategy, do a bit of research, contact your quote provider to see where their servers are. Then ask your risk guys if your sterling account is on chicago or new york server and go from there.
 
If you're trading over the public internet such as you've describe above, I doubt order routing is going to make a significant difference in something like SPY. SPY, as you know, is a favorite for HFT folks, so you will be competing with people who know how to route their orders AND can do so in a matter of microseconds. If slippage is a big problem, cutting the hundred milliseconds of 'there and back' may matter a little, but you still will be probably 10s or 100s of milliseconds behind HFT players, so I'm not sure how much of an improvement will be made by using a hosting solution. That's something you'll have to assess on your own, obviously. I'm not really familar with sterling, but the other issue is what latency your broker adds going through risk checks, possibly internalizing, etc etc. Whatever the number is, it will of course be higher than the latency of DMA HFT folks with leased access/membership/whatever business solution they have that's invetably times better than the retail equivalent. I'm not saying that hosting will not improve you to a level that makes it cost effective, nor am I saying that your strategy is inherently doomed because of the HFT participants, I'm just saying you'll need to understand the cost/benefit of hosting and that there will always be some degree of uncertainty with fills.
 
cjbuckley4 said:
If you're trading over the public internet such as you've describe above, I doubt order routing is going to make a significant difference in something like SPY. SPY, as you know, is a favorite for HFT folks, so you will be competing with people who know how to route their orders AND can do so in a matter of microseconds. If slippage is a big problem, cutting the hundred milliseconds of 'there and back' may matter a little, but you still will be probably 10s or 100s of milliseconds behind HFT players, so I'm not sure how much of an improvement will be made by using a hosting solution. That's something you'll have to assess on your own, obviously. I'm not really familar with sterling, but the other issue is what latency your broker adds going through risk checks, possibly internalizing, etc etc. Whatever the number is, it will of course be higher than the latency of DMA HFT folks with leased access/membership/whatever business solution they have that's invetably times better than the retail equivalent. I'm not saying that hosting will not improve you to a level that makes it cost effective, nor am I saying that your strategy is inherently doomed because of the HFT participants, I'm just saying you'll need to understand the cost/benefit of hosting and that there will always be some degree of uncertainty with fills.
More...

Thank you so much for the insight!
 
Best of luck. Slippage is a complicated concept that I'm far from an expert on. There are several statistical techniques designed to estimate it, but I've never implimented any of them. Your best bet might be just to collect slippage data yourself so you can approximate the pdfs of slippage in the amount of x pips in a,b,c... different market conditions.
 
But if he is proximity hosted vs Toronto, he would be a much happier trader. He is not saying he is attempting to compete with HFT. So if he can get down to 20-30ms to market, it would help out his strategy a great deal.

my 2 cents.
 
You must log in or register to reply here.
Back
Top