It's important to understand - on the whole - he average probability of guessing the direction of a trade is 50/50. Any obvious (and virtually all non-obvious) edges, arbitrage for instance, are removed by the market participants. That's what a market IS.
In that respect, 'random walk' is correct. However, that does not mean all participants cannot expect profitability over the short or long term. Which are also themselves, very different things.
For retail traders, transaction cost (commissions + bid/ask spread) make it a negative expectancy.
In fact, bid/ask spreads and commissions ARE PRICED right at a level to remove any positive expectancy, but leave the APPEARANCE of a possible positive one.
Getting that up just a couple percentage points to a positive expectancy is difficult in the short term, and VERY difficult over the long term. Not impossible. That's why about 70-80% of active retail traders are losing overall at any given time.
For firms, order flow and much lower transaction costs make for a positive expectancy. The value of order flow alone is responsible for the zero-commission trading industry.
Skill isn't the issue. The trader with skill level at lose a little or break even trading - from his bedroom - would very possibly be winning on the big banks trading desk. Even if executing the SAME trades they were at retail. More so, the more frequent the trading activity.
So if big institutions have a built in edge why do they sometimes fail? Easy, they still sometimes make stupid bets (ie.. Silicon Valley Bank's bond bet which killed it).