Black to play 52 and an example of how shockingly wrong a world class bot can evaluate a position.

Think your copy of XG is broken bro - that two should be a three

I just auto double hit otb in this spot with 23/21* and 6/1* under time pressure but could well be a huge leak. Not sure if sending more checkers back is even a good idea here, so maybe something like 23/18 /16 is better, but am curious what XG does and what the big mistake here is out of the options as I probably play these types of spots pretty badly.

I hit with 23-31-16 bringing another builder to bear on the 10-11 points and hope to make one of those next roll as that seems to be key in this position.

Hitting with 6-1 seems wrong. Even if white anchors there he's going to be moving off it very soon anyway so what's to gain.

I should have posted it as a 53 problem ;-) as the themes for both rolls are pretty much the same.

if you hit on 21 you have to hit on 1 imo

Why?

With either 5-2 or 5-3 I hit twice. I want to get the checkers on the 6-point back into the game, and with White's weak board this looks like a good time.

I'm not seeing it. If you get hit back great, but white is favored not to roll a one and then you've buried a checker on the ace point. And that checker while not ideally placed on the 6-point isn't completely out of the game either as white will be moving off the 4-point soon and making that point ourselves once that happens doesn't seem like a bad idea.

Is it that crucial to recycle a checker that is justifies the risk?

I'm not sure how crucial it is, but it's the way I like to play against backgames. Once my opponent has a couple of points in my home board, I want all my checkers in play.

+1

This is a tactic I learned from one of the games in Bill's book BACKGAMMON FOR SERIOUS PLAYERS.

don't like the idea of another anchor in my home board

Ok first lets look at some evals:

GNUBG has been on the wacky backy. It's absolute equity evals are off the chain plus it seems to vehemently disagree with which play is best compared to XG and Snowie.
Not hitting didn't even make it past it's 0-ply evals in World Class mode.
Maybe GNUBG can redeem itself by its own rollout?
Sadly not. 2354 Game 2-Ply Rollout results:
0.306 23/16
0.218 23/21*/16
0.115 23/21* 6/1*

What does XG rollout say? 2360 games 3-ply
0.339 23/16
0.295 23/21*/16
0.239 23/21* 6/1*



So it appears the not hitting play is correct.
It's easy to overlook non-hitting plays. Hitting (23/21) sends another checker back which is beneficial for white.
Hitting on the ace is wrong as getting hit back could easily backfire and allow white to win more going forward.
Best to simply not hit and keep the timing status quo and hope to bring the position home with white hopefully having to abandon one of his anchors.

So this is conclusive then? Or is it?

I am curious, how reliable are can bot rollouts be when they disagree among each other?

According to these results then, 21-16 with no hit is best and hitting on the ace point is the worst which is ... I don't know what to make of this.

I can see 23-16 without hitting being correct. I preferred hitting cause intuitively I didn't like leaving the opponents 4-point slotted and I didn't have a strong opinion on whether hitting would have a positive or negative effect on timing.

Since Mr. Robertie said he ALWAYS hits on the ace in positions such as this, I figured I guess he's probably right even though I didn't understand why. Then again, in the other problem Bill said stepping up with the back checker was wrong for sure and I agreed with him 100% in that case and yet the bots said that stepping up is correct.

So who is right in this case? Maybe the lesson is to keep an open mind and accept that there is still much for us to learn as far as basic principles go in certain positions.