Handicapping: What is a take back worth?

[CRoberson]

Any research on the Elo value of a take back?

From my experience with my son, I will guess 200 Elo per take back. Of course, this may vary based on the position.

[Guenther]

Any research on the Elo value of a take back?

From my experience with my son, I will guess 200 Elo per take back. Of course, this may vary based on the position.

My guess is it is less and it is not linear too.

Guenther

[Uri Blass]

Any research on the Elo value of a take back?

From my experience with my son, I will guess 200 Elo per take back. Of course, this may vary based on the position.

You need to define the conditions of takeback and time control.

It is not clear to me what to do about times on the clock after takeback.

Suppose we started a game at time control of 90+30(the normal time control that I have in tournaments).


Suppose I thought 10 minutes played 1.e4 and you thought 1 minutes about e6
and I think another 20 minutes about the second move and decide that I want to take back 1.e4
What is going to be the times on the clock?

Do we take back the times on the clock of both player to the time before 1.e4 or maybe we do not take back the time and continue normally?

Note that I guess diminishing returns for take back and I do not expect 2000 player to beat anand in a match even if he is allowed 20 take backs during the game.

[CRoberson]

The conditions for this definition are as follows.

Two humans are playing. One thinks for m minutes then makes a move. Within a few seconds, he/she takes their move back and punches the other persons clock.

Time is not adjusted for the take back.

[CRoberson]

Any research on the Elo value of a take back?

From my experience with my son, I will guess 200 Elo per take back. Of course, this may vary based on the position.

My guess is it is less and it is not linear too.

Guenther

Now that I've had some sleep, I have a hypothesis for a formula.

I believe that it is nonlinear. Some positions keep you from losing a pawn while another will keep you from losing a rook and so forth.


I suppose that can be the basis of an average value and that it depends on playing strength. We might assume that players of 1800 level are less likely to miss dropping the queen and more likely to miss dropping a pawn. OTOH, players of 1000 might don't pay complete attention for any piece.

As players improve, do they learn to protect mate then protect the queens ..... then pawns?

In the early days with my son, I could win a queen easily. Later, rooks were easier than queens to win. Now, minor pieces and pawns are easier. I am guessing this is a normal progression for most people.

Assuming that is true, the value is based on the players stage/rating/strength. So, somebody who drops rooks or less, but not queens has an average take back value of (Rook+minor)/2 = 4 pawns. IIRC, a pawn is worth about 100 Elo meaning that is an average take back value of 400 Elo.

When a player gets to only dropping minor pieces and pawns, then average take back value = (3+1)/2 * 100 = 200 Elo.

Maybe, the way to do it is count each take back and add it up independently instead of using average values. Last night, my son had 3 take backs in a game. Two saved him a minor piece and the third saved a pawn. Thus, the take backs were worth a handicap of (6+1) * 100 = 700 Elo.

Anybody see a better way to calculate it?

[Uri Blass]

Any research on the Elo value of a take back?

From my experience with my son, I will guess 200 Elo per take back. Of course, this may vary based on the position.

My guess is it is less and it is not linear too.

Guenther

Now that I've had some sleep, I have a hypothesis for a formula.

I believe that it is nonlinear. Some positions keep you from losing a pawn while another will keep you from losing a rook and so forth.


I suppose that can be the basis of an average value and that it depends on playing strength. We might assume that players of 1800 level are less likely to miss dropping the queen and more likely to miss dropping a pawn. OTOH, players of 1000 might don't pay complete attention for any piece.

As players improve, do they learn to protect mate then protect the queens ..... then pawns?

In the early days with my son, I could win a queen easily. Later, rooks were easier than queens to win. Now, minor pieces and pawns are easier. I am guessing this is a normal progression for most people.

Assuming that is true, the value is based on the players stage/rating/strength. So, somebody who drops rooks or less, but not queens has an average take back value of (Rook+minor)/2 = 4 pawns. IIRC, a pawn is worth about 100 Elo meaning that is an average take back value of 400 Elo.

When a player gets to only dropping minor pieces and pawns, then average take back value = (3+1)/2 * 100 = 200 Elo.

Maybe, the way to do it is count each take back and add it up independently instead of using average values. Last night, my son had 3 take backs in a game. Two saved him a minor piece and the third saved a pawn. Thus, the take backs were worth a handicap of (6+1) * 100 = 700 Elo.

Anybody see a better way to calculate it?

I do not think that you can calculate it in that way.
The value of a piece is not constant in elo terms and at high level it is higher.

I also think that chess is also a positional game and if
a player with rating 2000 plays against a player with rating 1000 then the 2000 player may get a decisive positional advantage and if the weak player tries only to take back tactical mistakes he is going to lose even with 100 take backs(I am not sure that 100 take backs are going to be enough with a different strategy but the best strategy for the weak player is also to take back positional mistakes).

[CRoberson]

I do not think that you can calculate it in that way.
The value of a piece is not constant in elo terms and at high level it is higher.

I also think that chess is also a positional game and if
a player with rating 2000 plays against a player with rating 1000 then the 2000 player may get a decisive positional advantage and if the weak player tries only to take back tactical mistakes he is going to lose even with 100 take backs(I am not sure that 100 take backs are going to be enough with a different strategy but the best strategy for the weak player is also to take back positional mistakes).

I agree. That system by itself is insufficient. A more correct system may have multiple components with each component geared to a certain level. However, my current question pertains to those below 1800. IOW, this would be a starting point for thought on it or to build on.

As far as the utility of it, it could be used by those teaching chess as opposed to other handicapping techniques.

[syzygy]

The conditions for this definition are as follows.

Two humans are playing. One thinks for m minutes then makes a move. Within a few seconds, he/she takes their move back and punches the other persons clock.

Time is not adjusted for the take back.

So if the first player plays a bad move and the other player spots it faster than the first player and makes a move, it will be too late for the first player to take his move back? Under those conditions the advantage seems very small. If you didn't spot that your queen was en prise for m minutes, you will likely not see it during the next few seconds either.

What about the option of taking a full move back? So if it is the first player's move, he can either play a regular move OR (once) take back two half moves and play a different move?

Regarding the observation that players make "less severe" mistakes as they get better, my guess is that this is compensated by the fact that "less severe" mistakes at a higher level will have more severe consequences. With both players at ELO 1000, losing a queen does not yet lose the game. With both players at ELO 2000, losing a pawn without compensation is likely decisive.