Re: Reverse Psychology System
- From: HC <TheHCapper@xxxxxxx>
- Date: Tue, 25 Sep 2007 10:57:07 -0700
On Sep 25, 1:48?pm, Inventor <Inventor...@xxxxxxxxxxx> wrote:
I tried out a new betting angle and it worked very well for the one
sample case that I tested. Actually I tried exploring several such
angles and they all failed except for this one. It may be old school
but it's new to me and I haven't found such a method on the net in my
searches yet. Here is what I did:
Note: the following idea is based on what I am calling the "ANDing"
method for lack of a better term. It works like Consensus but instead
of choosing the teams that have the most handicapper selections, it
chooses the teams that all handicappers have selected. For example if
I choose (MIA, PHI) and you choose (DAL, PHI) then the ANDing method
will only have (PHI) selected, whereas the Consensus method would have
(MIA, PHI, DAL) selected. In set theory we would call ANDing the
"intersection" of the sets, while Consensus is more like the "union"
of the sets (though not exactly - Consensus is more like voting). I
hope that's clear enough, now the concept...
There is a curious betting angle that I have been testing, which we
might call "Reverse Psychology". I was thinking about the situation
revealed by that quick little test program I wrote where the ANDing
method appears to be valuable only if your handicappers are above 57%
and unfortunately most handicapping sources are below 57%. If that is
the case, how can we use ANDing to improve our success rates? Well,
the flip side of that 57% is 100% - 57% = 43% so that if we have a
bunch of handicappers averaging 43% then their ANDed result will be
much lower than 43% according to the test program. In fact at 43% or
so the ANDed result is only about 30% successful or lower. This is
because the ANDed success rate of several handicappers is shown by the
test program to be approximately a sigmoid function centered at 67%.
I have also noticed that the more handicappers there are, the sharper
the sigmoid transition is, but I'm rambling... back to the point. The
thought then occurred that maybe we can purposely create bad
handicappers who are reliably bad, AND their predictions, and then
invert the ANDed predictions to acheive good handicapping results. In
other words, fade the cappers, AND them, then fade the ANDed result.
The method is as follows: Take your 55% or similar handicapper's
predictions and type them into a text file on one line. Then type in
all of the opponents to those selections on the next line. If your
capper is at 55% then these opponents on the second line will together
be at 45%. Do this for all of your handicapping sources, adding more
lines as you go. Then take the AND of all the inverted capper lists -
that is, on the next line type in the teams that appear in all of the
inverted capper selection lists. The resulting list should be even
worse than 45%, perhaps 30% or so, and it should be a reliable if not
consistent 30%. Now on the final bottom line reverse that list by
typing in all the opponents of the teams on that list. This bottom
line reversed list will have a success rate of 100% - 30%, or 70%. At
least that's what happened in my test.
I opened a text file and typed in my computer prediction list for the
three weeks of the 2007 season that have passed so far, which is at
58% right now, and reversed it by selecting opponents of the picks to
create a 42% list. Then I typed in the public consensus from
wagerline.com for games where 59% or more of the public wagered on a
favored team. If you fade this group, you should get about 55%
success rate according to aricles that I have read on the web
(including one article that had a data table derived from past results
to support that conclusion). If you don't fade the group but rather
go along with it, you should get 100% - 55% = 45% success rate, so
that's what I did. I then ANDed the reverse computer group with the
public consensus group to produce a selection with even worse success
rate, then recorded the opponents of that ANDed list. This final
list, which could be denoted in pseudocode as
"reverse( reverse(computer picks) AND (public picks) )" turned out to
have a record of 7-3-0, or 70%. Now here we have a situation where
two handicapping sources that average about 55% each have been
mathematically combined in such a way as to produce a much higher 70%
result. I also tried it with the wagerline.com list of public
selections that were greater than 50% and that test yielded a record
of 8-5-1, or 62%.
Of course, the usual caveat that I am working with a small data set
applies so it wouldn't make sense to wager on this method just yet.
Also the whole concept relies on the assumption that you can invert a
pick list's success rate by choosing all the opponents of the picks,
but I think that's valid. A lot of future study will be necessary to
prove or disprove the angle, but at least the initial result seems
promising, and besides I'm having fun exploring these ideas so I think
I will continue with it. As always, I encourage your constructive
criticism and comments. What do you think of this wagering angle?
Well, if you're enjoying yourself, more power to you.
But frankly, all you have here is the massaged data from the results
of three weeks of football.
Inventorwww.freedomodds.com
.
- Follow-Ups:
- Re: Reverse Psychology System
- From: Inventor
- Re: Reverse Psychology System
- References:
- Reverse Psychology System
- From: Inventor
- Reverse Psychology System
- Prev by Date: Reverse Psychology System
- Next by Date: Re: Picks / results
- Previous by thread: Reverse Psychology System
- Next by thread: Re: Reverse Psychology System
- Index(es):
Relevant Pages
|
Loading
