I'm gonna guess that it's your strat that needs help before order execution.
But, if you want simple, just pick a place between bid-ask to IOC. If you want more complex, then..
http://lmgtfy.com/?q=optimal+limit+order+execution
Sharpe getting killed by bid ask spread
- Thread starter traider
- Start date
I'm gonna guess that it's your strat that needs help before order execution.
But, if you want simple, just pick a place between bid-ask to IOC. If you want more complex, then..
http://lmgtfy.com/?q=optimal+limit+order+execution
L
lovethetrade
Must be awesome being able to speak for all developers and know how they think.
Having dealt with about 10 developers in the past for algorithm development I reckon that's makes me quite experienced with the "real world" and each developer showed a clear pattern of behavior that was consistent with stealing the IP and they all traded their own trading systems.
Unlike other forms of software development, algorithms are not exposed to the market and can't be patented so it's very difficult to prove whether someone is stealing the IP. I think you'll find most algorithm developers know this.
Last edited by a moderator:
Wow why so salty when nothing to contribute? I'm not the HFT that is arbing u
didn't mean it to come across as salty. my apologies if that's how it came across. But, I don't get how sharpe would be getting killed (as your title says) unless you were doing lots of frequency in short-term trades. Maybe you can explain? So.. it's back to the drawing board. Work on the strat..
as far as the algo, you probably don't realize how complex this can get. And you asked for simple. So.. just place something between the spread. When you get into the question of whether your order will be filled and which order type is optimal.. you are introducing a beast that is not at all simple. So, get to work! Do some research..
let me explain more indetail. The model trades a variety of instruments. For a few of them, the holding period is shorter and average trade expectancy is $6 even though I've tried to maximize holding period without affecting signal generation too much. So 100 shares with bid ask spread of $1 will be like 14.4% of my pnl. I'm trying to increase the margin of safety by improving the execution part too. I know this part is tough so I asked if anyone knows of order execution algos that I can study. It's easier to start with something basic first before progressing to the harder parts. Up till now I have more experience with the alpha model and not so much with order exe so I'm hoping that there are ppl in the community willing to share their exp
Why develop a platform that has order execution and the IP containing strategy exposed to each other? Structure it with a platform that handles all of the API stuff, including order submissions, executions, etc., the whole position maintaining ensemble. Call it a Position-Management platform? Put your IP algo in strategy code you write; have it pass a position size command (symbol, size, side) to your Position-Management platform. Additional commands define which actions (enter long, exit long, scale long, scale short, etc.) use what order types and with what default values; defaults by security and changeable on the fly as/if your algo decides a change is in order. Allows your IP algo code to work at a high level instead of being embedded with order submission, handling executions ,etc.. Need a new order type or order price improvement behaviour? Pay your guy to code it into the Position-Management platform. Enable, disable, set parameters, for it from your strategy code. He doesn't need to know what your IP algo is or the values of the parameters you'll use for which order types. And if you change brokers (or the API changes, hint hint), you only need to port portions of the Position-Management platform. You can have a developer build the framework for implementing your strategies too, including the usual indicators. Custom indicators you can code or extend from existing code for a similar indicator. Then you code the strategy with the IP algo. Software wise, these should all be separate functions to be robust and reliable.
Being a spread trader with some electronic execution experience, my hunch is that your strategy is hopelessly outgunned by faster bots with much deeper ECN pockets than mere mortals. It is just a reality in today's markets that having a trading strategy absolutely reliant upon certain electronic execution at exact prices needs to be overhauled. You asked - just my opinion. Good luck with things.
See if you can get a working link to the WSJ article published March 21 entitled:
"High-Frequency Traders Fall on Hard Times"
Not much that's original in that piece, though it's rare to see data costs featured so prominently in WSJ,
HFT firms also grumble about mounting costs for the market data they buy from operators like the New York Stock Exchange and Nasdaq Inc., as well as for co-location, the practice of putting a computer server directly in the exchange’s data center to cut down the time it takes to execute trades.
Chicago-based Wolverine Trading LLC told the Securities and Exchange Commission in December that its total costs related to NYSE equities market data had more than tripled from 2008 to 2016. “This is a monopoly,” Wolverine said in a letter to the SEC.
NYSE rejected the accusation, telling the SEC in January that firms like Wolverine had “chosen to build business models based on speed” and could choose not to buy NYSE’s highest-speed data services. Intercontinental Exchange Inc., the owner of NYSE, and Nasdaq say their fees are reasonable.