[Bug-gnubg] strange result involving reported s.d. in very long rollout

[bob koca]

 I rolled out 27aFAgDbtoUCAA:cAkgAAAAAAAA 1296^2 times at DMP
After 644721 rollouts it was at .610614 win chance and s.d. in win chance of .000015
After 1679616 (about 2.5 times as many games) it was at .610612 win chance and s.d in win chance
of .000020   How would the s.d. get so much larger when it should have gone down to about 5/8 of the .000015 value
to about .000008 or .000009
 
 Is it a statisical fluke or is something going wrong?
 
Bob Koca

---

Chat with Messenger straight from your Hotmail inbox. Check it out