FindMaximum/NMaximize vs. Excel Solver
- To: mathgroup at smc.vnet.net
- Subject: [mg106714] FindMaximum/NMaximize vs. Excel Solver
- From: John <jhurley13 at gmail.com>
- Date: Thu, 21 Jan 2010 04:57:20 -0500 (EST)
FindMaximum/NMaximize vs. Excel Solver
I am looking at a simple model that calculates the best wagers to
place on a set of 20 horses, where the goal is to maximize reward vs.
risk. Since the bets need to be constrained, I can't use NMaximize.
The probabilities are the "real" probabilities of the horses winning;
odds are what would be available at the track.
I have working versions in Excel and Mathematica, but the Mathematica
implementation runs twice as slowly, and it seems that there must be a
more elegant way to do it. The examples I've seen for FindMaximum use
3 or fewer variables and list them explicitly.
I am mostly interested in understanding if this is the Mathematica way
of doing this problem (as opposed to finding a really fast way to bet
on horses).
solveHorse[]:=Module
[{probabilities,odds,totalWager=100,vars,f,inRange,sol},
probabilities={0.2,0.2,0.2,0.2,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0.1,0.1};
odds={7,5,5,5,5,5,5,5,5,5,5,5,5,5,1,1,2,2,2,2};
inRange[x_]:=0<=x<=totalWager;
f[a__Symbol]:=Module[{ep,var,ws,sp},
ws=List[a];
ep=MapThread[(#1*(1+#2)-totalWager)&,{ws,odds}];
sp= Dot[probabilities,ep];
var=((#-sp)^2)&/@ep;
sp/Sqrt[Dot[probabilities,var]]
];
(* 23.08,25.64,25.64,25.64,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0.0,0.0 <--
Excel answer*)
vars=Table[ToExpression["x"<>ToString[i]],{i,1,20}];
sol=FindMaximum[{f[vars]/.List->Sequence,
(And@@inRange/@vars) && Total[vars]==totalWager },
vars,AccuracyGoal->6,PrecisionGoal->9,MaxIterations->1000];
Print["Final Rtn/SD: ",First[sol]];
If[#<10^-6,0,#]&/@vars/.Last[sol]
];
Timing[solveHorse[]]
Final Rtn/SD: 0.444878
{5.28921,
{23.0767,25.6411,25.6411,25.6411,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0}}