Quote from: defxor on September 04, 2011, 10:21:58 PM

Quote from: the joint on September 04, 2011, 09:46:39 PM

.00000009279 = .000009279% chance I will flip at least 36 heads (i.e. trade positively at least 36/40 times)

Edit 2: You do realize that the chances of trading at 90% profit become less likely the more times I trade...right? I hope you weren't implying that my chances of doing this across 40 trials were good, were you?

Nah I wanted to see if you considered after-the-fact to be the same as before-the-fact, which you did. Even if we go by "approx" and "about" being at the top of your own estimation - your numbers show that if there are 11 million traders in the world one will indeed be so lucky by chance alone.

That's what the whole "traders don't outperform chance" means. We talk about the successful ones, not the ones that don't beat the market.

So, I propose the following: For 40 days you'll post if you believe the price of BTC will be higher or lower than the day before, MtGox GMT timezone. Now if we're indeed on a steady downward slope due to inflation there should be a slight skew, but I'm quite sure you won't hit 36/40.

(The point being that after-the-fact is a lot easier, all you have to do is to be selective as to which trades you count)

I've never so far had anyone doing TA take me up on such a wager anyway.

The problem is that you are assuming that market movement is independent of the individual trades.

Suppose I guess "the market will move up 40 consecutive days"

I then make 40 consecutive trades, 1 on each day, buying enough coins such that the market goes up for 40 consecutive days.

At this point, my prediction was correct. The market moved up 40 consecutive days in a row. But, I did not profit 40 consecutive days in a row.

"Outperform" implies the action of the traders. This can either mean "outperform" in terms of profiting, or "outperform" in terms of guessing market movement. But the 2 are not always mutually exclusive.

When a large seller dumps coins, his action impacts the market. Suppose the price has been relatively steady for a few days at $9, bouncing between $8.80 and $9.20. Hard to guess which way it's gonna go in an hour, right?

Now, suppose one such seller dumps 5000 BTC dropping the price from $9 to $8. Care to make a new guess of where the price is going to go?

It seems you are suggesting that market movement is independent of the prior trades. So, for example, if I were to guess what the market price would be 1 month from now, I would obviously not have the data available to me between now and those 30 days. During the passing of those 30 days, I would accumulate more data with which I could adjust my guess. It would be hard to guess the value in 30 days because I do not know what the choices of other traders will be between now and then. However, the trades that occur between now and then are NOT random.

The trades that occur between now and then will be made by traders who ARE considering previous trades, and the trades they make continuously refine the guesses made by others. I cannot guess what the next trade that is made is going to be. But, when you confound the 'guess' with my desire (and the desire of everyone else) for profit, things get more interesting. You are, in effect, placing a new constraint on the 'guess.' The guess no longer is a pure guess...the parameters become more refined because you know everyone wants to make a profit.

Suppose I knew where every individual's buy-in price was. This is a factor that contributes to the current market price (after all, the current market price is only the current market price due to the trades made to date). Add to this the total sum of $ available to each individual to purchase additional BTC. This piece of information means there is a known limit to value of BTC. The minimum value is 0, the maximum value is the total amount of $ in the economy divided by the total number of BTC. This places another limitation on the possibility of the value. If there is only (for exmaple) $1,000,000 that can possibly be put into the BTC market, and only 1000 BTC available, you know that the value of BTC can never exceed $1000. So, no 'guess' can be made that the value would be over $1000. It is an absolute certainty that the value will be below $1,000. It is impossible that the value will be over $1,000.

The point is, with every additional piece of information available AT THE GIVEN MOMENT (before the fact) you can make a more educated guess. Can I guess whether or not the next trade will be a buy or sell? Not without knowing who the next seller will be or what their intention is. If I know trader x will buy 100 BTC at y time, then I know the market will go up then. Suppose I also know that the most BTC that any single person has is 10,000. While I don't know the exact value of the next trade, I do know that it will be impossible for that trade to exceed 10,000 BTC and I can be absolutely certain it will be less than 10,000 BTC. It is certain that the next trade will be a trade of amount x BTC. The more factors I can take into account, the more I will be aware of the constraints leading to this certainty.

So, back to what you're getting at. If I make guesses on up/down movement for the next 40 days, will I hit 90% again? Probably not.
BUT, saying this is a lack of outperforming 'chance' is an inappropriate statement.

You are, in effect, correlating chance with "unknown" factors that have nothing to do with chance at all. They are, in fact, variables that are available and can be known, but simply aren't to you.
There are, however, factors which are known and available at the present moment. These include current price, market depth, your own intentions, the current number of BTC in the economy, etc.

In particular, the market depth is the best indicator for predicting future movement. This is because from this you can infer the general 'intention' of the market. It's like an emotion - you can get a gist for how the market 'feels' at a given moment. This is why, as you say, if I were to predict that the market would go down tomorrow, I may have a better chance of being right since, as you acknowledged, it seems as if we are on a slight downward slope.

In conclusion, the problem I see with your argument is that you are assuming a cross-sectional piece of data is strictly that...cross-sectional, and that a singular cross-sectional event cannot be used to predict a future event. But, in reality, the cross-sectional event is not truly cross-sectional - it speaks of after-the-fact conditions. That is, the event itself indicates constraint of potential future movement. This makes the 'chance' analogy an invalid one.

Edit: Here's food for thought.
Only 1 of 3 things can happen. Event A (market goes up) Event B (market goes down) or Event C (market stays the same).
If event A happens, events B and C are impossible.
If event B happens, events A and C are impossible.
If event C happens, events A and B are impossible.
There is no possibility. There is only impossibility since only A B or C will happen.
In other words, there is only certainty (no chance) that one of these will happen, so there is only certainty (no chance) that 2 will not.
Chance is impossible because chance implies that A B or C could happen when in actuality only 1 WILL happen.

Could I have gone to the market today at noon instead of sleeping in? Nope. Never could have happened. Why? It didn't happen. Could I go to the market in 1 hour from now? Not if I don't go 1 hour from now. One of these options (going or not going) is impossible and there is absolutely zero chance that it will happen. The other isn't possible...it is certain to be.