So I have a simulation in mind, but don't quite know how to do it so wanted to see if anyone knows how or is interested. It's a test to determine luck vs. skill in a game (and quantify the luck or skill) with a slight negative edge..... please keep in mind this is a bit of a brain storm and not a complete and polished study that I'm proclaiming as truth. It's more like a hypothesis.

LET'S SAY:
There is a very simple game that you play against other people, and the true odds of winning for everyone that plays are 49%.

Hundreds of thousands of people play the game against each other, millions of times.

At any point in time there will always be winners and losers. The distribution of those should be normally distributed (I think) with the massive majority weighted towards the middle ranging between small winners and small losers.

But some (a very small percent, lets say top and bottom 1.5%) lucky and unlucky people will be very big winners and losers....even over extended periods. These would be those that fall into the tails of the distribution, due only to variance, not skill.

If you ran a simulation of games (lets say arbitrarily 20,000,000) for 10,000 participants. What percent of participants would be considered big winners and losers, and over what periods of time? This would be interesting, but could be even more useful.

If you could develop this model, you could change the 49% win rate and see how the distribution of big winners/losers changes. Who knows, at a 51% win rate, maybe there are 3 times more big winners, and 1/3 as many big losers. You could compare the distribution you calculate to the real world, and get an approximation of what the game of poker most closely resembles.

Does the real distribution of winners and losers suggest the best poker players have a 60% edge on average? or does it suggest a 50.5% edge or worse?

Does this make any sense at all?

Did you ever see Noah's blogs about variance in MTTs? The simulations he used were quite eye-opening, as even players with huge edges could lose over fairly large sample sizes.

No I didn't but glad you mentioned it. Read it, and it is eye opening (scary). Players will have to be more well-rounded to make it in the future...not just type of game but format. I personally like 6max SNG, 6max cash, and MTT. Variance seems to not have as much teeth this way (famous last words).

The simulation I described isn't so much about the brutalities of variance though, it's more about exploring what edge winners really have. I would hope that Amaya and other poker sites do calculations like these to see how rake changes affect their player base. Maybe they found that the majority of winners really won't be crushed by rake hikes because their edge is higher than people think.....(probably not though lol).

I'm sorry I couldn't help you more directly with your main line of query

I've often wondered how one would compute a cash game player's edge, but it seems like there are too many variables (such as rake, different playing styles with a variety of standard deviations), and too much variance to get a decent idea, even over what are generally considered "decent" sample sizes.

Have you tried playing with a cashgame variance simulator? I was quite shocked when first entered some of my own figures into that.

If someone has a winrate of 5bb/100 and a SD of 75bb/100, for example, there's roughly a 2.5% chance of them losing money in their next 100,000 hands, purely due to bad luck. (And these days, very few people are winning more than 5bb/100, which is why downswings are so commonplace).

If you enter zero for the winrate (indicating a breakeven player who is basically having an extended coin-flip), then the gap between max profit and max loss over a 100k sample is huge. Due to luck alone, the breakeven player could win anything up to about 80 buyins, but he has just as much chance of losing 80 buyins. The 95% confidence interval indicates that most "breakeven players" will have a winrate roughly in the range -4.75 and +4.75bb/100.

For your purposes, you could enter slight loss rates to see if it's still possible (through luck alone) for "losing players" to actually win over certain sample sizes. e.g. Someone whose skills indicate he "should" lose at 5bb/100 actually has about a 2.5% chance of making money in his next 100k hands. The "fish on a heater" really does exist, and sometimes those heaters can last quite a while.

FWIW, I think a "small" rake increase can actually have catastrophic effects in the long term. When winrates of grinders are measured in single digits, a rake rise that works out as 1bb/100 can turn a huge proportion of the field into losers overnight, although they won't actually realise it until they've played a large enough sample size to see the results!
You can see this by comparing the variance simulation graphs for 0bb/100 players and -1bb/100 players. With the former, 50% of players break even over 100k hands. When the rake knocks their winrate down 1bb/100, two thirds will now be losers. A 2bb/100 rake rise leads to 80% of the previously breakeven players going negative.
While that doesn't fully explain how rake impacts the game in the real world, or answer your original query, it should give you an idea how a high rake can destroy the perceived edge of winning players.