It's like you take a bunch of different people, you merge their faces and you get a perfectly average yet oddly beautiful and symmetrical one. I can't help noticing that in the end the heat maps are more similar than disimilar.
So maybe the solution is to examine the games once you _remove_ the heatmap. Once you eliminate the common, you might get the uncommon. Or complete garbage. I don't know :D
So maybe the solution is to examine the games once you _remove_ the heatmap. Once you eliminate the common, you might get the uncommon. Or complete garbage. I don't know :D
Well, if you remove the piece heatmaps, you're left with the pawn structures. So you might end up writing something like Sokolov's "Winning Chess Middlegames."
Interesting study. This can be useful for having a general idea about openings that you are not familiar with.
Would you consider open-sourcing your code? I am curious about heatmaps of the opening that I play. It would be nice to check out.
Would you consider open-sourcing your code? I am curious about heatmaps of the opening that I play. It would be nice to check out.
@KagitOynaCom said in #4:
> Interesting study. This can be useful for having a general idea about openings that you are not familiar with.
>
> Would you consider open-sourcing your code? I am curious about heatmaps of the opening that I play. It would be nice to check out.
The code is open-source: github.com/jk182/chessProjects/blob/main/pieceHeatmaps/pieaceHeatmaps.py
> Interesting study. This can be useful for having a general idea about openings that you are not familiar with.
>
> Would you consider open-sourcing your code? I am curious about heatmaps of the opening that I play. It would be nice to check out.
The code is open-source: github.com/jk182/chessProjects/blob/main/pieceHeatmaps/pieaceHeatmaps.py
Is this only valid in about openings. Would those notions of location stability versus volatility or trepidation rate per move (kidding a bit, tyring to warp my head around) also have some meaning thorugh a game. I guess starting in opening would be the maximum chess time (ply clock) window. I still need to keep reading. I only read the initial idea.
So is the question about how much do the opening choices determine pattern of heatmap visuals.
If pawn structure is the slowest dynamics in a whole game, it could actually be in relation to the heat map as a conseqeunce on the faster pieces abiliyt to change location more often. if I got any of this right. still looking.
So is the question about how much do the opening choices determine pattern of heatmap visuals.
If pawn structure is the slowest dynamics in a whole game, it could actually be in relation to the heat map as a conseqeunce on the faster pieces abiliyt to change location more often. if I got any of this right. still looking.
Great analysis.
I think you might get even more interesting results if you only look at moves rather than positions.
i.e. bishop on d3 that then stays there for 40 moves = 1 count of Bd3 rather than 40.
I would also consider looking at uplift, so rather than count of Bd3, doing ratio of Bd3 in QGD exchange to Bd3 in all other lines.
I think you might get even more interesting results if you only look at moves rather than positions.
i.e. bishop on d3 that then stays there for 40 moves = 1 count of Bd3 rather than 40.
I would also consider looking at uplift, so rather than count of Bd3, doing ratio of Bd3 in QGD exchange to Bd3 in all other lines.
@LCY_DollyParton said in #7:
> Great analysis.
> I think you might get even more interesting results if you only look at moves rather than positions.
>
> i.e. bishop on d3 that then stays there for 40 moves = 1 count of Bd3 rather than 40.
I thought about this, but decided against it because I feel it's more insightful to see how long a piece spends on each square.
Taking the Ruy Lopez for example, when White plays Bb5, Ba4 and Bb3 the bishop moves to every square once, but it spends significantly more time on b3 which gives a better idea where the piece will end up, not how it got there.
> I would also consider looking at uplift, so rather than count of Bd3, doing ratio of Bd3 in QGD exchange to Bd3 in all other lines.
Do you mean how often the bishop is on d3 in the QGD exchange compared to any other opening?
> Great analysis.
> I think you might get even more interesting results if you only look at moves rather than positions.
>
> i.e. bishop on d3 that then stays there for 40 moves = 1 count of Bd3 rather than 40.
I thought about this, but decided against it because I feel it's more insightful to see how long a piece spends on each square.
Taking the Ruy Lopez for example, when White plays Bb5, Ba4 and Bb3 the bishop moves to every square once, but it spends significantly more time on b3 which gives a better idea where the piece will end up, not how it got there.
> I would also consider looking at uplift, so rather than count of Bd3, doing ratio of Bd3 in QGD exchange to Bd3 in all other lines.
Do you mean how often the bishop is on d3 in the QGD exchange compared to any other opening?
I might have inverted the measure construction goal.
I find that one board to view them all, is a bit a cramped space to do heatmaps into. We might need more room.
This is punchline and crytpic to make sure i do not derail, i put the conclusion first.
So, since I am error prone in my reading, I realize that I might have been for a long time, and that always considering "whatelse" might be a compensation for that. whether I inverted volatility of piece location over the whole duration of game, or if it was about many games per player ID stability notion, not about the stability within one game, but sampled over many games on the same opening prefix at some depth for the whole game being binned in that opening name.. I am trying to fill in blanks, and that might be a nightmare for the author to fix.. but I try to wrap my head around what I could pay attention to so far.
There might be 2 dimensions about piece location changes that might be interesting to see at the same "time".
The location of the piece itself, versus it activitiy. but that would mean first going to do heatmaps in a more exploded set of planes or boards.
The appropriate visualization slices (usually 2D figures) might not always be all easily superposable on the plan of one chessboard.
I am not going chat-bot rambling. this is my own. Is the piece location really telling of the ensemble configuration.
I understand that the questin is about dynamics of configuration features. Long times scales of structure veruss fleeting time scales. (which i might invert w.r.t. to this blog, but both are linked and danse together). I think poetry is maybe better compromise between explanaotry rambling and concise cryptic...
You measure one compared to the others.. But that location is more about mobility than it is about actvity.
The reason I mention high dimensional work before projecting on our retinas (poetry here, of sorts), some explanatory slice, is that activity can also be visualize. but the overlaps of many pieces of such visualization as the piece abstract mobillity planes in the A0/LC0 terminology (I think) (btw, it may be confusing, that I call those move planes which are abstra potential modbility, the activity on that plane, but they are the same potential, the same figure of moblity is also the empty square potential activity, and once all pieces planes are combined the immediate activity left by substracting appropriate locatoins of all pieces.
this may not be readable very much. but it is about maybe tying with your other works. I have been stream of thoughts sharing so please do not call it spam. Not that you would, but i never know.
I find that one board to view them all, is a bit a cramped space to do heatmaps into. We might need more room.
This is punchline and crytpic to make sure i do not derail, i put the conclusion first.
So, since I am error prone in my reading, I realize that I might have been for a long time, and that always considering "whatelse" might be a compensation for that. whether I inverted volatility of piece location over the whole duration of game, or if it was about many games per player ID stability notion, not about the stability within one game, but sampled over many games on the same opening prefix at some depth for the whole game being binned in that opening name.. I am trying to fill in blanks, and that might be a nightmare for the author to fix.. but I try to wrap my head around what I could pay attention to so far.
There might be 2 dimensions about piece location changes that might be interesting to see at the same "time".
The location of the piece itself, versus it activitiy. but that would mean first going to do heatmaps in a more exploded set of planes or boards.
The appropriate visualization slices (usually 2D figures) might not always be all easily superposable on the plan of one chessboard.
I am not going chat-bot rambling. this is my own. Is the piece location really telling of the ensemble configuration.
I understand that the questin is about dynamics of configuration features. Long times scales of structure veruss fleeting time scales. (which i might invert w.r.t. to this blog, but both are linked and danse together). I think poetry is maybe better compromise between explanaotry rambling and concise cryptic...
You measure one compared to the others.. But that location is more about mobility than it is about actvity.
The reason I mention high dimensional work before projecting on our retinas (poetry here, of sorts), some explanatory slice, is that activity can also be visualize. but the overlaps of many pieces of such visualization as the piece abstract mobillity planes in the A0/LC0 terminology (I think) (btw, it may be confusing, that I call those move planes which are abstra potential modbility, the activity on that plane, but they are the same potential, the same figure of moblity is also the empty square potential activity, and once all pieces planes are combined the immediate activity left by substracting appropriate locatoins of all pieces.
this may not be readable very much. but it is about maybe tying with your other works. I have been stream of thoughts sharing so please do not call it spam. Not that you would, but i never know.
Could it be used to train a chess neural network?
How it could work?
How it could work?