There is a huge difference between theoretically and reality. HFT can never put 3 points stops in the ES. So management cannot be the same, or HFT will be bankrupt within the first hour of trading.
Yes, but with much smaller profits. And much more damaging stops in relation to the average profit per trade. Especially in trending markets. I tested all these things for years. How good is your HFT and how is your "longer timeframe trading", that's the question. or: Number of winning trades and average profit per winning trade. Number of losing trades and average loss per losing trade.
But I'd take a strategy that wins 51% of the time (or 1% of the time for that matter) with an expected return of 1 tick ($12.50) and trades 100 times a day, over your 3 point strategy that trades three times a day with a higher probability. This is the game Virtu, Jump, et. al are playing...not the game you're playing. 100*$12.50 = $1250; 3*$150 = $450; Each trade is worth less....but the cumulative returns are more. Obviously, commisions/fees will have an effect that i'm not accounting for here...
Have you compared simulated fills vs real fills? Why not run simulated data over the same period as the trading, do the results match? It's most likely an error in your code, peeking issue, it also would explain the extremely high Sharpe. ES has no liquidity issues to speak of.
That's what I recommend a few posts back. I'm shocked that people don't do this for every single trade, regardless of performance issues. You can learn so much from it. My system does it automatically at the end of every trading day and generates a report for me. I used to design guidance, navigation, and control algorithms for high performance missile systems. Obviously, when your dealing with high-dollar systems most of your design is going to take place in the simulation to mitigate risk..... thusly, a great deal of effort was expended after every flight test to update the simulation models to ensure they matched observed reality. Trading shouldn't be any different.
The opportunity cost is also not as high with a higher trade frequency. Having a faster feedback loop is an advantage to transform noise into signal. Consider the technological arc of developing radio to cellphones. On one level of awareness it’s all signal and it required a sequence of events to understand and harness - one analog cycle to digital step at a time. There is a Goldilocks zone between the two extremes of HFT and LFT. A holistic system will outperform an edge based one through time for it acknowledges the value of time and uses it as an advantage instead of discounting it. I’ve not developed automated systems so my view is biased, instead I’ve chosen to develop the software between the ears as the foundation and implementing through a combo of discretion and quasi-automation. Having a past-life as a application developer, it’s much easier to change mental models vs code based on those models. It’s pretty fascinating that in this profession, context changes the accuracy of any ‘truism.’ I supposed that’s the structural nature of trading with someone whom has an opposing interpretation of what they are perceiving from the same market data regardless if it’s a human clicking that mouse or a human/ml coded app.
Normally I would look at a 100 tick chart or less and scalp it for one or two ticks. Ideally I was trying to profit off of the spread, in simulated it works, but realistically there needs to a buyer for every seller and vice versa. I know the picture attached isn't considered scalping but its a system that has worked 80% of the time. So instead of capturing quick profits off of the spread, I am redesigning the algorithm to find volume surges from institutional algos to make quick profits. I primarily trade off of a 1 minute and 5 minute chart. L means long S means short E means exit my stops are 6-8 ticks back with 4-10 tick profit targets. Are my stops too tight? to ensure execution I take liquidity 60% of the time. Dependent upon market conditions...
For as far as I know HFT's don't trade the ES. Problem might be the $12.50 fix prices per tick. But I am not sure about that. And if they would, and I use your example, make $1,250 per contract per day, and trading 1,000 contracts (which is very small money for Virtu). They would make then 240 days*$1,250*1,000 contracts. Or in total 300 million$ a year. That's roughly twice the total yearly net profit of Virtu, trading 230 markets with 12,000 instruments by 148 people. Virtu should kick everybody out and just trade the ES then. Something is not adding up correctly in your example. I will come back to this later, no time now. PS: my postings were initially a reaction to OP. So not about Virtu, but about the 20 year old OP who wants to do HFT.