Statistical arbitrage on Russell1000

M

Monstrum

I thought it was impossible to find a statistical arbitrage strategy for us non-professionals, but I managed to do it :) . There are certain times of the day when there are misalignments on some stocks due to strong execution algorithm pressures. I wondered why this wasn't already being exploited and the answer lies in exposure limits. I am testing the strategy that generates about 40 trades per day with a winning rate of 56%. The trades are relatively short-lived with executions spread out over about 5 minutes. During highly volatile periods, it has opened up to 300 trades per day. So it is possible for non-professionals to find these anomalies as well. This week, I will provide more precise statistics with various graphs.
 
strange ...
Your first post is about boasting HFT on Russell1000 (I presume futures).

For high-frequency trading, you have to talk about speed, delay, latency ....
For swing trading, speed, delay, latency .... are not critical.


Also if you attempt to HFT on Russell1000 CFD using statistical arbitrage thing,
you can never earn $$$$ from it.


Definitely, for me, I can't do HFT because I am trading from home.
And my trading platform delay could be a few seconds to a few minutes.
And my paid charting software delay could be a few seconds to a few minutes.
Of course, I could have used a paid trading platform like TT but there will be other horrific speed bottlenecks.

If you do demo trading, there are zero delays.
If you do live trading, the delay could be horrific.
 
Last edited:
earth_imperator said:
But are you aware of the fact that if you tell more about it then you just will kill the edge you had found...
More...
upload_2023-3-27_15-42-16.jpeg
 
Monstrum said:
I thought it was impossible to find a statistical arbitrage strategy for us non-professionals, but I managed to do it :) . There are certain times of the day when there are misalignments on some stocks due to strong execution algorithm pressures. I wondered why this wasn't already being exploited and the answer lies in exposure limits. I am testing the strategy that generates about 40 trades per day with a winning rate of 56%. The trades are relatively short-lived with executions spread out over about 5 minutes. During highly volatile periods, it has opened up to 300 trades per day. So it is possible for non-professionals to find these anomalies as well. This week, I will provide more precise statistics with various graphs.
More...

What is exposure limits?
 
Monstrum said:
I thought it was impossible to find a statistical arbitrage strategy for us non-professionals, but I managed to do it :) . There are certain times of the day when there are misalignments on some stocks due to strong execution algorithm pressures. I wondered why this wasn't already being exploited and the answer lies in exposure limits. I am testing the strategy that generates about 40 trades per day with a winning rate of 56%. The trades are relatively short-lived with executions spread out over about 5 minutes. During highly volatile periods, it has opened up to 300 trades per day. So it is possible for non-professionals to find these anomalies as well. This week, I will provide more precise statistics with various graphs.
More...
Could you say what is your lookback period (or formation period in other words) and what time frame/interval you use for these ?
 
TrAndy2022 said:
How many months out-of-sample or realtime tracking you were following the results ? What sharpe ratio you got out of it ?
More...

These are 1-minute data taken from Databento spanning the last 4 years, up until September 29th, 2022. At the end of this month, they should add data up to March 2023. I have run some quick tests and am now writing a high-level backtesting script that will provide me with more accurate statistics. On average, the gains will be 0.27% for S&P500 stocks and 0.20% for other Russell1000 stocks. Some stocks perform better than others.

When I finish the backtesting and have more accurate data, I will create a portfolio simulator to make the right allocations and coverings if there are too many sell or buy orders. From the simulator, I will obtain the Sharpe ratio.

https://im.ge/i/IF5WuW
https://im.ge/i/IF58kr
https://im.ge/i/IF5gPm
https://im.ge/i/IF5Yn1
https://im.ge/i/IF5cdf
 
Monstrum said:
These are 1-minute data taken from Databento spanning the last 4 years, up until September 29th, 2022. At the end of this month, they should add data up to March 2023. I have run some quick tests and am now writing a high-level backtesting script that will provide me with more accurate statistics. On average, the gains will be 0.27% for S&P500 stocks and 0.20% for other Russell1000 stocks. Some stocks perform better than others.

When I finish the backtesting and have more accurate data, I will create a portfolio simulator to make the right allocations and coverings if there are too many sell or buy orders. From the simulator, I will obtain the Sharpe ratio.

https://im.ge/i/IF5WuW
https://im.ge/i/IF58kr
https://im.ge/i/IF5gPm
https://im.ge/i/IF5Yn1
https://im.ge/i/IF5cdf
More...


As mentioned earlier (which you obviously didn't bother to read),
after you have done your demo trading,
you have to do live trading.
Anyway, who cares. Do whatever you like.

And when you do live trading,
that's where you will face hell.
If you survive the hellfire, then good for you.

All the best/worst to you Mister.

PS: We had done what you are doing now.
So LEARN, don't PREACH
 
Last edited:
You must log in or register to reply here.
Back
Top