Inaccurate estimates of the Kelly wager?

Questions, suggestion and tips regarding the use of the ProlineParlayMaker program
Post Reply
Slips
Posts: 125
Joined: Sat Mar 07, 2015 10:30 am

Inaccurate estimates of the Kelly wager?

Post by Slips » Sun Oct 11, 2015 10:51 am

I have been noticing some weird numbers lately for the optimal kelly wagers, and I'm concerned that the method used to estimate them isn't very accurate. Which is only fair, as the disclaimer notes: "I want to stress that this is only intended as a quick approximation of the Kelly method to compensate for the overlapping parlays, and should not be regarded as accurate."

An example of the strange numbers I'm seeing...

In NFL point spread at the moment it's recommending a total wager of $526 on 5 and 7 team parlays (5% MinPlayerEdge, $100,000 bankroll, 0.25 bankroll adjust). Of that $526, $185 is 5 team parlays and $341 is 7 team parlays. But if I uncheck 5 team parlays, leaving just the 7 team parlays checked, then it recommends a total wager of only $300! How is the kelly wager on just 7 team parlays by themselves ($300) less than the 7 team parlays when played together with 5 team parlays ($341)? That sure doesn't seem right.

I have also run some kelly calculations using an Excel sheet that I trust is accurate to double check, and it seems that ParlayMaker is often overestimating by 100% or more.

So PLP it sounds like you're already aware of this problem? Could you give some more info on how exactly the wager amounts are estimated, and have you made any progress on a more accurate kelly calculator? There was some discussion back in the day on a different site alluding to the fact that it's actually possible to calculate the exact kelly optimal stakes using some calculus and a recursive algorithm, which would obviously be ideal to eliminate any error from estimating, but I haven't quite wrapped my head around that yet.

Post Reply