# Joseph_Ward comments on The Prediction Problem: A Variant on Newcomb’s

• Thank you for post­ing this! I’m post­ing here for the first time, al­though I’ve spent a sig­nifi­cant amount of time read­ing the Se­quences already (I just finished See­ing with Fresh Eyes). The com­ments on de­ter­minism cleared up a few un­cer­tain­ties about New­comb’s Prob­lem for me.

When I have ex­plained the prob­lem to oth­ers, I have usu­ally used the phras­ing where Alpha is sig­nifi­cantly bet­ter than av­er­age at pre­dict­ing what you will choose, but not perfect. (This helps re­duce in­cre­dulity on the part of the av­er­age listener.) I have also used the as­sump­tion that Alpha does this by ex­am­in­ing your men­tal state, rather than by draw­ing causal ar­rows back­ward in time. One of my friends sug­gested pre­com­mit­ting to a strat­egy that one-boxes 51% of the time and two-boxes 49% of the time, cho­sen at the time you re­ceive the boxes by some source that is agreed to be ran­dom such as rol­ling two d10′s. His logic is that Alpha would prob­a­bly read your mind ac­cu­rately, and that if he did, he would de­cide based on your men­tal state to put the money in the box, since you are more likely to one-box than not.

This seemed like a very good strat­egy (as­sum­ing the logic and the model of the prob­lem are cor­rect, which is far from cer­tain), and I won­dered why this strat­egy wasn’t at least be­ing dis­cussed more. It seems that most other peo­ple were as­sum­ing de­ter­minism while I was as­sum­ing liber­tar­ian free will.

What do all of you think of my friend’s strat­egy?

Is the as­sump­tion of de­ter­minism a com­ment on the ac­tual state of the uni­verse, or sim­ply a nec­es­sary as­sump­tion to make the prob­lem in­ter­est­ing?

• Well that would work for a pre­dic­tor that 100% pre­dicts your most likely strat­egy. If the pre­dic­tor has even a slight chance of pre­dict­ing the 49% strat­egy in­stead of the 51% strat­egy, you’ll lose out as you’re risk­ing a mil­lion to gain a thou­sand. But yes, the dis­cus­sion in my post as­sumes that the pre­dic­tor can pre­dict any sources of ran­dom­ness that you have ac­cess to.