First off sorry for the late response. I have to say I'm impressed with your industriousness in working on this question and the way you laid everything out. I once had that kind of energy when I was a youngster lol.
I don't have a python interpreter installed, but I'm guessing I could run that on an online one?
Very interesting that in case one even to depth 40 it is a ~certainty that two identical games have been played. I think at some point in this thread we were saying it'd be like a 0.0001 chance. So it's amazing we went from those very low odds to virtual certainty.
"For case 2 (any of the games played by a specific player is identical to any of the other games played by other players) the probability is already relatively low around move 20."
But in your post you gave a value of "1" for that probability? I'm guessing that was a typo.
Now here is another thing to ponder that strikes me now. For all the first 3 cases the specific move order is NOT a factor. So in other words if two different move orders transpose to the same position you are counting that as identical games, whereas in the last 3 cases the move order must be identical. So in the first 3 cases, where move order is not a factor, what comes to mind for me is that I believe the probability as a function of depth should have a lowest value.
In other words as you start increasing the depth the probability will go down BUT in reality when you get to depth 80 at that point because the number of pieces remaining on the board is greatly reduced the probability should actually start going UP again. Think of how many games have been won with just a K and R vs K for example, and there are only so many checkmate positions possible with those pieces into which all possible move orders that end like that must converge. With that kind of depth I'd bet we are back at p=~1 again.
And this is a great example to illustrate the idea that theoretical calculations while appearing convincing and giving us great confidence, can be wayyyyyyyyy off by orders of magnitude because factor X was not taken into account :P
Still as you said it's very nice to get some numbers down to look at. Even if they are just a frame of reference that you might think of as a lower or upper bound.
I don't have a python interpreter installed, but I'm guessing I could run that on an online one?
Very interesting that in case one even to depth 40 it is a ~certainty that two identical games have been played. I think at some point in this thread we were saying it'd be like a 0.0001 chance. So it's amazing we went from those very low odds to virtual certainty.
"For case 2 (any of the games played by a specific player is identical to any of the other games played by other players) the probability is already relatively low around move 20."
But in your post you gave a value of "1" for that probability? I'm guessing that was a typo.
Now here is another thing to ponder that strikes me now. For all the first 3 cases the specific move order is NOT a factor. So in other words if two different move orders transpose to the same position you are counting that as identical games, whereas in the last 3 cases the move order must be identical. So in the first 3 cases, where move order is not a factor, what comes to mind for me is that I believe the probability as a function of depth should have a lowest value.
In other words as you start increasing the depth the probability will go down BUT in reality when you get to depth 80 at that point because the number of pieces remaining on the board is greatly reduced the probability should actually start going UP again. Think of how many games have been won with just a K and R vs K for example, and there are only so many checkmate positions possible with those pieces into which all possible move orders that end like that must converge. With that kind of depth I'd bet we are back at p=~1 again.
And this is a great example to illustrate the idea that theoretical calculations while appearing convincing and giving us great confidence, can be wayyyyyyyyy off by orders of magnitude because factor X was not taken into account :P
Still as you said it's very nice to get some numbers down to look at. Even if they are just a frame of reference that you might think of as a lower or upper bound.