The first way we know is that because if a strategy existed which consistently resulted in EV+, then one single entity could follow it and eventually own the entire market, at which point it's no longer market. Markets continue to exist, therefore no EV+ strategy exists.
That's a valid conclusion, but it is based on assumptions that are not applicable to reality, so the conclusion isn't either.
There is also not
one strategy in poker that allows a good player to dominate all other players to the point where he essentially extracts all winnings. There isn't
one immutable algorithm either in computer poker doing this.
In reality, they are complex, and constantly changing strategies that are competing against each other. Strategy A might win against clearly inferior strategies B and C most of the time, but it could underperform against a strategy D that, on the other hand performs less well against B and C than A (though it still wins against them, just less).
A better approximation to
actual trading and predicting, rather than some armchair notion of "the strategy that dominates all others" are predictive
factors, anyway.
Harvey et al mentions a few of them, with the non-surprising result that 'momentum' comes out pretty high.
The other way we know it's a game of pure chance is because when enough data is collected on the lifetime behavior of all traders, the results are consistent with playing a game of chance, with about the number of outliers one would expect given the sample size (every once in a while 20 random coin flips will come up all heads in a row).
That's one of the stronger arguments against trading, I admit. However, if the claim simply is "a vast majority of traders will not be EV+", then there's no disagreement between us, but that's not quite the same as declaring trading to be purely a process of chance.
Another game comparison to illustrate that point (chess this time, as it's now about skill): Let's say the average beginning chess player's Elo score is 900 (not sure if it is, but let's say). Let's set a chess algorithm to that value, plus a bit on top (say by limiting the amount of halfturns it is allowed to calculate ahead). If you let the entirety of all chess players play against this computer, and assuming that beginners outnumber the higher Elo players greatly (which they actually have to if we define skill through Elo), then a vast number of players will never stand a chance against that computer. Furthermore, if you would only look at the aggregate results, you might conclude that the remaining 5%, maybe 10% of players that
do beat our crippled chess computer are simply the lucky outliers, and that skill had nothing to with it (similar to the "coin flip" analogy you employ).
Of course, you'd be wrong in the case of chess, but the point is: from the experimental setup, you couldn't tell the difference, because you are looking at /all/ players against some arbitrary goal (long term profitability), instead of looking at a carefully selected group trying to find out if there is a correlation between some skill set and performance.
Poll
