passed pawn question

[asanjuan]

Thanks Tom, very useful!

[asanjuan]

One 30K game match shows a -20 ELo change doing this. I am re-running to verify. All I did was what was mentioned. Normally one requires no pawns on either adjacent file one or more ranks in front of the passed pawn. I changed it to 2 or more ranks for those adjacent files. The eval in Crafty, for your position, changed by +1.something which is a big jump...

Yes +1 makes a lot!
I am still in the "bug hunting mode", I wonder how many silly bugs like this one I can find inside Rhetoric.
Thanks for your feedback, it's been very useful.

[Lyudmil Tsvetkov]

I would say that is a typical "Candidate" Passed Pawn.

As long as this pawn isn't a passed pawn from a static point of view i would handle it in a seperate way.
Having a bonus for such a pawn is a good idea anyway.

For reference, take a look at the 2 books from Hans Kmoch (maybe the god in pawn play ?) :

1. In french (maybe there is a English version but I don't know) : "L'art de jouer les pions"

2. In English : "Pawn power in chess"

Here is a small excerpt from the second one :


I-§2: On the road to promotion

The promotion of a pawn depends basically only on the obstruction
by opposing pawns. There are two types of obstruction, mechanical
and dynamical.

Most hampering is the mechanical obstruction offered by the
pawn's counterpawn, e.g. Pd2 vs Pd7. Every pawn is pawn and
counterpawn at the same time; it depends on the observer's point of
view.

Originally every pawn is unfree owing to mechanical obstruction.
Removal of its counterpawn makes a pawn half- free.

The half-free pawn meets dynamical obstruction on the part of
its opposing neighbors whom we call sentries; normally an innerpawn
faces two of them, a rimpawn only one. The sentries both guard the
same square or two different squares. The sentries of White's d-pawn,
for instance, are Black's c-pawn and e-pawn; in case of Pd4 vs Pc6,
Pe6 they both guard against d4-d5, while in case of Pd4 vs Pc7, Pe6
one guards agamst d4-d5, the other against d5-d6.

Dynamical obstruction is not absolute; the half-free pawn may
march through, and is therefore called a candidate-a candidate for
full freedom and promotion.

The promotion of a candidate depends on assistance by its own
neighbors whom we call helpers. The helpers of White's d- pawn, for
instance, are White's c-pawn and e-pawn. In the position ofPc2, Pd2,
Pe2 vs Pc7, Pe7 the d-pawn's bypassing of its sentries is assured.

The operation of helping a candidate to cross the guarded
square or squares should start wIth the advance of the candidate itself.
"Candidate first" is the rule for such cases. Other initial pawn moves
are basically unreliable.

Helpers and sentries neutralize each other if there is a helper
for every sentry. A half-free pawn with inadequate help is no true
candidate but afaker. In the position ofPd4 vs Pe6, for instance, both
pawns are fakers, each one lacking the necessary helper. The same
with Pd4 in the formation of Pc2, Pd4, Pe2 vs Pc7, Pe7, Pf6 when Pe2
is paralyzed by f7 -f6; this helper needs a helper's helper, e.g. Pf2, which
assures the consecutive crossing of eS and d6.

In other words: the passing of a candidate depends on its
belonging to a majority of pawns (which mayor may not be just local).
Once a candidate faces no more obstruction on the part of
sentries, it is free, as are all pawns in the position of Pd4, Ph2 vs Pc4,
Pf6. A free pawn is called a passed pawn or, as we prefer to call it for
short, a passer.

A passer is basically superior to an unfree pawn or a candidate.
The shorter its frontspan is, the greater the value of a passer.
For instance Pf5 vs Pc6 favors White.




Another factor of importance in the relative value of passers is
their horizontal distance from the bulk of the pawns; the greater this
distance is, the better. It constitutes the so-called advantage of the
outside passed pawn (Diagram 5). The outside passer counts particularly
in pawn endings, inasmuch as it forces the opposing King to
stray far away from its own pawns.

There is also distinction with regard to protection. A passer
protected by one or two pawns is a protected passer (Diagram 6) and
superior to an ordinary passer. But it is exclusively the protection by
pawns that counts.

An unfree pawn or a faker may suddenly become a passer of
decisive power by means of a a sacrificial combination. We call such
a pawn a sneaker. The sneaker's outstanding quality is almost invariably its short frontspan and outside location, e.g. PaS, Pb4, PdS
vs Pa6, Pc7, Pd6, when the unfree Pa5 sneaks through after 1 b5,
a6xb5; 2 a6.



In the popular example of Diagram 7 White has two potential
sneakers, one of them marching through as follows: 1 d6, exd6 (or 1
... cxd6; 2 e6); 2 c6, dxc6; 3 e6 and wins (assuming that no piece can
interfere). Note the importance of the original span situation; if
Black has the move, he correctly starts with 1 ... d6 but must answer
2 exd6 with 2 ... exd6, as 2 ... c6 obviously makes no sense.

I think if somebody try to use all the ideas from Kmoch books about pawn play, this engine will be very very good ! But difficult to translate all these ideas in programming lines ...

It is interesting how these threads suddenly resuscitate from the past...

Now that Alberto brought back to life this thread, I read something I did not red before and decided I could prolong the thread's life a bit longer

However, referring to Ben's post, I must say that there are some problems with Mr. Kmoch's concepts. I will take just a single case.

You have there the concept of a 'sneaker'. Well, what does that actually mean, and how widely applicable it is?

Below you have two positions with sneakers. In the first one, a simple pawn endgame, the sneaker (h5) will win the game, but only because it is more advanced than the enemy pawns and the enemy king is unable to reach it on time. But this will be valid only for pawn edngames and with very specific additional arrangements that would favour the sneaker.

In the second one, just adding a knight each side, or you might add whatever other piece/pieces you want, the sneaker suddenly disappears, because the knight will be able to stop its advance instead of the king. So that, basically, in the second diagram, the sneaker is pretty much meaningless, and all I see is a backward white pawn on g4; black is actually better.

[d]8/8/7p/1pk4P/6P1/2K5/8/8 w - - 0 1

[d]8/5n2/7p/1pk4P/6P1/2K2N2/8/8 w - - 0 1

The conclusion might be that a sneaker would be an useful term in only about 5-10% of cases. Who needs such a knowledge, when it is not generalised in the least?

A backward pawn, on the other hand, would be useful in some 90-95% of cases, definitely an useful one.

Just my 2 cents, it is all to blame on Alberto