How can I make a robustness test? Do you have a methodology?
Backtesting is useless
- Thread starter JCDST1979
- Start date
How can I make a robustness test? Do you have a methodology?
That's my point you can't know that it will make money doing backtesting, you are testing on only one version of the history.
I agree that backtesting can be free practice but it is biased because you already know what the market will do. I think paper trading will be a better way to practice, or, even better, real trading with very small positions.
I disagree that improves your odds of winning, you can't control your odds in real life, you don't even know how many faces the dice of life have.
I think it can tell you what would not have worked but no more than that.
The market is random but not unpredictable, meaning there's a process which drives the random outcomes. When you backtest you're trying to identify that process. As long as the *process* is deterministic (and they usually are to a high degree), you can make money.
That's the whole idea behind options trading and Black-Scholes-like processes.
The problem with making money this way is that:
1) If you work with a known theory like Black Scholes and it's adaptations, so does everyone. You can verify in backtests that even plain Black-Scholes really works AS LONG AS YOU CAN ADD ENOUGH SPREAD. But because the theory is well known, the spreads on the market are so tight that most of the times you can't make a trade which would guarrantee a profit. Say my calculated price for an options would be $10 with a standard deviation of $1. If I could sell at $11 or buy at $9, and do this a lot of times, I would make money guarranteed. Problem's the spread is like $0.1 and my calculation is not so exact for such a low spread. It may be the realized price is actually $10.5, if I sold at $11 I'd still make a profit on average, but with a $0.1 spread I cannot.
2) When according to the known theory, there are rare times when the market seems to be mispricing the options (in my example above, I'd get a $10 average price with a standard dev of $1 and I could actually sell it at even $12), well the known theory is wrong
And you know who tells me this? Backtests! So far from useless, backtests are very useful in telling you that almost everything you try will fail!3) Really the only chance of making money is to go some way that is not known by everyone. The way I see it, still identify the market or it's edge cases as "random" but driven by a deterministic process in the statistical way. Extremely freaking hard since if you don't start from the known theory, chances are very very low you'll arrive at something and in the best case that will be the already known facts. And if you start from the known facts, for each "original" thought process you think you may have, think how likely it is that one of the 10s of thousands of mathematicians / quants / traders out there have not arrived at it yet and either already discarded it as non working or actually implement it in their system so the market is already too tight when you arrive at it.
"deterministic process in the statistical way" Is this an oxymoron?.
I think that when you use statistics you're accepting your ignorance about the process, so there are no cause and effect relationships, no determinism.
For example, when playing the Russian roulette you have 1 of 6 probabilities of dying but you can never know when you are going to die. In real life and the market is even harder because you don't how many bullets and chambers there are.
I think that when you use statistics you're accepting your ignorance about the process, so there are no cause and effect relationships, no determinism.
For example, when playing the Russian roulette you have 1 of 6 probabilities of dying but you can never know when you are going to die. In real life and the market is even harder because you don't how many bullets and chambers there are.
I don't recall at the moment the exact term but "random process" and "no fucking clue" are different terms. If you throw a coin, you can't predict what the next toss will produce (head or tail) but you know if you throw it 100 times, you'll get head approximately 50 times. If you throw it 1000 times you'll get head approximately 500 times and so on. So the underlying process is deterministic. If the average number of times you'd get head when you threw it 100, 1000, 10000 times would be random, then you'd be in the "no fucking clue" territory.
So the "outcome" of a random process is well, random, but the "distribution" of the outcomes is deterministic.
As long as you guess the right distribution, right? It is a lot easier in a coin toss you only have 1 variable and two results.
In the markets, the only thing that you have is the distribution of the past. Every trader is a dynamic and emotional variable. Do you think you can come up with distribution for that?
Actually Nobel prizes were won precisely for coming up with that
That's why I recommend learning and understanding the current foundation before starting experimenting on your own, or else you'll discover on your own time (expense) that "two years in the lab are saving you two weeks in the library", otherwise said you risk to lose a lot of time only to rediscover what is already known.
@jcdst...Lets forget about backtesting for one moment....
Whats your thoughts on "experience"/ time at the game" as it relates to success and trading?
Is that not another form of "backtesting"??
Not sure where you are going on your backtesting opinions
And FWIW,and as i am sure you know,there are some very successful stat arb/mean reversion traders floating around.How do i know? I worked at a hedge fund and the head of the fund told me one groups profitability and the fact that they had less than 3 losing days in a full year..
Whats your thoughts on "experience"/ time at the game" as it relates to success and trading?
Is that not another form of "backtesting"??
Not sure where you are going on your backtesting opinions
And FWIW,and as i am sure you know,there are some very successful stat arb/mean reversion traders floating around.How do i know? I worked at a hedge fund and the head of the fund told me one groups profitability and the fact that they had less than 3 losing days in a full year..