Evaluation of Local Endgames with Long Sequences
Posted: Mon Apr 06, 2026 11:20 pm
It is insufficient to consider values of followers and individual moves. It is also mandatory to verify whether long alternating sequences of several plays are worth playing successively to know from which followers to derive the initial move value.
I criticise every unverified local endgame evaluation anywhere in the world involving long sequences. Without verification, alleged evaluation is just guessing whether a calculated value is, or is not, the right one.
In a simple gote, it is obvious that Black's alternating sequence comprises one move and White's alternating sequence comprises one move, there is no other way to derive values so it clear we have a local gote without verifying it.
In a simple local endgame with one player's possible alternating sequence of two moves (he starts then the opponent replies), it might be a simple gote with the player's one-move sequence or it might be a simple sente with his two-move sequence (he starts then the opponent replies). We can assume one of the two possibilities, compare the tentative initial move value to the follow-up move value, and determine the local endgame as a local gote (decreasing move values), ambiguous (constant move values) or local sente (increasing move values). After this verification, we know whether the tentative is the correct move value or whether we have another type of local endgame with the other move value (gote move value instead of sente move value or vice versa). Without this verification, we would only have guessed whether our tentative move value is the correct move value or whether we have calculated a wrong move value.
If Black's or White's alternating sequences are long, there are more possibilities for calculating various tentative move values. Usually, only one possibility is correct while every other possible calculation generates a wrong move value. Therefore, with long alternating sequences available, it is all the more important to verify for every suggested tentative move value whether it is the correct one.
Verification of one local endgame by thermography takes days or weeks so is impractical. Verification according to Sensei's Library (comparing a tentative initial value to other move values) is wrong, as I have shown. Verification works in practice by my method of making hypotheses, which compares a tentative initial value to the gains of Black's alternating sequence and to the gains of White's alternating sequence. Verification is confirming that a tentative initial value is at most each of the gains. Start hypotheses with longest possible alternating sequences. If necessary, try shorter ones.
Therefore, the gains of all moves of Black's alternating sequence and White's alternating sequence must be calculated! Since every gain is derived from the counts of the positions before versus after a move, the counts of all the positions along Black's alternating sequence and White's alternating sequence must be calculated!
Since follow-up positions might already depend on long sequences, iterate the process as necessary backwards, like you would do for careful tactical reading.
The major work of determining a correct move value depending on long sequences is verification! Instead, only guessing whether some arbitrary calculation of a tentative move value determines the possibly correct move value is just that: guessing - and the value might be right or, more likely, wrong because there are many more possibilities of wrong derivations than the only one correct derivation.
Do not stop where everybody else stops - providing some random calculation and guessing that the thereby derived move value would be correct. Instead, do determine the counts of the followers, do determine the gains of all the moves of both alternating sequences and do verify every tentative move value by comparing it to all the gains! Move value calculation is not guessing but it is verification!
I criticise every unverified local endgame evaluation anywhere in the world involving long sequences. Without verification, alleged evaluation is just guessing whether a calculated value is, or is not, the right one.
In a simple gote, it is obvious that Black's alternating sequence comprises one move and White's alternating sequence comprises one move, there is no other way to derive values so it clear we have a local gote without verifying it.
In a simple local endgame with one player's possible alternating sequence of two moves (he starts then the opponent replies), it might be a simple gote with the player's one-move sequence or it might be a simple sente with his two-move sequence (he starts then the opponent replies). We can assume one of the two possibilities, compare the tentative initial move value to the follow-up move value, and determine the local endgame as a local gote (decreasing move values), ambiguous (constant move values) or local sente (increasing move values). After this verification, we know whether the tentative is the correct move value or whether we have another type of local endgame with the other move value (gote move value instead of sente move value or vice versa). Without this verification, we would only have guessed whether our tentative move value is the correct move value or whether we have calculated a wrong move value.
If Black's or White's alternating sequences are long, there are more possibilities for calculating various tentative move values. Usually, only one possibility is correct while every other possible calculation generates a wrong move value. Therefore, with long alternating sequences available, it is all the more important to verify for every suggested tentative move value whether it is the correct one.
Verification of one local endgame by thermography takes days or weeks so is impractical. Verification according to Sensei's Library (comparing a tentative initial value to other move values) is wrong, as I have shown. Verification works in practice by my method of making hypotheses, which compares a tentative initial value to the gains of Black's alternating sequence and to the gains of White's alternating sequence. Verification is confirming that a tentative initial value is at most each of the gains. Start hypotheses with longest possible alternating sequences. If necessary, try shorter ones.
Therefore, the gains of all moves of Black's alternating sequence and White's alternating sequence must be calculated! Since every gain is derived from the counts of the positions before versus after a move, the counts of all the positions along Black's alternating sequence and White's alternating sequence must be calculated!
Since follow-up positions might already depend on long sequences, iterate the process as necessary backwards, like you would do for careful tactical reading.
The major work of determining a correct move value depending on long sequences is verification! Instead, only guessing whether some arbitrary calculation of a tentative move value determines the possibly correct move value is just that: guessing - and the value might be right or, more likely, wrong because there are many more possibilities of wrong derivations than the only one correct derivation.
Do not stop where everybody else stops - providing some random calculation and guessing that the thereby derived move value would be correct. Instead, do determine the counts of the followers, do determine the gains of all the moves of both alternating sequences and do verify every tentative move value by comparing it to all the gains! Move value calculation is not guessing but it is verification!