One of the world's greatest traders said:
"A 1917 edge that still works is superior to a 2017 algo".
"A 1917 edge that still works is superior to a 2017 algo".
Last edited:
Is Quantitative Analysis the only way to get an edge in the market?
- Thread starter _eug_
- Start date
I would be very interested in a program that can interpret whether an upcoming news event will be tradable, not tradable or a non event. Or one that can decide based on market comditions to use market orders to enter or limit orders. I have yet to see one and I have seen a lot of them.
Hello _eug_,
I don't care about all that execution and psychology stuff any more. I don't manual trade anymore. My robot/algo (whatever you wanna call it) will not have psychology or emotions issues. The robot job will be to take trades per design and programmed. I don't care about all that news fear, greed, news release time, stuff, robot just do its job.
Once I find a profitable strategy that perform well during back test on NinjaTrader, i will trade via algo. I have to program the strategy to test it. If it performs well, I put money in it and click Start and NEVER click Stop again. It will either lose all money invested or make money. I will increase contract size once I double my initial investment. If it goes into major drawdown and lose all the money, it is what it is. I start over again.
While that algo is running making money, I will proceed developing and back testing more strategies in NinjaTrader. Rinse and repeat. Nice and easy.
I hope this makes sense.
Last edited:
Your problem here is you want "easy." None of this is easy and all of it requires continual maintenance and constant improvement. If you can't hang with this, find another game.
i960,
Great comment. Why would I want anything but simplicity, easy, systemagic approach to develop an effective algo trading system?
I know longer worry about hard work and time. Hard work is a constant in the goal and sucess equation.
Having fun and keeping things simple and easy is my priority, that money will follow as long as I keep it simple and work smart.
I disagree with the improvement part. But this is my opinion. Once I hit start on my algo. I'm not ever clicking stop. It will either make money long term per the evidence base back test or lose. Is just that simple. My job is to make sure my tools are accuracy and developement process gives me odds of sucess. Nice and easy
@SimpleMeLike I believe the point people are making and you seem to be missing is that trading, manually or automated is a DYNAMIC process. There is no easy, turn it on and let it run and go to the beach. If it were the quants and algo traders at Renisonce, ARQ, you name it shop would be at the beach, they are not.
Sounds all good in theory, actual execution is a different kettle of fish.
Back testing is an example, we design an algo and backtest it for 5 years and it shows a profit.
In reality it is different, would you really take all those trades? Answer: No!
Humans always want to fuck and fiddle.
These sound to me equivalent in theory but human has limited capacity in processing numbers so must go quantitative. I heard there is a 3 1/2 year old in Russia playing chess with pros but mind you this is outlier and less than 0.1% of traders will be able to beat the machines (pros) as with chess players. It will be hard to beat quants like this one who is one of my favorite. This is the procedure most of them follow:
Step 1: Let the computer analyze price action
Step 2: Computer results and trader input determine some model
Step 3: Model is tested in unused data to limit bias (Note 1)
Step 4: Model is then forwarded tested
Step 5: If forward test is good, then model is used to trade
Note 1. Bias will never drop to 0 but using a fresh sample is an unbiased test, or as said, out-of-sample results are unbiased estimates of actual system results.
Still success with above requires substantial work but in the right area. It is silly to try to do things computers can do much more efficiently.
If you compare a computer's ability to do QA with that of a human's, that's one thing. But no matter what form of analysis a computer is programmed to do, how can it do any better than recognizing and riding a trend, which (some) humans have been doing for 100 years without QA? Not to say that a computer can do such though
Last edited: