On the "Planet of PRs", why would any ape make "PR sacrificing moves for tactical reasons"?
Warning! This is a trap question. Don't try to answer. ;)
MK
On Wednesday, October 13, 2021 at 9:26:49 AM UTC+1, MK wrote:
On the "Planet of PRs", why would any ape make "PR sacrificing moves for tactical reasons"?
Warning! This is a trap question. Don't try to answer. ;)
That's when a player (like Tim for example)
says (to himself) "I don't actually think it's good enough to
double but I'll double anyway because I've seen that my opponent
(for example, Jim Plaskett) often passes this type of position."
It's a good tactical double but it sacrifices PR.
On the "Planet of PRs", why would any ape make "PR sacrificing moves
for tactical reasons"?
On 2021-10-13, MK <mu...@compuplus.net> wrote:
On the "Planet of PRs", why would any ape make "PR
sacrificing moves for tactical reasons"?
Because in some circumstances it's the right thing to do.
Why else ?
An example that Stick casually mentionned more than once
(in the bgonline.org forum for sure, maybe here as well), but
didn't really explain as far as I can remember, is passing clear
takes against weaker players in simple endgame positions,
like races or advanced anchor holding games.
Consider this position:
Between equal players, this is a straightforward D/T, any
different action is en error of about 0.1.
If X is weaker than O by 3 PR points, it is a pass!
The "tactical" qualifier in your question may not even be needed.
This is only mathematics, the uneven curvature of the equity
when X and O are of different skills, something like that.
There are a few real mathematicians here that could certainly
express it more rigorously.
Of course you have to realize that it may matter, then study
when and how much it does, then decide that, when it's the
right thing to do, you will do it and silently laugh at the raised
eyebrows of the kibitzers.
Since you are willing to engage, let me ask a few more questions
in trying to understand better.
1) Isn't PR is the average of checker and cube errors? If so, how
would that "uneven curvature of the equity mathematics" apply
"when X and O are of different skills" of checker and cube play
separately and disproportionately?
2) a- Since this is obvious and common knowledge, what would
keep the underdog from manipukating it to his advantage?
b- Especially if the average PR underdog by +3, is in fact stronger
in checker or cube PR and can use this to his advantage in positions
where being better in checker or cube skill matters and coincides
with his being better in checker or cube skill?
3) As a minor issue, does a player's PR stay static forever? What
if the underdog improves and/or the favorite declines in PR?
4) a- As a minor issue, does PR equate to "predictability"?
b- If a human player makes such tactical moves against a bot
rated at +3 PR and also perfectly consistent/predictable, can you
run a test to prove that they will even benefit in that case?
On 2021-10-18, MK <mu...@compuplus.net> wrote:
1) Isn't PR is the average of checker and cube errors? If so, how
would that "uneven curvature of the equity mathematics" apply
"when X and O are of different skills" of checker and cube play
separately and disproportionately?
I think it doesn't matter,
The general idea is "if the opponent butchers games that are
average to difficult, keep the cube low in the ones that are easy".
2) a- Since this is obvious and common knowledge, what would
keep the underdog from manipukating it to his advantage?
He can't do that, since it applies in simple situations where unnatural
plays will be errors and cannot distract the stronger opponent.
b- Especially if the average PR underdog by +3, is in fact stronger
in checker or cube PR and can use this to his advantage in positions
where being better in checker or cube skill matters and coincides
with his being better in checker or cube skill?
Again, this applies in positions where little skill is needed.
3) As a minor issue, does a player's PR stay static forever? What
if the underdog improves and/or the favorite declines in PR?
Of course PR vary, and is never really known precisely (be it that of
the opponent or even one's own).
4) a- As a minor issue, does PR equate to "predictability"?
Since a low PR means playing closer to the bot, yes, a low PR
implies more predicatble play.
b- If a human player makes such tactical moves against a bot
rated at +3 PR and also perfectly consistent/predictable, can you
run a test to prove that they will even benefit in that case?
I'm not sure what you mean. Is "they" the human player and the bot
playing at a weakened level ? In this case, if the bot's weakening is cleverly done and it makes errors in "humanly" difficult positions but
not in simple ones, I think "they" will benefit as well. If the bot is weakend by adding some random noise to evaluations of all positions it
won't work since every position will be similarly "difficult" to it.
On October 19, 2021 at 4:35:26 PM UTC-6, Philippe Michel wrote:
On 2021-10-18, MK <mu...@compuplus.net> wrote:
1) Isn't PR is the average of checker and cube errors? If so, how
would that "uneven curvature of the equity mathematics" apply
"when X and O are of different skills" of checker and cube play
separately and disproportionately?
I think it doesn't matter,
How not? If one player is 4 cube PR and 6 checker PR, with an
average 5 PR. If the other player 12 cube PR and 4 checker PR,
with an average 8 PR (i.e. 3 worse than the other). How can you
claim that an average 5 PR can exploit the 8 PR but a 4 checker
PR can't exploit a 6 checker PR or that a 4 cube PR can exploit
a 12 cube PR even worse...??
The general idea is "if the opponent butchers games that are
average to difficult, keep the cube low in the ones that are easy".
Right here, before we go on, let's make a record of what you are
saying: "stronger player should not give the weaker player a chance
to get lucky in positions where not much checker skill is needed,
in other words "cube magnifies luck!".
Please confirm and agree.
4) a- As a minor issue, does PR equate to "predictability"?
Since a low PR means playing closer to the bot, yes, a low PR
implies more predicatble play.
With that, will you agree that logically the "less predictable" player
can better exploit the "more predictable" player?
On 2021-10-20, MK <mu...@compuplus.net> wrote:
How not? If one player is 4 cube PR and 6 checker PR, with an
average 5 PR. If the other player 12 cube PR and 4 checker PR,
with an average 8 PR (i.e. 3 worse than the other). How can you
claim that an average 5 PR can exploit the 8 PR but a 4 checker
PR can't exploit a 6 checker PR or that a 4 cube PR can exploit
a 12 cube PR even worse...??
You can't average PRs like this.
First, as far as I know, XG doesn't have separate numbers for
checker and cube play.
GNUbg does but they must be weigthed by the number of
decisions (non-forced checker moves and what it calls
"actual or close cube decisions").
In general, checker play error rate matters more (3 times more ?
5 times more ? that's in this area).
The general idea is "if the opponent butchers games that are
average to difficult, keep the cube low in the ones that are easy".
Right here, before we go on, let's make a record of what you are
saying: "stronger player should not give the weaker player a chance
to get lucky in positions where not much checker skill is needed,
in other words "cube magnifies luck!".
Please confirm and agree.
I can't since I didn't say that...
I never implied checker skill specifically.
Anyway, cube doesn't magnify luck. When accepted (not all the
time, then), cube magnifies the stake. Lucky events become
twice as lucky and errors become twice as expensive.
Moreover, cubeful play adds a series of pure skill decisions to
the checker play : should I double ? shoud I double ?
... he doubled, should I take ? These dilute the cubeless
luck-to-skill ratio. Cube diminishes the influence of luck.
On 2021-10-20, MK <mu...@compuplus.net> wrote:
Since a low PR means playing closer to the bot, yes, a low PR
implies more predicatble play.
With that, will you agree that logically the "less predictable" player
can better exploit the "more predictable" player?
Predictability here is somehow a weaker concept than in other
games (typically card games) where parts of the "position" to
evaluate are hidden and Bayesian reasoning (he did this, then
he is more likely to have that) is an important part of trying to
evaluate it.
All I mentionned is that some kind of positions are easier to play
and other are harder, and it is predictable that people make more
errors in the second case. Here, the more predictable the opponent
is, the less there is to exploit.
Moreover, in so-called incomplete information games, unpredictability
is a "defensive" skill, merely making you harder to exploit. Your own unpredictability doesn't help to see through that of the opponent. The
"can better" in your words above doesn't really make sense.
Sysop: | Keyop |
---|---|
Location: | Huddersfield, West Yorkshire, UK |
Users: | 286 |
Nodes: | 16 (2 / 14) |
Uptime: | 85:17:52 |
Calls: | 6,495 |
Calls today: | 6 |
Files: | 12,099 |
Messages: | 5,276,971 |