i,e, A+, a, b, c, d, etc......

just thaught i would get some ideas before i start processing my data (nba, nfl, mlb - value bets).

basically each selection is rated and a value pick chosen as a bet, and given a % stake (the more value the bigger the stake). However i want to try and give that % value a ranking i,e A+, A, B etc.....

i know the easiest method would be to use the following:
A+ = 5%+
A = 4 - 4.9%
B = 3 - 3.9%
C = 2 - 2.9%
D = 1 - 1.9%
E = 0.1 - 0.9%

and stake:

A+ 6%
A 5%
b 4%
c 3%
d 2%
e 1%

however with this method of staking i would have something like (out of around 600 recorded bets):

A+ 1 bet
A 3 bets
B 20 bets
C 60 bets
D 200 bets
E 300+ bets

as you can see the vast majority are way down the scale. (the sub 1% ammount may even be higher than 300). however the sub 1% bet's i am pretty sure make a decent profit. so when my staking says - put 0.0001 % of bank on the selection - it seems like i am missing out. - so i feel band ing the stakes into groups of bet stakes may work out.

this however would lead to basically just D+E bets and very rarely any higher ranked bet.

so another posibility could be, to to split the bandings into 'frequency' ratings, i,e,:

top 5% = A+ bet ....................30 selections
next 10% = A bet....................60 selections
next 15% = B bet...................90 selections
next 20% = C bet....................120 selections
next 20% = D bet.....................120 selections
next 30% = E bet.....................180 selections

this would give a more varied distrobution of bets.
the odd's range of the (top 5% for example) would be worked out from the previous seasons results.

.........................................................

i know that finding the most profitable bands can be worked out by data mining. but i have a problem with this as my biggest stakes 3.9%+ (top example A+-B bets) i have so little to work from (under 25 selections from 600+ bets), - i know that in theory that these 'should' be my most profitable long term but they may even have made a loss in my data - the biggest bet by far - the single 5%+ stake bet lost.

any ideas on how i could find bands for the ratings?