| |
BGonline.org Forums
Rollout - Some thoughts about backgames
Posted By: Keene In Response To: Rollout (John O'Hagan)
Date: Thursday, 1 August 2013, at 10:13 p.m.
In my experience with XG, I often find that it will play best equity plays immediately. It doesnt apply concepts like 'adjust timing' or 'need a longer race' to particular positions. It performs its analysis based on the 3/4/5 ply that will offer the immediate equity results that it desires, however, ultimately (in this position as an example), its not looking 20 plys down the road to see the future shot count, or stronger homeboard that can be made by increasing timing.
This is one of those situations whereby XG is finding optimal immediate (short term) equity plays. I wouldnt be surprised if the more blotful play is more effective in the long run.
As many here, I dont truly trust XG to adjust to move into a backgame at this early position, as it simply wont know that the reducing equity of the player will ultimately result in a series of positions that are practically worthless, and it doesnt have the human capacity to predict an endgame position from here.
In short, the play may be to make the prime in the short term, however, the concept in the long term may be to do what John and Rod were talking about - generate timing.
Undoubtedly the bot will play out these short term positions better than almost all humans, however it wont ever know that its a better concept to adjust timing (if indeed it is here). Perhaps if XG could classify positions, and apply ideas as it pertains to net equity over the endgame positions, then it would be a better bot. Surrendering a chain of equities thats equivalent to say 0.5 now in return for a chain of equities thats worth 0.4 in the future (versus 0.2 in the endgame future of this position).
Numbers above are just random, to illustrate a point.
Thoughts anyone?
Keene
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.